Democratic Nat'l Comm. v. Reagan

Full title: THE DEMOCRATIC NATIONAL COMMITTEE; DSCC, AKA Democratic Senatorial…

Court: UNITED STATES COURT OF APPEALS FOR THE NINTH CIRCUIT

Date published: Sep 12, 2018

904 F.3d 686 (9th Cir. 2018)

Opinion

No. 18-15845

09-12-2018

The DEMOCRATIC NATIONAL COMMITTEE; DSCC, AKA Democratic Senatorial Campaign Committee; The Arizona Democratic Party, Plaintiffs-Appellants, v. Michele REAGAN, in her official capacity as Secretary of State of Arizona; Mark Brnovich, Attorney General, in his official capacity as Arizona Attorney General, Defendants-Appellees, The Arizona Republican Party; Bill Gates, Councilman; Suzanne Klapp, Councilwoman; Debbie Lesko, Sen.; Tony Rivero, Rep., Intervenor-Defendants-Appellees.

Bruce V. Spiva (argued), Alexander G. Tischenko, Amanda R. Callais, Elisabeth C. Frost, and Marc E. Elias, Perkins Coie LLP, Washington, D.C.; Sarah R. Gonski and Daniel C. Barr, Perkins Coie LLP, Phoenix, Arizona; Joshua L. Kaul, Perkins Coie LLP, Madison, Wisconsin; for Plaintiffs-Appellants. Dominic E. Draye (argued), Joseph E. La Rue, Karen J. Hartman-Tellez, Kara M. Karlson, and Andrew G. Pappas, Office of the Attorney General, Phoenix, Arizona, for Defendants-Appellees. Brett W. Johnson (argued) and Colin P. Ahler, Snell & Wilmer LLP, Phoenix, Arizona, for Intervenor-Defendants-Appellees.

IKUTA, Circuit Judge

Bruce V. Spiva (argued), Alexander G. Tischenko, Amanda R. Callais, Elisabeth C. Frost, and Marc E. Elias, Perkins Coie LLP, Washington, D.C.; Sarah R. Gonski and Daniel C. Barr, Perkins Coie LLP, Phoenix, Arizona; Joshua L. Kaul, Perkins Coie LLP, Madison, Wisconsin; for Plaintiffs-Appellants.

Dominic E. Draye (argued), Joseph E. La Rue, Karen J. Hartman-Tellez, Kara M. Karlson, and Andrew G. Pappas, Office of the Attorney General, Phoenix, Arizona, for Defendants-Appellees.

Brett W. Johnson (argued) and Colin P. Ahler, Snell & Wilmer LLP, Phoenix, Arizona, for Intervenor-Defendants-Appellees.

Before: Sidney R. Thomas, Chief Judge, and Carlos T. Bea and Sandra S. Ikuta, Circuit Judges.

Dissent by Chief Judge Thomas

IKUTA, Circuit Judge:

The Democratic National Committee (DNC) and other appellants¹ sued the state of Arizona,² raising several challenges under the First, Fourteenth and Fifteenth Amendments, and § 2 of the Voting Rights Act of 1965 (VRA), 52 U.S.C. § 10301, against two state election practices: (1) Arizona's longstanding requirement that in-person voters cast their ballots in their assigned precinct, which Arizona enforces by not counting ballots cast in the wrong precinct (referred to by DNC as the out-of-precinct or OOP policy), and (2) H.B. 2023, a recent legislative enactment which precludes most third parties from collecting early ballots from voters. After a lengthy trial involving the testimony of 51 witnesses and over 230 evidentiary exhibits, the district court rejected each of DNC's claims. Democratic Nat'l Comm. v. Reagan , ––– F. Supp.3d ––––, No. CV-16-01065-PHX-DLR, 2018 WL 2191664 (D. Ariz. May 10, 2018).

1 The appellants here (plaintiffs below) are the Democratic National Committee, the Democratic Senatorial Campaign Committee, and the Arizona Democratic Party. For convenience, we refer to the appellants as "DNC."

2 The appellees here (defendants below) are Arizona Secretary of State Michele Reagan, in her official capacity, and Arizona Attorney General Mark Brnovich, in his official capacity. The intervenor-defendants/appellees are the Arizona Republican Party; Debbie Lesko, an Arizona member of the U.S. House of Representatives; Tony Rivero, a member of the Arizona House of Representatives; Bill Gates, a member of the Maricopa County Board of Supervisors; and Suzanne Klapp, a City of Scottsdale Councilwoman and Precinct Committeewoman. For convenience, we refer to the appellees as "Arizona."

In deciding this case, the district court was tasked with making primarily factual determinations. For instance, a First and Fourteenth Amendment challenge to an election rule involves the "intense[ly] factual inquiry" of whether a plaintiff has carried the burden of showing that challenged election laws impose a severe burden on Arizona voters, or a subgroup thereof. Gonzalez v. Arizona , 485 F.3d 1041, 1050 (9th Cir. 2007). A Fifteenth Amendment claim involves the "pure question of fact" of whether the plaintiff has carried the burden of showing that the state legislature enacted the challenged law with a discriminatory intent. Pullman-Standard v. Swint , 456 U.S. 273, 287–88, 102 S.Ct. 1781, 72 L.Ed.2d 66 (1982). And in a VRA challenge, we defer to "the district court's superior fact-finding capabilities," Smith v. Salt River Project Agric. Improvements & Power Dist. , 109 F.3d 586, 591 (9th Cir. 1997), regarding whether the plaintiff has carried the burden of showing that an election practice offers minorities less opportunity "to participate in the political process and to elect representatives of their choice." 52 U.S.C. § 10301(b) ; see also Chisom v. Roemer , 501 U.S. 380, 397, 111 S.Ct. 2354, 115 L.Ed.2d 348 (1991). We must affirm these factual findings unless they are "clearly erroneous." Anderson v. Bessemer City , 470 U.S. 564, 573, 105 S.Ct. 1504, 84 L.Ed.2d 518 (1985).

In its detailed 83-page opinion, the district court found that DNC failed to meet its burden on these critical factual questions. Its analysis on these factual inquiries was thorough and evenhanded, with findings well-supported by the record. Given the district court's extensive factual findings, much of DNC's appeal amounts to a request that we reweigh and reevaluate the evidence in the record. But we may not "duplicate the role of the lower court" or reject factual findings that, as here, are not clearly erroneous. Id . at 573, 105 S.Ct. 1504. Nor did the district court err in identifying and applying the correct legal standard to each of DNC's claims.

Accordingly, we conclude that the district court did not err in holding that H.B. 2023 and the OOP policy did not violate the First and Fourteenth Amendments because they imposed only a minimal burden on voters and were adequately designed to serve Arizona's important regulatory interests. We also conclude that the district court did not err in holding that H.B. 2023 and the OOP policy did not violate § 2 of the VRA. Given the minimal burden imposed by these election practices, DNC failed to show that minority voters were deprived of an equal opportunity to participate in the political process and elect candidates of their choice. Finally, we conclude that the district court did not err in holding that H.B. 2023 did not violate the Fifteenth Amendment, because DNC failed to carry its burden of showing that H.B. 2023 was enacted with discriminatory intent. We reject DNC's urging to toss out the district court's findings, reweigh the facts and reach opposite conclusions. As such, we affirm the district court.

I

The district court's order denying DNC's claims sets forth the facts in detail, Reagan , ––– F.Supp.3d at –––– – ––––, 2018 WL 2191664, at *1–9, so we provide only a brief factual and procedural summary here. The district court's factual findings are discussed in detail as they become relevant to our analysis.

A

We begin by reviewing Arizona's election system. Arizona permits voters to vote either in person on Election Day or by early mail ballot. Id . at ––––, ––––, 2018 WL 2191664, at *7, *12. The vast majority of Arizonans vote by early ballot. For instance, only about 20 percent of the votes in the 2016 general election were cast in person. Id . at ––––, 2018 WL 2191664, at *12.

Most Arizona counties conduct in-person voting through a precinct-based system. Arizona gives each county the responsibility to "establish a convenient number of election precincts in the county and define the boundaries of [those] precincts." Ariz. Rev. Stat. § 16-411(A). Before an election, the County Board of Supervisors (the County's legislative unit) must designate at least one polling place per precinct. Id. § 16-411(B). Arizona law provides some flexibility for counties to combine precincts if each county's board of supervisors makes specific findings. See id. § 16-411(B)(2).

Arizona has long required in-person voters to cast their ballots in their assigned precinct and has enforced this system, since at least 1970, by counting only votes cast in the correct precinct. See Ariz. Rev. Stat. §§ 16-122, 16-135, 16-584 (codified in 1979); 1970 Ariz. Sess. Laws, ch. 151, § 64 (amending Ariz. Rev. Stat. § 16-895); Ariz. Rev. Stat. § 16-102 (1974). If an Arizona voter's name does not appear on the voting register at the polling place on Election Day (either because the voter recently moved or due to inaccuracies in the official records), the voter may vote only by provisional ballot. Ariz. Rev. Stat. §§ 16-122, 16-135, 16-584. Later, the state reviews all provisional ballots and counts those votes cast by voters confirmed to be eligible to vote. Id . §§ 16-135(D), 16-584(D). A provisional ballot cast outside of the voter's correct precinct is not counted. Id . (As mentioned above, DNC refers to Arizona's rejection of improperly cast ballots as Arizona's OOP policy.)

Recently, Arizona has permitted counties to choose between the traditional precinct model and "voting centers," wherein voters from multiple precincts can vote at a single location. Id . § 16-411(B)(4). Each voting center must be equipped to print a specific ballot, correlated to each voter's particular district, that includes all races in which the voter is eligible to vote. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *9. Six rural and sparsely populated counties—Graham, Greenlee, Cochise, Navajo, Yavapai, and Yuma—have adopted the voting center model. Id .

As noted above, most Arizona voters (roughly 80 percent in the 2016 general election) do not vote in person. Arizona law permits "[a]ny qualified elector" to "vote by early ballot." Ariz. Rev. Stat. § 16-541(A).³ Early voting can occur by mail or in person at an on-site early voting location in the 27 days before an election. See id. § 16-542(D). All Arizona counties operate at least one on-site early voting location. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *7. Voters may also return their ballots in person at any polling place without waiting in line, and several counties additionally provide special drop boxes for early ballot submission. Id . Moreover, voters can vote early by mail, either for an individual election or by having their names added to a permanent early voting list. Id . An early ballot is mailed to every person on that list as a matter of course no later than the first day of the early voting period. Ariz. Rev. Stat. § 16-544(F). Voters may return their early ballot by mail at no cost, id . § 16-542(C), but it must be received by 7:00 p.m. on Election Day, id. § 16-548(A).

3 A "qualified elector" is any person at least eighteen years of age on or before the date of the election "who is properly registered to vote." Ariz. Rev. Stat. § 16-121(A).

Since 1992, Arizona has prohibited any person other than the voter from having "possession of that elector's unvoted absentee ballot." See 1991 Ariz. Legis. Serv. Ch. 310, § 22 (S.B. 1390) (West). In 1997, the Arizona legislature expanded that prohibition to prevent any person other than the voter from having possession of any type of unvoted early ballot. See 1997 Ariz. Legis. Serv. Ch. 5, § 18 (S.B. 1003) (West) (codified at Ariz. Rev. Stat. § 16-542(D) ). As explained by the Supreme Court of Arizona, regulations on the distribution of absentee and early ballots advance Arizona's constitutional interest in secret voting, see Ariz. Const. art. VII, § 1, "by setting forth procedural safeguards to prevent undue influence, fraud, ballot tampering, and voter intimidation," Miller v. Picacho Elementary Sch. Dist. No. 33 , 179 Ariz. 178, 180, 877 P.2d 277 (1994) (en banc).

Arizona has long supplemented its protection of the early voting process through the use of penal provisions, as set forth in section 16-1005 of Arizona's statutes. For example, since 1999, "[a]ny person who knowingly marks a voted or unvoted ballot or ballot envelope with the intent to fix an election for that person's own benefit ... is guilty of a class 5 felony." 1999 Ariz. Legis. Serv. Ch. 32, § 12 (S.B. 1227) (codified as amended at Ariz. Rev. Stat. § 16-1005(A) ). And in 2011, Arizona made offering or providing any consideration to acquire a voted or unvoted early ballot a class 5 felony. See 2011 Ariz. Legis. Serv. Ch. 105, § 3 (S.B. 1412) (codified at Ariz. Rev. Stat. § 16-1005(B) ).

Since at least 2002, individuals and groups in Arizona have collected early ballots from voters. While distribution of early ballots had been strictly regulated for decades, see 1997 Ariz. Legis. Serv. Ch. 5, § 18 (S.B. 1003) (West) (codified at Ariz. Rev. Stat. § 16-542(D) ), ballot collection by third parties was not. This changed in 2016, when Arizona revised its early voting process, as defined in section 16-1005, by enacting H.B. 2023 to regulate the collection of early ballots. This law added the following provisions to the existing penalties for persons abusing the early voting process:

H. A person who knowingly collects voted or unvoted early ballots from another person is guilty of a class 6 felony. An election official, a United States postal service worker or any other person who is allowed by law to transmit United States mail is deemed not to have collected an early ballot if the official, worker or other person is engaged in official duties.

I. Subsection H of this section does not apply to:

1. An election held by a special taxing district formed pursuant to title 48 for the purpose of protecting or providing services to agricultural lands or crops and that is authorized to conduct elections pursuant to title 48.

2. A family member, household member or caregiver of the voter. For the purposes of this paragraph:

(a) "Caregiver" means a person who provides medical or health care assistance to the voter in a residence, nursing care institution, hospice facility, assisted living center, assisted living facility, assisted living home, residential care institution, adult day health care facility or adult foster care home.

(b) "Collects" means to gain possession or control of an early ballot.

(c) "Family member" means a person who is related to the voter by blood, marriage, adoption or legal guardianship.

(d) "Household member" means a person who resides at the same residence as the voter.

Ariz. Rev. Stat. § 16-1005(H)–(I).

This amendment to section 16-1005 makes it a felony for third parties to collect early ballots from voters unless the collector falls into one of several exceptions. See id. The prohibition does not apply to election officials acting as such, mail carriers acting as such, any family members, any persons who reside at the same residence as the voter, or caregivers, defined as any person who provides medical or health care assistance to voters in a range of adult residences and facilities. Id. § 16-1005(I)(2). H.B. 2023 does not provide that ballots collected in violation of this statute are disqualified or disregarded in the final election tally.

B

We next turn to the history of this case. In April 2016, DNC and other appellants sued the state of Arizona, challenging H.B. 2023 and Arizona's OOP policy.

In separate motions, DNC sought preliminary injunctions against H.B. 2023 and the OOP policy, respectively. On September 23, 2016, the district court denied the motion to preliminarily enjoin enforcement of H.B. 2023. The district court subsequently denied DNC's motion for a preliminary injunction pending appeal. On October 11, 2016, the district court likewise declined to issue a preliminary injunction with respect to the OOP policy.

DNC appealed both denials. A motions panel denied DNC's request to issue an injunction pending appeal of the district court's ruling on the challenge to H.B. 2023, but the two appeals were expedited and calendared for arguments before a three-judge panel on October 19 and 26, 2016, respectively. The expedited appeals proceeded at a rapid pace. On October 28, 2016, a divided panel affirmed the district court's denial of a preliminary injunction as to H.B. 2023. See Feldman v. Ariz. Sec'y of State's Office (Feldman I ), 840 F.3d 1057 (9th Cir. 2016). The case was called en banc the same day, and on November 2, 2016—after a highly compressed five-day memo exchange and voting period—a majority of the active judges on this court voted to hear the appeal of the district court's denial of a preliminary injunction against H.B. 2023 en banc. Two days later, the en banc panel reconsidered the motions panel's earlier denial of an injunction pending appeal and granted DNC's motion for an injunction pending a resolution of the preliminary injunction appeal. See Feldman v. Ariz. Sec'y of State'sOffice (Feldman III ), 843 F.3d 366 (9th Cir. 2016) (en banc). In so doing, the six-judge majority stated that "we grant the motion for a preliminary injunction pending appeal essentially for the reasons provided in the dissent in [ Feldman I ]." Id. at 367 (citing Feldman I , 840 F.3d at 1085–98 ). The Supreme Court summarily stayed this injunction pending appeal the next day. See Ariz. Sec'y of State's Office v. Feldman , ––– U.S. ––––, 137 S.Ct. 446, 446, 196 L.Ed.2d 326 (2016) (mem.) ("The injunction issued by the United States Court of Appeals for the Ninth Circuit on November 4, 2016, in case No. 16-16698, is stayed pending final disposition of the appeal by that court.").⁴

4 Although Feldman III referenced the dissent in Feldman I , it did not incorporate it nor adopt any specific reasoning from the dissenting opinion, Because Feldman III did not provide a "fully considered appellate ruling on an issue of law," we are guided by our general rule that "decisions at the preliminary injunction phase do not constitute the law of the case." Ranchers Cattlemen Action Legal Fund United Stockgrowers of Am. v. U.S. Dept. of Agric. , 499 F.3d 1108, 1114 (9th Cir. 2007) (first quoting 18 Charles Alan Wright & Arthur R. Miller Federal Practice and Procedure § 4478.5 (2002) ; then citing S. Or. Barter Fair v. Jackson County , 372 F.3d 1128, 1136 (9th Cir. 2004) ). Moreover, the Supreme Court's immediate stay of Feldman III 's injunction pending appeal "undercut[s] [Feldman III 's] theory or reasoning" to a significant extent. Miller v. Gammie , 335 F.3d 889, 900 (9th Cir. 2003) (en banc). Therefore, we conclude that Feldman III 's reference to the dissent in Feldman I does not make that dissent law of the case or of the circuit.

The appeal of the district court's denial of a preliminary injunction as to the OOP policy also proceeded apace. On November 2, 2016, a divided panel affirmed the district court. See Feldman v. Ariz. Sec'y of State's Office (Feldman II ), 842 F.3d 613 (9th Cir. 2016). Two days later a majority of active judges voted to hear the OOP policy appeal en banc, and the en banc panel denied DNC's motion for an injunction pending resolution of the appeal. See Feldman v. Ariz. Sec'yof State's Office , 840 F.3d 1165 (9th Cir. 2016) (mem.) (per curiam) (en banc). As a result of these proceedings, both H.B. 2023 and the OOP policy remained in effect for the November 2016 election. The en banc panel did not reach the merits of DNC's appeal of the district court's denial of the preliminary injunctions against H.B. 2023 and the OOP policy.⁵

5 After the district court rendered its decision on the merits and final judgment, the en banc panel dismissed the interlocutory appeals of the denied preliminary injunctions as moot.

DNC's challenge proceeded in district court. DNC argued that H.B. 2023 imposed undue burdens on the right to vote, in violation of the First and Fourteenth Amendments. DNC also claimed that H.B. 2023 violated § 2 of the VRA because it resulted in a discriminatory burden on voting rights prohibited by that section. Finally, DNC claimed that H.B. 2023 was enacted with discriminatory intent, in violation of the Fifteenth Amendment. DNC raised similar claims that the OOP policy imposed an unconstitutional burden on the right to vote and violated § 2 of the VRA, but did not claim that the OOP policy had a discriminatory purpose.

The district court developed an extensive factual record on all five claims. Over the course of a ten-day bench trial in October 2017, the parties presented live testimony from 7 expert witnesses and 33 lay witnesses, in addition to the testimony of 11 witnesses by deposition. Reagan , ––– F.Supp.3d at –––– – ––––, 2018 WL 2191664, at *2–7. The district court also considered over 230 exhibits admitted into evidence.

Seven months later, on May 10, 2018, the district court issued its amended 83-page findings of fact and conclusions of law, holding that DNC had failed to prove its constitutional and VRA claims. Reagan , ––– F.Supp.3d ––––, 2018 WL 2191664.

DNC timely appealed that same day. Fed. R. App. P. 4(a)(1)(B). It also moved for an injunction pending resolution of its appeal. The en banc panel voted not to exercise jurisdiction over the appeal, and the case was assigned to the original three-judge panel. We granted DNC's motion to expedite the appeal in light of the upcoming 2018 election.⁶

6 We deferred consideration of DNC's motion for an injunction pending appeal. Because we affirm the district court, we now DENY that motion as moot.

II

The district court exercised jurisdiction under 28 U.S.C. § 1331, and we have jurisdiction pursuant to 28 U.S.C. § 1291.

Following a bench trial, we review de novo the district court's conclusions of law and review its findings of fact for clear error. Navajo Nation v. U.S. Forest Serv. , 535 F.3d 1058, 1067 (9th Cir. 2008) (en banc). "The clear error standard is significantly deferential." Cohen v. U.S. Dist. Court , 586 F.3d 703, 708 (9th Cir. 2009). "[T]o be clearly erroneous, a decision must ... strike [a court] as wrong with the force of a five-week old, unrefrigerated dead fish." Ocean Garden, Inc. v. Marktrade Co., Inc. , 953 F.2d 500, 502 (9th Cir. 1991) (quoting Parts and Elec. Motors, Inc. v. Sterling Elec., Inc. , 866 F.2d 228, 233 (7th Cir. 1988) ). "This standard plainly does not entitle a reviewing court to reverse the finding of the trier of fact simply because it is convinced that it would have decided the case differently." Bessemer City , 470 U.S. at 573, 105 S.Ct. 1504. "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." Id . at 573–74, 105 S.Ct. 1504. That is, "[w]here there are two permissible views of the evidence, the factfinder's choice between them cannot be clearly erroneous." Id. at 574, 105 S.Ct. 1504.

III

We first address DNC's challenges to H.B. 2023. DNC argues that (1) H.B. 2023 unduly burdens the right to vote, in violation of the First and Fourteenth Amendments; (2) H.B. 2023 disproportionately impacts minority voters in a manner that violates § 2 of the VRA; and (3) H.B. 2023 was enacted with discriminatory intent, in violation of the Fifteenth Amendment.⁷ We address each claim in turn.

7 DNC does not "specifically and distinctly" argue that H.B. 2023 was enacted with a discriminatory purpose in violation of § 2 of the VRA, and therefore we do not consider this issue. Greenwood v. FAA , 28 F.3d 971, 977 (9th Cir. 1994).

A

We begin with DNC's claim that H.B. 2023 violates Arizona voters' First and Fourteenth Amendment rights.

1

The Constitution vests the States with a "broad power to prescribe the 'Times, Places and Manner of holding Elections for Senators and Representatives.' " Clingman v. Beaver , 544 U.S. 581, 586, 125 S.Ct. 2029, 161 L.Ed.2d 920 (2005) (quoting U.S. Const., art. 1, § 4, cl. 1 ). This power under the Elections Clause to regulate elections for federal offices "is matched by state control over the election process for state offices." Id ."Governments necessarily 'must play an active role in structuring elections,' " Pub. Integrity All., Inc. v. City of Tucson , 836 F.3d 1019, 1024 (9th Cir. 2016) (en banc) (quoting Burdick v. Takushi , 504 U.S. 428, 433, 112 S.Ct. 2059, 119 L.Ed.2d 245 (1992) ), and "as a practical matter, there must be a substantial regulation of elections if they are to be fair and honest and if some sort of order, rather than chaos, is to accompany the democratic processes," Storer v. Brown , 415 U.S. 724, 730, 94 S.Ct. 1274, 39 L.Ed.2d 714 (1974). However, when a state exercises its power and discharges its obligation "[t]o achieve these necessary objectives," the resulting laws "inevitably affect[ ]—at least to some degree—the individual's right to vote and his right to associate with others for political ends." Anderson v. Celebrezze , 460 U.S. 780, 788, 103 S.Ct. 1564, 75 L.Ed.2d 547 (1983).

Because a state has the authority and obligation to manage the election process, "not all election laws impose constitutionally suspect burdens on that right." Short v. Brown , 893 F.3d 671, 676 (9th Cir. 2018). There is no " 'litmus-paper test' that will separate valid from invalid restrictions." Anderson , 460 U.S. at 789, 103 S.Ct. 1564 (quoting Storer , 415 U.S. at 730, 94 S.Ct. 1274 ). Rather, "a more flexible standard applies." Burdick , 504 U.S. at 434, 112 S.Ct. 2059. "A court considering a challenge to a state election law must weigh [1] 'the character and magnitude of the asserted injury to the rights protected by the First and Fourteenth Amendments that the plaintiff seeks to vindicate' against [2] 'the precise interests put forward by the State as justifications for the burden imposed by its rule,' taking into consideration [3] 'the extent to which those interests make it necessary to burden the plaintiff's rights.' " Id. (quoting Anderson , 460 U.S. at 789, 103 S.Ct. 1564 ). This framework is generally referred to as the Anderson / Burdick balancing test.

The first prong of this test, the magnitude of the burden imposed on voters by the election law, "is a factual question on which the plaintiff bears the burden of proof." Democratic Party of Haw. v. Nago , 833 F.3d 1119, 1122–24 (9th Cir. 2016) (citing Cal. Democratic Party v. Jones , 530 U.S. 567, 120 S.Ct. 2402, 147 L.Ed.2d 502 (2000) ); Gonzalez , 485 F.3d at 1050 (noting that whether an election law imposes a severe burden is an "intense[ly] factual inquiry"). In addition to considering the burden on the electorate as a whole, courts may also consider whether the law has a heavier impact on subgroups, Pub. Integrity All. , 836 F.3d at 1025 n.2, but only if the plaintiff adduces evidence sufficient to show the size of the subgroup and quantify how the subgroup's special characteristics makes the election law more burdensome. Thus, Crawford v. Marion County Election Board acknowledged the argument that a voter photo identification (ID) requirement might impose a heavier burden on "homeless persons[,] persons with a religious objection to being photographed," and those "who may have difficulty obtaining a birth certificate," but declined to undertake a subgroup analysis because the evidence was insufficient to show the size of such subgroups or to quantify the additional burden on those voters. 553 U.S. 181, 199, 200–03, 128 S.Ct. 1610, 170 L.Ed.2d 574 (2008). Accordingly, it is an error to consider "the burden that the challenged provisions uniquely place" on a subgroup of voters in the absence of "quantifiable evidence from which an arbiter could gauge the frequency with which this narrow class of voters has been or will become disenfranchised as a result of [those provisions]." Ne. Ohio Coal. for the Homeless v. Husted , 837 F.3d 612, 631 (6th Cir. 2016).

After determining the severity of the burden, the court must then identify the state's justifications for the law, and consider whether those interests make it "necessary to burden the plaintiff's rights." Anderson , 460 U.S. at 789, 103 S.Ct. 1564. As we have emphasized, this inquiry does not necessarily mean that the state is "required to show that its system is narrowly tailored—that is, is the one best tailored to achieve its purposes." Dudum v. Arntz , 640 F.3d 1098, 1114 (9th Cir. 2011). Rather, this step involves a "balancing and means-end fit framework." Ariz. Green Party v. Reagan , 838 F.3d 983, 988 (9th Cir. 2016) (quoting Pub. Integrity All. , 836 F.3d at 1024 ). The severity of the burden dictates the closeness of the fit required, and the more severe the burden, the "more compelling the state's interest must be." Id .

By contrast, "when a state election law provision imposes only 'reasonable, nondiscriminatory restrictions' upon the First and Fourteenth Amendment rights of voters, 'the State's important regulatory interests are generally sufficient to justify' the restrictions." Burdick , 504 U.S. at 434, 112 S.Ct. 2059 (quoting Anderson , 460 U.S. at 788, 103 S.Ct. 1564 ); see also Ariz. Green Party , 838 F.3d at 988. In conducting this analysis, we are particularly deferential when "the challenge is to an electoral system , as opposed to a discrete election rule ." Dudum , 640 F.3d at 1114.

2

Applying the Anderson / Burdick framework, the district court found that H.B. 2023 did not unconstitutionally burden the right to vote. First, the court found that H.B. 2023 posed only a minimal burden on Arizona voters as a whole. Twenty percent of Arizonans voted in person in the prior 2016 general election, and so were wholly unaffected. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *12. As to the 80 percent of Arizonans who voted by mail, the district court noted that there were no records of the number of voters who returned their ballots with the assistance of third parties. Id . After presenting various witnesses on this issue, DNC's counsel's "best estimate of the number of voters affected by H.B. 2023 based on the evidence at trial" was "thousands ... but I don't have a precise number of that." Id . The court found that the evidence suggested that "possibly fewer than 10,000 voters are impacted" out of over 2.3 million voters. Id . Therefore, the vast majority of Arizona voters were unaffected by the law. Id .

Second, the district court found that H.B. 2023 imposed a minimal burden on even the small number of voters who had previously returned ballots with the assistance of third parties. Because "[e]arly voters may return their own ballots, either in person or by mail, or they may entrust a family member, household member, or caregiver to do the same," the burden imposed by H.B. 2023 "is the burden of traveling to a mail box, post office, early ballot drop box, any polling place or vote center (without waiting in line), or an authorized election official's office, either personally or with the assistance of a statutorily authorized proxy, during a 27-day early voting period." Id . Therefore, the court found that H.B. 2023 "does not increase the ordinary burdens traditionally associated with voting." Id .

The district court then considered whether DNC had shown that H.B. 2023 had a more severe impact on particular subgroups of Arizona voters who have some common circumstance that would cause them to face special difficulties in voting without ballot collection services, such as "communities that lack easy access to outgoing mail services; the elderly, homebound, and disabled voters; socioeconomically disadvantaged voters who lack reliable transportation; [and] voters who have trouble finding time to return mail because they work multiple jobs or lack childcare services."⁸ Id . at ––––, 2018 WL 2191664, at *14. The court determined that the plaintiffs had not made such a showing, because there was "insufficient evidence from which to measure the burdens on discrete subsets of voters" or to "quantify with any degree of certainty" how many voters had previously used ballot collection services. Id . Moreover, the district court could not determine the number of those voters who used those services merely "out of convenience or personal preference, as opposed to meaningful hardship," and therefore could not evaluate whether any of them would face a substantial burden in relying on other means of voting offered by Arizona. Id .

8 DNC also identified as a potential subgroup "voters who are unfamiliar with the voting process and therefore do not vote without assistance or tend to miss critical deadlines." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *14. The district court found that remembering relevant deadlines was not a burden on the right to vote, and therefore not a basis for finding a special burden. Id .

Having identified these major gaps in DNC's evidence, the district court evaluated the evidence presented. According to the district court, "the evidence available largely shows that voters who have used ballot collection services in the past have done so out of convenience or personal preference." Id . The court discussed five voters who testified, Nellie Ruiz, Carolyn Glover, Daniel Magos, Carmen Arias, and Marva Gilbreath, explained their individual circumstances and noted that each had successfully returned their ballot except for Gilbreath, who simply forgot to timely mail her ballot.⁹ Id . at ––––, 2018 WL 2191664, at *15. The district court also found that Arizona provides accommodations to subgroups of voters whose special characteristics might lead them to place a greater reliance on ballot collection. Id . at ––––, 2018 WL 2191664, at *14. Specifically, for voters with mobility issues, Arizona requires counties to provide special election boards, which, upon timely request, will deliver a ballot to an ill or disabled voter. Id . While finding that "relatively few voters are aware of this service," the district court pointed out that DNC could educate voters as to its availability. Id . Further, Arizona permits polling places to offer curbside voting, allowing voters to pull up to the curb by a polling place and have an election official assist them at their car. Id . Arizona law also requires employers to give their employees time off to vote in person if an employee is scheduled for an Election Day shift without at least a three-hour window to vote. Id . at ––––, 2018 WL 2191664, at *15. Finally, the district court noted the many exceptions in H.B. 2023, allowing voters to give their early ballots to family members, household members, caregivers, or election officials. Id .

9 The district court expressed "concerns about the credibility" of the deposition testimony of a deceased witness, Victor Vasquez. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *16. "When findings are based on determinations regarding the credibility of witnesses, Rule 52(a) demands even greater deference to the trial court's findings." Bessemer City , 470 U.S. at 575, 105 S.Ct. 1504.

Because the court found that H.B. 2023 imposed only a minimal burden on Arizonans' First and Fourteenth Amendment rights, it held that defendants had to show only that H.B. 2023 served important regulatory interests. As summarized by the district court, Arizona advanced two regulatory interests: (1) "that H.B. 2023 is a prophylactic measure intended to prevent absentee voter fraud by creating a chain of custody for early ballots and minimizing the opportunities for ballot tampering, loss, and destruction"; and (2) "that H.B. 2023 improves and maintains public confidence in election integrity." Id . at ––––, 2018 WL 2191664, at *18. The court found that these interests were important. Id . at ––––, 2018 WL 2191664, at *19.

Turning to a means-end fit, the court found that given the de minimis nature of the burden imposed by H.B. 2023, it did not need to be "the most narrowly tailored provision," so long as it reasonably advanced the state's interests. Id . at ––––, 2018 WL 2191664, at *20. Finding that it did so, the court held that H.B. 2023 did not violate the First and Fourteenth Amendments. Id . at –––– – ––––, 2018 WL 2191664, at *18–20.

3

We conclude that the district court did not err in its Anderson / Burdick analysis. First, the district court's determination that H.B. 2023 imposes only a de minimis burden on Arizona voters was not clearly erroneous. See Crawford , 553 U.S. at 198, 128 S.Ct. 1610 (holding that "the inconvenience" of the process of going to the state Bureau of Motor Vehicles to obtain an ID "does not qualify as a substantial burden on the right to vote, or even represent a significant increase over the usual burdens of voting"). DNC does not directly dispute this conclusion.

Rather, DNC argues that H.B. 2023 imposes severe burdens on subgroups of voters unable to vote without the third-party ballot collection services prohibited by H.B. 2023. This argument fails. The district court did not clearly err in finding that there was "insufficient evidence from which to measure the burdens on discrete subsets of voters," Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *14, which is a threshold requirement to conducting a subgroup analysis. See Crawford , 553 U.S. at 200–03, 128 S.Ct. 1610. The record shows that DNC's witnesses could not specify how many voters would have been unable to vote without ballot collection services. For instance, a Maricopa County Democratic Party organizer, Leah Gillespie, testified that some voters who used ballot collection services told her that they had no other means of voting, but her only example was of a friend whose husband was supposed to deliver her ballot but forgot it at home.¹⁰ Similarly, Arizona State Senator Martin Quezada stated that his campaign received ballot collection requests after H.B. 2023 took effect and had been unable to provide rides to the polling place or other assistance to all such voters. But he did not know "how many of those people had family members who could have turned in their ballot," and could only give his sense "that several of them lacked anybody" who could do so. Moreover, DNC failed "to produce a single voter to testify that H.B. 2023's limitations on who may collect an early mail ballot would make voting significantly more difficult for her." Only one voter (Marva Gilbreath) testified that she did not vote in the 2016 general election, because she "was in the process of moving," had no mailbox key due to "misunderstandings with the realtor and things like that," and "didn't know where the voting place was." This witness's highly idiosyncratic circumstances do not indicate that H.B. 2023 imposes a severe burden on an identifiable subgroup of voters. Rather, burdens "arising from life's vagaries are neither so serious nor so frequent as to raise any question about the constitutionality of [the challenged law]." Id . at 197, 128 S.Ct. 1610.

10 Of course, had the husband not forgot, but had delivered the vote, there would have been no violation of H.B. 2023, which exempts family members. Ariz. Rev. Stat. § 16-1005(H)–(I).

In sum, DNC's evidence falls far short of the necessary "quantifiable evidence from which an arbiter could gauge the frequency with which this narrow class of voters has been or will become disenfranchised as a result of [H.B. 2023]." Ne. Ohio Coal. , 837 F.3d at 631 ; cf. Crawford , 553 U.S. at 201–02, 128 S.Ct. 1610 (declining to conduct a subgroup analysis despite evidence of one indigent voter who could not (or would not) pay for a birth certificate and one homeless woman who was denied a photo ID card because she lacked an address.).

The dissent disagrees, but its disagreement here—as with the district court's opinion generally—is based on throwing out the district court's factual findings, reweighing the evidence, and reaching its own factual conclusions. This approach is not only contrary to the most basic principles of appellate review, but is an approach that the Supreme Court has frequently warned us to avoid. See Bessemer City , 470 U.S. at 574–75, 105 S.Ct. 1504 (holding that the rationale for deference to the trial court's finding of fact is based not only on "the superiority of the trial judge's position to make determinations of credibility," but also on the judge's expertise in determination of fact, and ensuring that "the trial on the merits should be 'the main event ... rather than a tryout on the road' ") (quoting Wainwright v. Sykes , 433 U.S. 72, 90, 97 S.Ct. 2497, 53 L.Ed.2d 594 (1977) ).

Here, for instance, the dissent seeks to revisit the district court's conclusion that DNC failed to carry its burden of showing that H.B. 2023 imposed a heavy burden on Native Americans. Dissent at 753–54. Conducting its own factual evaluation, the dissent claims that H.B. 2023 imposes a heavy burden on Native Americans because a majority of them lack home mail service. Dissent at 753–54. The dissent then speculates that many Native Americans may have trouble getting to post offices, and may have different family relationships than are indicated in H.B. 2023. Dissent at 702. Of course, the dissent's determination that "it would have decided the case differently" does not make the district court's findings clearly erroneous. Bessemer City , 470 U.S. at 573, 105 S.Ct. 1504. Indeed, even evidence that third-party ballot collection is more useful to Native Americans than to other voters does not compel the conclusion that H.B. 2023 imposes a heavy burden on Native Americans' ability to vote. Most tellingly, the dissent does not meaningfully address the district court's most notable factual finding: that not a single voter testified at trial that H.B. 2023's limitations would make voting significantly more difficult. Although the dissent insists that there was evidence to this effect, Dissent at 753–54, it cites only to the testimony of a third-party ballot collector who conceded that his organization had not attempted to determine whether the voters they served could have returned their ballots some other way. There is thus no basis for holding that the district court's findings were clearly erroneous, and the dissent errs in arguing otherwise.

The dissent also faults the district court's decision not to conduct a subgroup analysis because it "could not determine a precise number of voters that had relied on ballot collection in the past or predict a likely number in the future." Dissent at 753. According to the dissent, this decision was based on a misunderstanding of Crawford , and therefore constitutes legal error. We disagree. The district court correctly relied on Crawford in concluding that "on the basis of the evidence in the record it [was] not possible to quantify either the magnitude of the burden on this narrow class of voters or the portion of the burden imposed on them that [was] fully justified." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *14 (quoting Crawford , 553 U.S. at 200, 128 S.Ct. 1610 ). Accordingly, the court properly held that DNC did not carry its burden of showing the existence of a relevant subgroup.

Nor did the district court clearly err in finding that any burden imposed by H.B. 2023 was further minimized by Arizona's many accommodations available for those subgroups of voters that DNC claims are burdened by H.B. 2023.¹¹ Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *14. For instance, the district court reasonably found that the subgroup of voters who are "confined as the result of a continuing illness or physical disability," Ariz. Rev. Stat. § 16-549(C), could request ballots from special election boards, and the burden of doing so was minimal, see Short , 893 F.3d at 677 ("To the extent that having to register to receive a mailed ballot could be viewed as a burden, it is an extremely small one, and certainly not one that demands serious constitutional scrutiny."). The district court did not clearly err in finding that it was irrelevant whether voters were widely aware of this alternative, as nothing prevented DNC from informing voters of and facilitating this procedure. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *14.

11 Given that DNC did not meet its burden of showing how large the subgroup of specially burdened voters might be, see Democratic Party of Haw. , 833 F.3d at 1122–24, its unsupported claims that Arizona's many accommodations cannot adequately serve an unquantified number of voters are unpersuasive.

We conclude that the district court did not clearly err in finding that DNC had failed both to quantify the subgroups purportedly burdened by H.B. 2023 and to show that Arizona's alternatives did not ameliorate any burden on them. Accordingly, there was no clear error in the district court's finding that H.B. 2023 imposed only a minimal burden.

4

Next, DNC and the dissent contend that the district court clearly erred in finding that H.B. 2023 serves Arizona's important regulatory interests because Arizona did not adduce any direct evidence of voter fraud. We reject this argument.

DNC does not dispute—nor could it—that Arizona's interest in "a prophylactic measure intended to prevent absentee voter fraud" and to maintain public confidence are facially important. Id . at ––––, 2018 WL 2191664, at *18 ; see Crawford , 553 U.S. at 196, 128 S.Ct. 1610 ("There is no question about the legitimacy or importance of the State's interest in counting only the votes of eligible voters."); Purcell v. Gonzalez , 549 U.S. 1, 4, 127 S.Ct. 5, 166 L.Ed.2d 1 (2006) (explaining that "[c]onfidence in the integrity of our electoral processes is essential to the functioning of our participatory democracy" and noting "the State's compelling interest in preventing voter fraud").

Further, a state "need not show specific local evidence of fraud in order to justify preventive measures," Voting for Am., Inc. v. Steen , 732 F.3d 382, 394 (5th Cir. 2013), nor is such evidence required to uphold a law that imposes minimal burdens under the Anderson / Burdick framework, see Munro v. Socialist Workers Party , 479 U.S. 189, 195, 107 S.Ct. 533, 93 L.Ed.2d 499 (1986) (explaining that legislatures are "permitted to respond to potential deficiencies in the electoral process with foresight rather than reactively"). For example, in Crawford , the challenged law addressed only in-person voter fraud, and "[t]he record contain[ed] no evidence of any such fraud actually occurring in Indiana at any time in its history." 553 U.S. at 194, 128 S.Ct. 1610. Yet the controlling opinion concluded that the law served Indiana's interests in preventing fraud, citing evidence of in-person and absentee voter fraud in other jurisdictions and in historical examples. Id. at 195–96 & nn. 11–13, 128 S.Ct. 1610. Accordingly, H.B. 2023 serves Arizona's important interest in preventing voter fraud even without direct evidence of ballot collection voter fraud in Arizona.¹²

12 DNC's reliance on a vacated Sixth Circuit opinion is unpersuasive. See Ohio State Conference of the NAACP v. Husted , 768 F.3d 524 (6th Cir. 2014), vacated , No. 14-3877, 2014 WL 10384647 (6th Cir. Oct. 1, 2014). The Sixth Circuit has explained that any persuasive value in Ohio State Conference 's analysis of this point is limited to cases involving "significant although not severe" burdens, Ohio Democratic Party v. Husted , 834 F.3d 620, 635 (6th Cir. 2016) (quoting Ohio State Conference , 768 F.3d at 539 ), and not those involving "minimal" burdens, id. (explaining that the district court's reliance on Ohio State Conference was "not sound").

The dissent proposes several meritless distinctions between H.B. 2023 and the voter I.D. law in Crawford . First, the dissent argues that unlike H.B. 2023, Crawford 's voter I.D. law was "tied to 'the state's interest in counting only the votes of eligible voters.' " Dissent at 754 (quoting Crawford , 553 U.S. at 196, 128 S.Ct. 1610 ). But H.B. 2023's regulation of third-party ballot collectors is likewise tied to the state's interest in ensuring the integrity of the vote. As explained by the district court, Arizona could reasonably conclude that H.B. 2023 reduced "opportunities for early ballots to be lost or destroyed" by limiting the possession of early ballots to "presumptively trustworthy proxies," and also lessened the potential for pressure or intimidation of voters, and other forms of fraud and abuse. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *20 ; see infra at 709–10. Second the dissent argues that Crawford is distinguishable because the legislature in that case was motivated in-part by "legitimate concerns," while here the Arizona legislature was "motivated by discriminatory intent," or by solely partisan interests. Dissent at 754. Again, we reject the dissent's factual findings because the district court found that the legislature was not motivated by discriminatory intent and only partially motivated by partisan considerations, and these findings are not clearly erroneous. Moreover, a legislature may act on partisan considerations without violating the constitution. See infra at 720–21.

Similarly, a court can reasonably conclude that a challenged law serves the state's interest in maintaining "public confidence in the integrity of the electoral process," even in the absence of any evidence that the public's confidence had been undermined. Crawford , 553 U.S. at 197, 128 S.Ct. 1610. As several other circuits have recognized, it is "practically self-evidently true" that implementing a measure designed to prevent voter fraud would instill public confidence. Ohio Democratic Party v. Husted , 834 F.3d 620, 633 (6th Cir. 2016) (citing Crawford , 553 U.S. at 197, 128 S.Ct. 1610 ); see Frank v. Walker , 768 F.3d 744, 750 (7th Cir. 2014) (noting that Crawford took "as almost self-evidently true" the relationship between a measure taken to prevent voter fraud and promoting voter confidence). The district court did not clearly err in finding that H.B. 2023 also serves this important state interest.

5

DNC next argues that Arizona could have used less burdensome means to pursue its regulatory interests and H.B. 2023 could have been designed more effectively. This argument also fails. Burdick expressly declined to require that restrictions imposing minimal burdens on voters' rights be narrowly tailored. See 504 U.S. at 433, 112 S.Ct. 2059. Consistent with Burdick , we upheld an election restriction that furthered the interest of "ensuring local representation by and geographic diversity among elected officials" even though less-restrictive means could have achieved the same purposes. Pub. Integrity All. , 836 F.3d at 1028. Similarly, in Arizona Green Party , we rejected the argument that the state must adopt a system of voting deadlines "that is the most efficient possible," in light of the "de minimis burden" imposed by the existing deadlines. 838 F.3d at 992 (citation omitted).

Here, the district court found that H.B. 2023 imposed a minimal burden, and that it was a reasonable means for advancing the state's interests. It concluded that "[b]y limiting who may possess another's early ballot, H.B. 2023 reasonably reduces opportunities for early ballots to be lost or destroyed." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *20. The district court also observed that H.B. 2023 "closely follows," id ., the recommendation of a bipartisan national commission on election reform to "reduce the risks of fraud and abuse in absentee voting by prohibiting 'third-party' organizations, candidates, and political party activists from handling absentee ballots," id . (quoting Building Confidence in U.S. Elections § 5.2 (Sept. 2005) ).¹³ These findings were sufficient to justify the minimal burden imposed by H.B. 2023. DNC's reliance on Common Cause Indiana v. Individual Members of the Indiana Election , 800 F.3d 913, 928 (7th Cir. 2015) as requiring a closer means-ends fit is misplaced. As the Seventh Circuit concluded, the election law in that case imposed a severe burden on the right to vote, and therefore it was appropriate to apply strict scrutiny. Id . at 927.

13 The district court took judicial notice of the report of the Commission on Federal Election Reform chaired by former President Jimmy Carter and former Secretary of State James A. Baker III. Reagan , ––– F.Supp.3d at –––– n.12, 2018 WL 2191664, at *20 n.12. The district court noted that the report was cited favorably in Crawford , which remarked that "[t]he historical perceptions of the Carter-Baker Report can largely be confirmed." 553 U.S. at 194 n.10, 128 S.Ct. 1610. The relevant portion of the report provides:

Fraud occurs in several ways. Absentee ballots remain the largest source of potential voter fraud.... Absentee balloting is vulnerable to abuse in several ways: ... Citizens who vote at home, at nursing homes, at the workplace, or in church are more susceptible to pressure, overt and subtle, or to intimidation. Vote buying schemes are far more difficult to detect when citizens vote by mail. States therefore should reduce the risks of fraud and abuse in absentee voting by prohibiting "third-party" organizations, candidates, and political party activists from handling absentee ballots.

Building Confidence in U.S. Elections § 5.2 (Sept. 2005), https://www.eac.gov/assets/1/6/Exhibit%20M.PDF. The district court did not abuse its discretion in taking judicial notice of the report publicly available on the website of the U.S. Election Assistance Commission. See Anderson v. Holder , 673 F.3d 1089, 1094 n.1 (9th Cir. 2012) ("We may take judicial notice of records and reports of administrative bodies.") (internal quotation marks and citation omitted). There is no dispute as to the report's authenticity or that it contained the cited recommendation, and DNC was not unfairly surprised, given that counsel indicated at trial that he was well acquainted with it and its contents.

We therefore affirm the district court's conclusion that DNC did not succeed on its Anderson / Burdick claim as to H.B. 2023.

B

We next consider DNC's claim that H.B. 2023 violates § 2 of the VRA. We begin by providing some necessary legal background.

1

"Inspired to action by the civil rights movement," Congress enacted the Voting Rights Act of 1965 to improve enforcement of the Fifteenth Amendment.¹⁴ Shelby County v. Holder , 570 U.S. 529, 536, 133 S.Ct. 2612, 186 L.Ed.2d 651 (2013). Section 2 of the Act forbade all states from enacting any "standard, practice, or procedure ... imposed or applied ... to deny or abridge the right of any citizen of the United States to vote on account of race or color." Id. (quoting Voting Rights Act of 1965, § 2, 79 Stat. 437). Section 5 of the Act prevented states from making certain changes in voting procedures unless the states obtained "preclearance" for those changes, meaning they were approved by either the Attorney General or a court of three judges. Id. at 537, 133 S.Ct. 2612."At the time of the passage of the Voting Rights Act of 1965, § 2, unlike other provisions of the Act, did not provoke significant debate in Congress because it was viewed largely as a restatement of the Fifteenth Amendment." Chisom , 501 U.S. at 392, 111 S.Ct. 2354. In 1980, black residents of Mobile, Alabama challenged the city's at-large method of electing its commissioners on the ground that it unfairly diluted their voting strength. City of Mobile v. Bolden , 446 U.S. 55, 58, 100 S.Ct. 1519, 64 L.Ed.2d 47 (1980). A plurality of the Supreme Court held that the electoral system did not violate § 2 of the VRA because there was no showing of "purposefully discriminatory denial or abridgment by government of the freedom to vote 'on account of race, color or previous conditions of servitude.' " Id. at 65, 100 S.Ct. 1519.

14 The Fifteenth Amendment provides that "[t]he right of citizens of the United States to vote shall not be denied or abridged by the United States or by any State on account of race, color, or previous condition of servitude," and authorizes Congress to enforce the provision "by appropriate legislation." U.S. Const. amend. XV.

In response to Bolden , "Congress substantially revised § 2 to make clear that a violation could be proved by showing discriminatory effect alone." Thornburg v. Gingles , 478 U.S. 30, 35, 106 S.Ct. 2752, 92 L.Ed.2d 25 (1986). In order to show actionable discriminatory effect, Congress enacted the "results test," applied by the Supreme Court in White v. Regester , 412 U.S. 755, 93 S.Ct. 2332, 37 L.Ed.2d 314 (1973), see Gingles , 478 U.S. at 35, 106 S.Ct. 2752, namely "whether the political processes are equally open to minority voters." S. Rep. No. 97-417, at 2 (1982), as reprinted in 1982 U.S.C.C.A.N. 177, 205.

As amended, § 2 of the VRA provides:

§ 10301. Denial or abridgement of right to vote on account of race or color through voting qualifications or prerequisites; establishment of violation

(a) No voting qualification or prerequisite to voting or standard, practice, or procedure shall be imposed or applied by any State or political subdivision 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, or in contravention of the guarantees set forth in section 10303(f)(2) of this title, as provided in subsection (b).

(b) A violation of subsection (a) is established 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 class of citizens protected by subsection (a) in that its members have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice....

52 U.S.C. § 10301.

Thus, § 2(a) prohibits a state or political subdivision from adopting a practice that "results in a denial or abridgement" of any U.S. citizen's right to vote on account of race, color, or membership in a language minority group, "as provided in subsection (b)." Id. § 10301(a). Subsection (b), in turn, provides that a plaintiff can establish a violation of § 2(a) if "based on the totality of circumstances," the members of a protected class identified in § 2(a) "have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice." Id. § 10301(b).

Thornburg v. Gingles further clarified that in analyzing whether a state practice violates § 2, a court must engage in a two-step process. First, the court must ask the key question set forth in § 2(b), whether "as a result of the challenged practice or structure plaintiffs do not have an equal opportunity to participate in the political processes and to elect candidates of their choice." 478 U.S. at 44, 106 S.Ct. 2752 (quoting S. Rep. No. 97-417, at 28). Second, a court must assess the impact of the practice on such electoral opportunities in light of the factors set forth in the Senate Report, which accompanied the 1982 amendments and "elaborates on the nature of § 2 violations and on the proof required to establish these violations." Id. at 43–44, 106 S.Ct. 2752.¹⁵

15 As explained in Gingles , the Senate Factors include the extent of any history of official discrimination, the use of election practices or structures that could enhance the opportunity for discrimination, the extent to which voting is racially polarized, and the extent to which minorities bear the effects of discrimination in education, employment and health. 478 U.S. at 36–37, 106 S.Ct. 2752. The factors are not exclusive, and "the question whether the political processes are equally open depends upon a searching practical evaluation of the past and present reality, and on a functional view of the political process."Id . at 45, 106 S.Ct. 2752 (quoting S. Rep. No. 97-417, at 30 (1982), as reprinted in 1982 U.S.C.C.A.N. 177, 208). Because the "essence of a § 2 claim is that a certain electoral law, practice, or structure interacts with social and historical conditions to cause an inequality in the opportunities enjoyed by black and white voters to elect their preferred representatives," 478 U.S. at 47, 106 S.Ct. 2752, if a court determines that a challenged practice does not cause unequal opportunities, it need not consider the practice's interaction with the Senate Factors. Because we affirm the district court's finding that DNC failed to carry its burden of satisfying step one of the § 2 analysis for either H.B. 2023 or the OOP policy, we do not review in detail its factual findings that DNC also failed to carry its burdens at step two.

In the wake of Gingles , some lower courts interpreted the key question set forth in § 2(b) (whether as a result of the challenged practice plaintiffs do not have an equal opportunity to participate in the political process and to elect candidates of their choice) as "provid[ing] two distinct types of protection for minority voters." Chisom , 501 U.S. at 396, 111 S.Ct. 2354 (citing League of United Latin Am. Citizens Council No. 4434 v. Clements , 914 F.2d 620, 625 (5th Cir. 1990) (en banc) ). These courts held that a "vote denial" claim, meaning a claim that a particular state election practice denied or abridged a minority group's right to vote, turned on whether members of that protected class had "less opportunity ... to participate in the political process." By contrast, a "vote dilution" claim, meaning a claim that a state election practice diluted the effectiveness of a minority group's votes, turned on whether those members had "less opportunity ... to elect representatives of their choice." Id. at 388, 395–96, 111 S.Ct. 2354 (citing Clements , 914 F.2d at 625 ).

The Supreme Court flatly rejected this interpretation. In Chisom , the Supreme Court explained that § 2(b) "does not create two separate and distinct rights." Id. at 397, 111 S.Ct. 2354. The Court reasoned that if members of a protected class established that a challenged practice abridged their opportunity to participate in the political process, it would be relatively easy to show they were also unable to elect representatives of their choice, because "[a]ny abridgment of the opportunity of members of a protected class to participate in the political process inevitably impairs their ability to influence the outcome of an election." Id . By contrast, evidence that members of a protected class are unable to elect representatives of their choice does not necessarily prove they had less opportunity to participate in the political process. Id. Accordingly, the Court concluded that the two-pronged results test required by the 1982 amendment "is applicable to all claims arising under § 2," and "all such claims must allege an abridgment of the opportunity to participate in the political process and to elect representatives of one's choice." Id. at 398, 111 S.Ct. 2354 ; see also Ortiz v. City of Phila. Office of City Comm'rs Voter Registration Div. , 28 F.3d 306, 314 (3d Cir. 1994) ("Section 2 plaintiffs must demonstrate that they had less opportunity both (1) to participate in the political process, and (2) to elect representatives of their choice." (emphasis added) (citing Chisom , 501 U.S. at 397, 111 S.Ct. 2354 ) ).

In reaching this conclusion, the Chisom majority rejected Justice Scalia's argument in dissent that requiring a plaintiff to prove both less opportunity to participate and less opportunity to elect representatives would prevent small numbers of voters from bringing a § 2 claim. According to Justice Scalia, the Court should have read "and" in § 2(b) to mean "or," so that if "a county permitted voter registration for only three hours one day a week, and that made it more difficult for blacks to register than whites, blacks would have less opportunity 'to participate in the political process' than whites, and § 2 would therefore be violated—even if the number of potential black voters was so small that they would on no hypothesis be able to elect their own candidate." Chisom , 501 U.S. at 408, 111 S.Ct. 2354 (Scalia, J., dissenting). The majority rejected this argument, however, stating that it had "no authority to divide a unitary claim created by Congress." Id. at 398, 111 S.Ct. 2354.¹⁶

16 The majority also rejected Justice Scalia's "erroneous assumption that a small group of voters can never influence the outcome of an election," Chisom , 501 U.S. at 397 n.24, 111 S.Ct. 2354, although it did not explain what evidence would be necessary to establish that an election practice that affected only a small group of voters deprived minorities of an equal opportunity to elect candidates of their choice.

In light of Chisom , plaintiffs cannot establish a § 2 violation without showing that an electoral practice actually gives minorities less opportunity to elect representatives of their choice. This requires plaintiffs to show that the state election practice has some material effect on elections and their outcomes. As Gingles explained, "[i]t is obvious that unless minority group members experience substantial difficulty electing representatives of their choice, they cannot prove that a challenged electoral mechanism impairs their ability 'to elect.' " 478 U.S. at 48 n.15, 106 S.Ct. 2752 (quoting 52 U.S.C. § 10301(b) ). It is "the usual predictability of the majority's success" which distinguishes a structural problem "from the mere loss of an occasional election." Id. at 51, 106 S.Ct. 2752. If an election practice would generally "not impede the ability of minority voters to elect representatives of their choice" there is no § 2 violation; rather a "bloc voting majority must usually be able to defeat candidates supported by a politically cohesive, geographically insular minority group." Id . at 48–49, 106 S.Ct. 2752.

In a § 2 challenge, a court's focus must be on the question whether minorities have less opportunity to elect representatives of their choice; therefore, evidence that a particular election practice falls more heavily on minority than non-minority voters, or that electoral outcomes are not proportionate to the numbers of minorities in the population,¹⁷ is not sufficient by itself to establish a § 2 violation. As we have previously explained, "a bare statistical showing of disproportionate impact on a racial minority does not satisfy the § 2 'results' inquiry." Salt River , 109 F.3d at 595. Rather, "plaintiffs must show a causal connection between the challenged voting practice and [a] prohibited discriminatory result," i.e., less opportunity to participate in the political process and elect representatives. Id. (quoting Ortiz , 28 F.3d at 312 ). Because "[n]o state has exactly equal registration rates, exactly equal turnout rates, and so on, at every stage of its voting system," Frank , 768 F.3d at 754, were it enough to merely point to "some relevant statistical disparity" implicated by the challenged law, Salt River , 109 F.3d at 595, then § 2 would "dismantle every state's voting apparatus," Frank , 768 F.3d at 754.¹⁸

17 The VRA itself states that "nothing in this section establishes a right to have members of a protected class elected in numbers equal to their proportion in the population." 52 U.S.C. § 10301(b).

18 Directly contrary to this longstanding precedent, the dissent insists that if a challenged practice disproportionately impacts members of a protected class, then it per se constitutes a violation under the first step of the § 2 test. See Dissent at 734–35 (arguing that because DNC showed that minorities are over-represented among those who cast out-of precinct ballots, "[t]he analysis at step one of the § 2 results test ought to end at this point"); id . at 735 (asserting that the district court's finding that "OOP ballot rejection has no meaningfully disparate impact on the opportunities of minority voters to elect their preferred representatives" is "irrelevant to step one of § 2's results test, which focuses solely on the differences in opportunity and effect enjoyed by groups of voters"); id . at 735 (arguing that under § 2, a state must correct any disparities that can be attributed to socioeconomic factors); id . at 751–52 (arguing that because H.B. 2023 imposes a disparate burden on members of protected classes, it meets step one). The dissent's argument is not only contrary to our precedent, but is inconsistent with the plain language of § 2, and to the Supreme Court's interpretation of the VRA. Gingles , 478 U.S. at 51, 106 S.Ct. 2752 (§ 2 plaintiffs must show more than "the mere loss of an occasional election"); Chisom , 501 U.S. at 398, 111 S.Ct. 2354 ("For all such [§ 2] claims must allege an abridgement of the opportunity to participate in the political process and to elect representatives of one's choice.").

If a challenged election practice is not burdensome or the state offers easily accessible alternative means of voting, a court can reasonably conclude that the law does not impair any particular group's opportunity to "influence the outcome of an election," Chisom , 501 U.S. at 397 n.24, 111 S.Ct. 2354, even if the practice has a disproportionate impact on minority voters. For instance, in Lee v. Virginia State Board of Elections , plaintiffs argued that Virginia's photo ID law violated § 2 because more minorities than non-minorities lacked the necessary IDs, and "the process of obtaining photo IDs requires those voters to spend time traveling to and from a registrar's office." 843 F.3d 592, 600 (4th Cir. 2016). The Fourth Circuit rejected this argument. Observing that the state provided the option for voters without ID to cast a provisional ballot and obtain a free ID to verify their identity, the Fourth Circuit reasoned that "every registered voter in Virginia has the full ability to vote when election day arrives," and therefore the election practice "does not diminish the right of any member of the protected class to have an equal opportunity to participate in the political process." Id .

In sum, in considering a § 2 claim, a court must consider whether the challenged standard, practice, or procedure gives members of a protected class less opportunity than others both "to participate in the political process and to elect representatives of their choice." Chisom , 501 U.S. at 397, 111 S.Ct. 2354 (quoting 52 U.S.C. § 10301(b) ). The plaintiff must show a causal connection between the challenged voting practice and the lessened opportunity of the protected class to participate and elect representatives; it is not enough that the burden of the challenged practice falls more heavily on minority voters. See Salt River , 109 F.3d at 595. Rather, the challenged practice must "influence the outcome of an election," Chisom , 501 U.S. at 397 n.24, 111 S.Ct. 2354, and create some "substantial difficulty" for a protected class to elect representatives of its choice, not just the "mere loss of an occasional election." Gingles , 478 U.S. at 48 n.15, 51, 106 S.Ct. 2752. If this sort of discriminatory result is found, then the practice must be considered in light of the Senate Factors, which are "particularly" pertinent to vote dilution claims, but "will often be pertinent" to other § 2 claims as well. Id. at 44–45, 106 S.Ct. 2752.¹⁹

19 Our two-step analysis, derived from the language of § 2, and Supreme Court precedent, is consistent with the two-step framework adopted by the Fourth, Fifth, and Sixth Circuits (and, in part, the Seventh Circuit):

[1] [T]he challenged standard, practice, or procedure must impose a discriminatory burden on members of a protected class, meaning that members of the protected class have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice, [and]

[2] [T]hat burden must in part be caused by or linked to social and historical conditions that have or currently produce discrimination against members of the protected class.

League of Women Voters of N.C. v. North Carolina , 769 F.3d 224, 240 (4th Cir. 2014) (citations and internal quotation marks omitted); Veasey v. Abbott (Veasey I ), 830 F.3d 216, 244 (5th Cir. 2016) ; Ohio Democratic Party , 834 F.3d at 637 ; Frank , 768 F.3d at 754–55 (adopting the test "for the sake of argument"). The first prong tracks the language of § 2, as interpreted by the Supreme Court, and the second prong implicates the Senate Factors.

2

We now turn to the district court's determination here. We review the district court's legal determinations de novo, Gonzalez v. Arizona , 677 F.3d 383, 406 (9th Cir. 2012), but defer to "the district court's superior fact-finding capabilities," and review its factual findings for clear error, Salt River , 109 F.3d at 591.

In analyzing the first step of a § 2 claim, the district court first found that DNC had provided no quantitative or statistical evidence showing how many people would be affected by H.B. 2023 and their minority status, noting that it was "aware of no vote denial case in which a § 2 violation has been found without quantitative evidence measuring the alleged disparate impact of a challenged law on minority voters." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *30. Despite the lack of any statistical evidence establishing a disproportionate impact of H.B. 2023 on minorities, the court stated that it would not rule against DNC on this ground. Id . at ––––, 2018 WL 2191664, at *31. Instead, it considered DNC's circumstantial and anecdotal evidence, and tentatively concluded that "prior to H.B. 2023's enactment minorities generically were more likely than nonminorities to return their early ballots with the assistance of third parties," emphasizing the caveat that it could not "speak in more specific or precise terms than 'more' or 'less.' " Id . at ––––, 2018 WL 2191664, at *33.

Having inferred, based on DNC's circumstantial and anecdotal evidence, that H.B. 2023 likely impacted more minority voters than non-minority voters, the district court nevertheless concluded that DNC's evidence did not establish that H.B. 2023 gave members of a protected class less opportunity than other members of the electorate both to participate in the political process and to elect representatives of their choice. Id . at –––– – ––––, 2018 WL 2191664, at *32–34. The district court provided two reasons. First, the court reasoned that the evidence presented indicated that only "a relatively small number of voters" used ballot collection services at all. Id . at ––––, 2018 WL 2191664, at *33. By logical extension, that meant that only a small number of minorities used ballot collection services to vote, and the vast majority of minority voters "vote without the assistance of third-parties who would not fall within H.B. 2023's exceptions." Id . Because only a small number of minority voters were affected to any degree by H.B. 2023, the court found "it is unlikely that H.B. 2023's limitations on who may collect an early ballot cause a meaningful inequality in the electoral opportunities of minorities as compared to nonminorities." Id .

Second, the court reasoned that even for the small number of minority voters who were affected by H.B. 2023 (i.e., who would use third-party ballot collectors no longer permitted by H.B. 2023 if they could), the evidence did not show that H.B. 2023 gave minorities less opportunity than other members of the electorate to participate in the political process and elect representatives. Id . at ––––, 2018 WL 2191664, at *34. While H.B. 2023 might make it "slightly more difficult or inconvenient for a small, yet unquantified subset of voters to return their early ballots," the court found that there was no evidence that H.B. 2023 "would make it significantly more difficult to vote," particularly given that no individual voter had testified that H.B. 2023 had this impact. Id . Therefore, the district court found that DNC had not carried its burden at the first step of the § 2 analysis. Id .

Although the district court did not need to reach the second step, it nonetheless reviewed the relevant Senate Factors in order to develop the record and concluded that DNC had likewise failed to carry its burden at step two. Id . at –––– – ––––, 2018 WL 2191664, at *36–40.²⁰

20 As noted above, supra at 712–12 n.15, because the district court correctly determined that H.B. 2023 does not satisfy step one of the § 2 analysis, we need not evaluate the district court's analysis of these factors in detail. Nevertheless, the district court's factual conclusions were not clearly erroneous, and as explained below, see infra at 729–30 n.32, we reject the dissent's factual reevaluations.

3

The district court's conclusion that the burden on a protected class of voters is so minimal that it would not give them less opportunity to elect representatives of their choice is not clearly erroneous. DNC produced anecdotal testimony that various sources collected between fifty and a few thousand ballots but DNC's counsel could not articulate an estimate more precise than that "thousands" of people used this opportunity. Id . at ––––, 2018 WL 2191664, at *12. Accordingly, the district court did not clearly err in estimating that fewer than 10,000 voters used ballot collection services in each election. Moreover, the district court even considered a more generous, although "unjustified," number of 100,000 voters, but nonetheless found that this was "relatively small" in relation to the 1.4 million early mail ballots and 2.3 million total voters. Id . The district court's view was, at minimum, a permissible view of the evidence. See Bessemer City , 470 U.S. at 573, 105 S.Ct. 1504. Given these small numbers, the district court did not clearly err in concluding that the unavailability of third party ballot collection would have minimal effect on the opportunity of minority voters to elect representatives of their choice.

Further, as explained in the Anderson / Burdick analysis, the evidence available indicated that the burden on even those few minority voters who used third-party ballot collection was minimal, because those voters had "done so out of convenience or personal preference, or because of circumstances that Arizona law adequately accommodates in other ways," rather than from necessity. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *14. As the district court pointed out, not a single voter testified at trial that H.B. 2023 made it significantly more difficult to vote, despite the fact that H.B. 2023 was in place for two 2016 elections. Id . at ––––, 2018 WL 2191664, at *34.²¹

21 In arguing that H.B. 2023 had a disparate impact on the ability of minorities to participate in the political process, the dissent fails to address this key fact.

In challenging the district court's conclusion, DNC and the dissent argue that under § 2, the total number of votes affected is not the relevant inquiry; the proper test is whether any minority votes are burdened. This argument is meritless. As we have explained, a "bare statistical showing" that an election practice "has a disproportionate impact on a racial minority does not satisfy the § 2 'results' inquiry." Salt River , 109 F.3d at 595. Rather, the test under § 2 is whether the "members [of a class of protected citizens] have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice." 52 U.S.C. § 10301(b) (emphasis added).²² To determine whether a challenged law will result in members of a class having less opportunity to elect representatives of their choice, a court must necessarily consider the severity and breadth of the law's impacts on the protected class.

22 While DNC cites extensively to the dissent in Chisom in arguing that they need not prove members of a protected class have less opportunity to elect representatives of their choice, we are bound by the majority, which rejected this argument. 501 U.S. at 397 & n.24, 111 S.Ct. 2354.

Accordingly, we affirm the district court's ruling that DNC failed to establish that H.B. 2023 results in less opportunity for minority voters to participate in the political process and to elect representatives of their choice, and therefore H.B. 2023 did not violate § 2 of the VRA.

C

Finally, we consider DNC's claim that H.B. 2023 violated the Fifteenth Amendment.

1

Plaintiffs can challenge a state's election practice as violating their Fifteenth Amendment rights by showing that "a state law was enacted with discriminatory intent." Abbott v. Perez , ––– U.S. ––––, 138 S.Ct. 2305, 2324, 201 L.Ed.2d 714 (2018). Discriminatory intent "implies more than intent as volition or intent as awareness of consequences." Pers. Adm'r of Mass. v. Feeney , 442 U.S. 256, 279, 99 S.Ct. 2282, 60 L.Ed.2d 870 (1979). Rather, plaintiffs must show that a state legislature "selected or reaffirmed a particular course of action at least in part 'because of,' not merely 'in spite of,' its adverse effects upon an identifiable group." Id. Thus, although racial discrimination need not be the "dominant" or "primary" factor underlying a legislative enactment, it must be a "motivating factor." Vill. of Arlington Heights v. Metro. Hous. Dev. Corp. , 429 U.S. 252, 265–66, 97 S.Ct. 555, 50 L.Ed.2d 450 (1977).

A law is not infected by discriminatory intent merely "because it may affect a greater proportion of one race than of another." Washington v. Davis , 426 U.S. 229, 242, 96 S.Ct. 2040, 48 L.Ed.2d 597 (1976). Rather, "[d]etermining whether invidious discriminatory purpose was a motivating factor demands a sensitive inquiry into such circumstantial and direct evidence of intent as may be available." Arlington Heights , 429 U.S. at 266, 97 S.Ct. 555. This inquiry is guided by factors set forth in Arlington Heights . Id . at 266–68, 97 S.Ct. 555 ; see Bolden , 446 U.S. at 62, 72–74, 100 S.Ct. 1519 (holding that a facially neutral law "violates the Fifteenth Amendment only if motivated by a discriminatory purpose" and applying Arlington Heights in an analysis of discriminatory intent).

Under the Arlington Heights framework, "the following, non-exhaustive factors" are relevant "in assessing whether a defendant acted with discriminatory purpose: (1) the impact of the official action and whether it bears more heavily on one race than another; (2) the historical background of the decision; (3) the specific sequence of events leading to the challenged action; (4) the defendant's departures from normal procedures or substantive conclusions; and (5) the relevant legislative or administrative history." Arce v. Douglas , 793 F.3d 968, 977 (9th Cir. 2015). Because of "the presumption of good faith that must be accorded legislative enactments" and the "evidentiary difficulty" in determining whether race was a motivating factor, courts must "exercise extraordinary caution" when engaging in this inquiry. Miller v. Johnson , 515 U.S. 900, 916, 115 S.Ct. 2475, 132 L.Ed.2d 762 (1995).

Discriminatory intent "is a pure question of fact" subject to review for clear error. Pullman-Standard , 456 U.S. at 287–88, 102 S.Ct. 1781 ; Abbott , 138 S.Ct. at 2326. "It is not a question of law and not a mixed question of law and fact." Pullman-Standard , 456 U.S. at 288, 102 S.Ct. 1781.

Given this standard, we must determine whether the district court's finding that the Arizona legislature did not have discriminatory intent is clearly erroneous. We consider the district court's findings on each Arlington Heights factor.

2

We start with two of the Arlington Heights factors, the historical background and legislative history of the enactment. Arce , 793 F.3d at 977. According to the district court, Arizona's history was "a mixed bag of advancements and discriminatory actions." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *38. Although there was evidence of discrimination and racially polarized voting, there was also evidence of improvement. While Arizona was subject to § 5 preclearance, "the DOJ did not issue any objections to any of [Arizona's] statewide procedures for registration or voting." Id . at ––––, 2018 WL 2191664, at *37. Moreover, Arizona enacted an Independent Redistricting Commission to combat problems with discrimination in drawing statewide redistricting plans. Id . at ––––, 2018 WL 2191664, at *38.

The district court also noted the relevant legislative history of H.B. 2023, including "farfetched allegations of ballot collection fraud" made by one legislator, Arizona State Senator Don Shooter, id . at ––––, 2018 WL 2191664, at *41, and a video (referred to as the "LaFaro Video") which "showed surveillance footage of a man of apparent Hispanic heritage appearing to deliver early ballots," id . at ––––, 2018 WL 2191664, at *38.²³ However, the court concluded that the legislature was not motivated by discriminatory intent. Rather, the court found that "Shooter's allegations and the LaFaro Video were successful in convincing H.B. 2023's proponents that ballot collection presented opportunities for fraud that did not exist for in-person voting, and these proponents appear to have been sincere in their beliefs that this was a potential problem that needed to be addressed." Id . at ––––, 2018 WL 2191664, at *41.

23 The district court found that the narration by Maricopa County Republican Chair A.J. LaFaro "contained a narration of 'Innuendos of illegality ... [and] racially tinged and inaccurate commentary by ... LaFaro.' " Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *38. The video was first introduced in 2014, but became "prominent in the debates over H.B. 2023." Id . at ––––, 2018 WL 2191664, at *39.

The district court's conclusion is well supported by the legislative record, which shows that legislative discussion focused on the danger of fraud. For example, the bill's sponsor, Senator Michelle Ugenti-Rita, stated that H.B. 2023 was designed to "limit fraud" in ballot collection, which "is important to maintaining integrity in our electoral process" because the ballot collection practice "is ripe to be taken advantage of." Senator Steve Smith testified that ballot fraud is "certainly happening," and Michael Johnson, an African American who had served on the Phoenix City Council, testified that he had constituents call to complain about ballot collectors in minority communities. Senator Smith cited this testimony in a speech supporting the law. Senator Sylvia Allen expressed concern that "we do not know what happens between the time the ballots are collected and when they're finally delivered." This concern was confirmed by State Election Director Eric Spencer, who testified that "there is a huge imbalance in the amount of security measures that are in place for polling place voting compared to early voting." Even though "77 percent of all the votes cast in Arizona" are early votes, there are "almost no prophylactic security procedures in place to govern that practice, whereas, at the polling place, where only 23 percent of the votes are taking place, we have every security measure in the world."

The legislature also heard testimony that other states had implemented similar security measures related to ballot collection. According to the legislative record, at the time H.B. 2023 was considered by the Arizona legislature, "California, New Mexico, Colorado, [and] Nevada all ha[d] laws that restrict or prohibit ballot collection," and therefore Arizona was "a little bit out of the norm especially among our neighbors." The legislature also heard that the California law was more draconian than H.B. 2023: it prohibited all ballot collection except by members of the household, family members, and spouses, and did not count votes in ballots that had been improperly collected.

DNC and the dissent claim that the district court erred in giving weight to this evidence because there was no evidence of actual fraud. According to DNC, this evidentiary gap established that the legislators' expressed concerns regarding fraud in ballot collection were merely a facade for racial discrimination. This argument fails. The Arizona legislature was free to enact prophylactic measures even when the legislative record "contains no evidence of any such fraud actually occurring." Crawford , 553 U.S. at 194, 128 S.Ct. 1610. Moreover, as the district court noted, "H.B. 2023 found support among some minority officials and organizations," including Michael Johnson, the African American councilman, and the Arizona Latino Republican Association for the Tucson Chapter, which undermines DNC's claim that concerns about fraud were a mere front for discriminatory motives. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *41.

DNC argues that the district court erred in not giving sufficient weight to the evidence that the LaFaro video had racial overtones. The district court's decision to give this evidence less weight was not a legal error, however, because the district court was not obliged to impute the motives of a few legislators to the entire Arizona legislature that passed H.B. 2023. See Arlington Heights , 429 U.S. at 265–66, 97 S.Ct. 555. "What motivates one legislator to make a speech about a statute is not necessarily what motivates scores of others to enact it." United States v. O'Brien , 391 U.S. 367, 384, 88 S.Ct. 1673, 20 L.Ed.2d 672 (1968).²⁴ The Sixth Circuit recently recognized this point, holding that the clearly discriminatory statements and motive of one legislator did not show that the enacting legislature "acted with racial animus." Ne. Ohio Coal. , 837 F.3d at 637.

24 DNC relies on Masterpiece Cakeshop, Ltd. v. Colorado Civil Rights Commission , ––– U.S. ––––, 138 S.Ct. 1719, 201 L.Ed.2d 35 (2018), for the principle that courts should put more weight on discriminatory statements of individual decisionmakers, but that case is not on point. In holding that statements of individual commissioners were relevant to determine whether a law intentionally discriminated on the basis of religion, the Court distinguished the adjudicatory context from the legislative context. See id. at 1730. Masterpiece Cakeshop explained that while "[m]embers of the Court have disagreed on the question whether statements made by lawmakers may properly be taken into account in determining whether a law intentionally discriminates on the basis of religion," the remarks in this case were made "in a very different context—by an adjudicatory body deciding a particular case." Id . Because our case involves a legislature enacting a general statute, rather than adjudicating a specific case, Masterpiece Cakeshop is not applicable.

The district court also did not err in giving little weight to evidence that "some individual legislators and proponents were motivated in part by partisan interests." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *43. The record shows that State Senator Shooter's concerns about ballot collection arose after he won a close election, that Michael Johnson complained that ballot collection put candidates without an effective get-out-the-vote effort at a disadvantage, and a 2014 Republican candidate for the Arizona House of Representatives claimed that he lost his election because of ballot collection activities. Id . Although DNC and the dissent seem to argue that, as a matter of law, legislators should be deemed to have a discriminatory intent for Fifteenth Amendment purposes when they are motivated by partisan interests to enact laws that disproportionately burden minorities, this is incorrect. Fifteenth Amendment plaintiffs must show that the legislature acted with racial motives, not merely partisan motives. See, e.g. , Cooper v. Harris , ––– U.S. ––––, 137 S.Ct. 1455, 1473, 197 L.Ed.2d 837 (2017) ("[A] trial court has a formidable task: It must ... assess whether the plaintiffs have managed to disentangle race from politics and prove that the former drove a district's lines."); Easley v. Cromartie , 532 U.S. 234, 243, 121 S.Ct. 1452, 149 L.Ed.2d 430 (2001) (evaluating the district court's critical finding "that race rather than politics" motivated the districting map). The "intent to preserve incumbencies" is not equivalent to racially-discriminatory intent, and only the latter supports a finding of intentional discrimination. Garza v. County of Los Angeles , 918 F.2d 763, 771 & n.1 (9th Cir. 1990). Even when "racial identification is highly correlated with political affiliation," Cooper , 137 S.Ct. at 1473 (quoting Easley , 532 U.S. at 243, 121 S.Ct. 1452 ), plaintiffs must still carry their burden of showing that the former was a motivating factor. Id. Accordingly, the determination whether racial or political interests motivated a legislature is one of fact subject to review for clear error. See Cooper , 137 S.Ct. at 1473–74. Here the district court disentangled racial motives from partisan motives, and its factual finding that even those few legislators harboring partisan interests did not act with a discriminatory purpose is not clearly erroneous.²⁵ Therefore, the historical and legislative history factors support the district court's conclusion.

25 Contrary to the dissent, the district court did not find that "partisan self-interest [ ] absolve[d] discriminatory intent." Dissent at 748. Rather, the district court determined that the Arizona legislature did not act with discriminatory intent, and passed H.B. 2023 in spite of any potential disparate-impact on minority voters, not because of it. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *41.

3

We next turn to the Arlington Heights factors of the "sequence of events" leading to the challenged action and "departures from normal procedures." Arce , 793 F.3d at 977. First, the district court found that the Arizona legislature followed its normal course in enacting H.B. 2023, and therefore the legislative process itself did not raise an inference of discriminatory intent. Reagan , ––– F.Supp.3d at –––– – ––––, 2018 WL 2191664, at *42–43. This conclusion is supported by the record; there is no evidence that the legislature used unusual procedures or unprecedented speed to pass a law, N.C. State Conference of NAACP v. McCrory , 831 F.3d 204, 214, 228 (4th Cir. 2016), which other courts have deemed raise such an inference, see, e.g. , Veasey I , 830 F.3d at 238 (holding that the Texas legislature's unwonted procedure of designating the bill "as emergency legislation," cutting debates short, passing it without the ordinary committee process, and suspending a two-thirds voting rule to get the bill passed, weighed in favor of a finding of discriminatory intent).

Second, in considering the historical sequence of events, the district court held that neither of the two prior efforts to limit ballot collection, S.B. 1412 (enacted in 2011) and H.B. 2305 (enacted in 2013), weighed in favor of finding that the legislature had a discriminatory intent in enacting H.B. 2023. Reagan , ––– F.Supp.3d at –––– – ––––, 2018 WL 2191664, at *42–43. The record showed that S.B. 1412 was subject to § 5 preclearance, and that after the DOJ requested additional information regarding the ballot collection provision, the Arizona Attorney General voluntarily withdrew the provision. Id . at ––––, 2018 WL 2191664, at *42. Two years later, the legislature enacted H.B. 2305, which also regulated ballot collection. Id . After citizen groups organized referendum efforts against the law, the legislature repealed it. Id . The court held that while these circumstances were somewhat suspicious, they "have less probative value because they involve different bills passed during different legislative sessions by a substantially different composition of legislators." Id .

The district court did not clearly err in giving little weight to these prior enactments. Even if the bills had been informed by a discriminatory intent, the Supreme Court has made clear that "[p]ast discrimination cannot, in the manner of original sin, condemn governmental action that is not itself unlawful." Abbott , 138 S.Ct. at 2324 (quoting Bolden , 446 U.S. at 74, 100 S.Ct. 1519 ). The intent of a prior legislature cannot be imputed to a new legislature enacting a different bill "notwithstanding the previous drafter's intent." Veasey v. Abbott (Veasey II) , 888 F.3d 792, 802 (5th Cir. 2016). Indeed, it is a clear error to presume that any invidious intent behind a prior bill "necessarily carried over to and fatally infected" the law at issue. Id . Further, "meaningful alterations" in an amended statute may render even a previously discriminatory statute valid. Id . (citation omitted). Because Arizona's previous laws on ballot collection were different rules, passed by different legislatures, and H.B. 2023 is "more lenient than its predecessors given its broad exceptions for family members, household members, and caregivers," these prior enactments do not materially bear on the legislature's intent in enacting H.B. 2023. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *43.

Moreover, the district court did not err in finding that neither S.B. 1412 or H.B. 2305 was enacted with racially discriminatory intent. Regarding S.B. 1412, the record shows only that the DOJ requested more information, but its primary concern was the law's "impact on minority voters," Feldman III , 843 F.3d at 369 (emphasis added), not the intent of the legislature in enacting it.²⁶ And as to H.B. 2305, the record does not disclose why citizens opposed the law or whether the referendum sought to combat a discriminatory purpose. The lack of evidence of past discrimination further undermines DNC's argument that the legislature had discriminatory intent in passing H.B. 2023.

26 To support its claim, DNC points to Representative Ruben Gallego's statements to the DOJ that S.B. 1412 was motivated by discriminatory intent. But Gallego opposed S.B. 1412, and "[t]he Supreme Court has ... repeatedly cautioned—in the analogous context of statutory construction—against placing too much emphasis on the contemporaneous views of a bill's opponents" in determining a legislature's intent. Veasey I , 830 F.3d at 234 (quoting Butts v. City of New York , 779 F.2d 141, 147 (2d Cir. 1985) ). DNC also points to statements by Amy Chan (formerly Amy Bjelland) to the DOJ, but the district court reasonably interpreted her statements as merely explaining that the impetus for S.B. 1412 was an accusation of voter fraud in San Luis, a predominately Hispanic area in the southern portion of Arizona. Feldman III , 843 F.3d at 384.

4

In reviewing the final Arlington Heights factor (whether the law would have a disparate impact on a particular racial group), Arce , 793 F.3d at 977, the district court found that "the legislature enacted H.B. 2023 in spite of its impact on minority [get out the vote] efforts, not because of that impact," and concluded that "proponents of the bill seemed to view these concerns as less significant because of the minimal burdens associated with returning a mail ballot," Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *43.

The district court did not clearly err in reaching this conclusion. Multiple senators expressed their view that H.B. 2023 imposes only a slight burden on voters. For instance, Senator Michelle Ugenti-Rita stated that voters have "[l]ots of opportunities" to vote in the 27 day early-voting window, and expressed her view that there is no reason to presume a voter who previously used ballot collection would have trouble voting. Given that these voters have already asked "that their ballot be mailed to them," Senator Ugenti-Rita stated "logic would tell you they are perfectly capable and understand that, in order to then get their ballot in, they need to put it back in to the mailbox or drop it off." Another proponent of the bill, John Kavanaugh, expressed a similar view: "The only way you get an early ballot is to have it delivered to you by mail, and the way you're supposed to return an early ballot is to reverse that process. And it's hard to imagine how, when you have an early ballot, somewhere in the area of 30 days, you somehow can't do that." Again, the record does not contain the sort of evidence that has led other courts to infer the legislature was acting with discriminatory intent, such as evidence that the legislators studied minority data and targeted the voting methods most used by minority voters. Cf. McCrory , 831 F.3d at 220. In fact, no voters, minority or non-minority, testified that they faced a substantial obstacle to voting because of H.B. 2023. Accordingly, we find no clear error in the court's holding that "[b]ased on the totality of the circumstances," DNC had "not shown that the legislature enacted H.B. 2023 with the intent to suppress minority votes." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *43.

In sum, the district court carefully weighed the evidence of discriminatory purpose and found the Arizona legislature was not motivated by an intent to discriminate. The findings supporting this conclusion are not clearly erroneous, and neither was the ultimate balancing of the Arlington Heights factors.

5

Because discriminatory intent is a "pure question of fact," a court must defer to the district court's fact-finding unless it is clearly erroneous. Pullman-Standard , 456 U.S. at 288, 102 S.Ct. 1781. But the dissent once again reviews the record de novo, reweighs the evidence, and reaches its own conclusion. For instance, the district court referenced Senator Shooter's allegations and the LaFaro video, but concluded, based on its review of the record, that the legislature was not motivated by discriminatory intent. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *41. The dissent simply reaches the opposite conclusion, based on the same evidence. Dissent at 748–49. Similarly, the dissent claims "the district court was wrong to determine that a law is not racially motivated if any people of color support it." Dissent at 749. But that mischaracterizes the district court's holding. Rather, after reviewing the evidence in the record, the district court found that H.B. 2023 was supported by minority officials and organizations. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *41. The district court did not err in considering that fact, among others, in determining whether the supporters of H.B. 2023 were motivated by racial discrimination, and the district court need not have concluded, as does the dissent, that such evidence "simply demonstrates that people of color have diverse interests." Dissent at 749. The Supreme Court has long held that an appellate court may not reject a district court's findings as clear error even when the court is "convinced that had it been sitting as the trier of fact, it would have weighed the evidence differently." Bessemer City , 470 U.S. at 574, 105 S.Ct. 1504. The dissent's approach contradicts this rule.

Further, the dissent supports its conclusion that "H.B. 2023 was enacted for the purpose of suppressing minority votes" by creating its own per se rules that a legislature's anti-fraud motive is pretextual when there is no direct evidence of voter fraud, and that a legislature's partisan motives are evidence of racial discrimination. Dissent at 746–47, 748–49. The dissent cites no support for these new rules, likely because Supreme Court precedents contradict them: Crawford rejected the idea that actual evidence of voter fraud was needed to justify restrictions preventing voter fraud, 553 U.S. at 195–96 & nn.11–13, 128 S.Ct. 1610 ; and Cooper made clear plaintiffs must "disentangle race from politics and prove that the former drove" the legislature, 137 S.Ct. at 1473. The dissent's attempt to reframe the evidence does not make the district court's resolution of this purely factual question clearly erroneous. Pullman-Standard , 456 U.S. at 287–88, 102 S.Ct. 1781.

IV

We now turn to DNC's challenges to the OOP policy. DNC argues that (1) the OOP policy violates the First and Fourteenth Amendment; and (2) the OOP policy violates § 2 of the VRA.

A

We begin with DNC's claims that the OOP policy violates the First and Fourteenth Amendment by imposing an unconstitutional burden under the Anderson / Burdick test.

1

As an initial matter, we agree with the district court's characterization of these claims as constituting a challenge to the precinct voter system. As discussed, most Arizona counties use a precinct-based system for the 20 percent of voters who vote in person on Election Day. In-person voters must cast their ballots in their assigned precinct, or their votes will not be counted. See Ariz. Rev. Stat. §§ 16-122, 16-135, 16-584 (codified in 1979); 1970 Ariz. Sess. Laws, ch. 151, § 64 (amending Ariz. Rev. Stat. § 16-895); Ariz. Rev. Stat. § 16-102 (1974). This rule does not apply to voters who cast their ballots in a county that use a vote center system, or who use other methods to vote.

On appeal, DNC argues that it is not challenging the rule requiring voting within a precinct, but rather Arizona's enforcement of the rule by not counting ballots cast in the wrong precinct (which it calls disenfranchisement).²⁷ This argument is sophistical; it conflates the burden of complying with an election rule with the consequence of noncompliance. As the Supreme Court has recognized, a state has an obligation to structure and organize the voting process within the state through a system of election rules. Storer , 415 U.S. at 730, 94 S.Ct. 1274. For instance, states typically have election rules that require voters to register to vote and to cast their votes in person during the hours when polls are open. These rules impose certain minimal burdens on voters—the ordinary burdens of registering to vote and showing up on time. If voters fail to comply, they may be unable to vote or their ballots may not be counted. But it is the election rules that impose a burden on the voter—not the enforcement of those rules. Under DNC's theory, a state could not enforce even a rule requiring registration, because the state's failure to count the vote of a non-registered voter would "disenfranchise" the noncompliant voter.

27 This is a misnomer. A state disenfranchises voters (for example, pursuant to a felon disenfranchisement law) by depriving certain individuals of their right to vote, not by requiring voters to comply with an election rule in order to have their votes counted. As the Supreme Court has explained, an election rule, such as the requirement to have a valid photo ID in order to vote, may be valid, even if a voter's noncompliance with such a rule means that the voter's ballot will not be counted. Crawford , 553 U.S. at 187, 189, 128 S.Ct. 1610.

Rather than adopt DNC's fallacious approach, we are guided by the Supreme Court's approach in Crawford. Crawford considered a state's election rule which provided that in-person voters who did not have valid photo ID, and did not thereafter verify their identities, were unable to have their votes counted. 553 U.S. at 186, 128 S.Ct. 1610. In conducting its Anderson / Burdick analysis, Crawford held that this photo ID rule imposed the burden of obtaining the requisite identification by "making a trip to the [issuing agency], gathering the required documents, and posing for a photograph," id. at 198, 128 S.Ct. 1610, and potentially could impose a heavier burden on subgroups, such as the homeless or those lacking birth certificates, id. at 199, 128 S.Ct. 1610. The Court's analysis would make little sense if the relevant burden were the state's enforcement of the photo ID rule; under that view, all voters would be subject to the same burden—that of having their non-compliant votes discounted. Accordingly, like the district court, we conclude that the appropriate analysis is whether compliance with the voter requirement in question—here, the requirement to vote in an assigned precinct—imposes an undue burden. See also Serv. Emps. Int'l Union Local 1 v. Husted , 698 F.3d 341, 344 (6th Cir. 2012) (explaining that courts cannot "absolve[ ] voters of all responsibility for voting in the correct precinct or correct polling place by assessing voter burden solely on the basis of the outcome—i.e. the state's ballot validity determination").

2

Applying the Anderson / Burdick framework to the proper characterization of DNC's challenge, the district court found that the precinct voting rule did not unconstitutionally burden the right to vote. As with H.B. 2023, the district court first observed that Arizona's OOP policy has no impact on the vast majority of Arizona voters because 80 percent of them cast their ballots through early mail voting. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *21. The court also noted that the policy has no impact on voters in Graham, Greenlee, Cochise, Navajo, Yavapai, and Yuma counties, rural counties that adopted the vote center model. Id .

As to those few Arizonans who vote in person outside of the vote center counties, the district court found that the burden of voting in the correct precinct was minimal. The district court acknowledged that people who move frequently may fail to update their voter registration in a timely manner and, as a result, may not have their early ballot forwarded to their new address, and that "changes in polling locations from election to election, inconsistent election regimes used by and within counties, and placement of polling locations all tend to increase OOP voting rates," as well as incorrect information provided by poll workers. Id . at ––––, 2018 WL 2191664, at *22. The district court nevertheless concluded that "the burdens imposed on voters to find and travel to their assigned precincts are minimal and do not represent significant increases in the ordinary burdens traditionally associated with voting." Id . at ––––, 2018 WL 2191664, at *24. Moreover, the district court found, "Arizona does not make it needlessly difficult for voters to find their assigned precincts," citing the myriad ways Arizona provides that information to voters: direct mailings, multiple state and county websites, town halls, live events, and social media and other advertising. Id . at –––– – ––––, 2018 WL 2191664, at *23–24 This information is generally provided in both English and Spanish. Id . at ––––, 2018 WL 2191664, at *24. Further, the court found that "for those who find it too difficult to locate their assigned precinct, Arizona offers generous early mail voting alternatives." Id . In light of these measures, the district court did not clearly err in finding that the burden of voting in the correct precinct was minimal.

Considering the electorate as a whole, the court found that the number of out-of-precinct votes was "small and ever-dwindling." Id . Only 14,885 of the 2,320,851 Arizonan votes cast in the 2008 general election were cast outside of the correct precinct—just 0.64 percent of total votes. Id . at ––––, 2018 WL 2191664, at *21. That number dropped to 10,979 ballots in the 2012 general election—0.47 percent of total votes. Id . By the 2016 general election, only 3,970 votes were cast in the wrong precinct in Arizona—just 0.15 percent of the 2,661,497 total votes. Id . The small and decreasing number of out-of-precinct votes confirms the district court's conclusion that the burden of identifying the correct precinct is minimal.

We conclude that the district court's finding that the requirement to vote in the correct precinct is a minimal burden is not clearly erroneous. As the district court noted, precinct-based voting is an established method of conducting elections and is used in a majority of states. Id . at ––––, 2018 WL 2191664, at *8 ; see also Serv. Emps. , 698 F.3d at 344 (precinct-voting system); Sandusky Cty. Democratic Party v. Blackwell , 387 F.3d 565, 568 (6th Cir. 2004) (per curiam) ("One aspect common to elections in almost every state is that voters are required to vote in a particular precinct. Indeed, in at least 27 of the states using a precinct voting system, including Ohio, a voter's ballot will only be counted as a valid ballot if it is cast in the correct precinct."). And a majority of the states that use precinct voting do not count out-of-precinct ballots. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *8. The requirement to use mail voting or locate the correct precinct and then travel to the correct precinct to vote does not "represent a significant increase over the usual burdens of voting." Crawford , 553 U.S. at 198, 128 S.Ct. 1610.

DNC's arguments to the contrary are meritless. First, DNC argues that the burden imposed by Arizona's policy of not counting ballots cast outside of the proper precinct is not minimal because the ratio of Arizona voters who cast ballots outside of the correct precinct compared to total votes cast in-person on Election Day is higher than in any other state. This statistic is misleading, because the vast majority of Arizonans vote early by mail—not in-person on Election Day. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *21. More important, the relative difference between Arizona and other states does not shed any light on the only relevant issue: the size of the burden imposed by Arizona's precinct voter system.²⁸

28 The dissent offers similarly misleading statistics to support its assertion that "Arizona voters are far likelier to vote [out of precinct] than voters of other states." Dissent at 732. The dissent's graph, Dissent at 732–33, shows only that the small subset of Arizona voters who cast their ballots in-person on Election Day are more likely to vote outside their precinct than voters in other states. Dissent at 732. The vast majority of Arizona voters, however, vote early by mail. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *21. Further, the dissent mentions the total number of votes cast out of precinct in the 2012 election, but not the more recent data from the 2016 election, which supports the district court's conclusion that the number of votes cast out of precinct is an "ever-decreasing fraction of the overall votes cast in any given election." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *35.

Second, DNC points to the evidence in the record regarding the external factors that contribute to out-of-precinct voting in Arizona, such as residential mobility, polling place locations, and pollworker training, and argues that such external factors impose a heavier burden on minorities.²⁹ But even if DNC presented evidence showing that the burden of finding the correct precinct fell more heavily on minorities than nonminorities, such evidence would not establish that the burden is any more than de minimis. DNC does not cite evidence that would allow a court "to quantify either the magnitude of the burden on [any such] class of voters or the portion of the burden imposed on them that is fully justified," id . at 200, 128 S.Ct. 1610 ; nor does DNC directly contest the evidence on which the district court relied in determining the burden was minimal. For instance, the district court cited substantial evidence in the record showing that in "Arizona counties with precinct-based systems, voters generally are assigned to precincts near where they live, and county officials consider access to public transportation when assigning polling places," and that "Arizona voters also can learn of their assigned precincts in a variety of ways," by accessing multiple websites operated by Arizona or various counties, by being mailed notice of any changes in polling places, or by calling the county recorder, among numerous other methods. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *23. Further, the district court relied on a 2016 Survey of Performance of American Elections in which no Arizona respondents stated that it was "very difficult" to find their polling place, and 94 percent of Arizona respondents reported that it was "very easy" or "somewhat easy" to find their polling place. Id . Accordingly, we decline the invitation by DNC and the dissent to reweigh the same evidence considered by the district court and reach the opposite conclusion. See Bessemer City , 470 U.S. at 573, 105 S.Ct. 1504. Instead, we affirm the district court's determination that the Arizona precinct voter rule imposed only minimal burdens.

29 As the district court noted, DNC did not challenge the manner in which individual counties locate polling places, or the manner in which Arizona trains its poll workers or informs voters of their assigned precincts, thus undercutting any argument that such practices violated § 2. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *23.

3

We next consider the district court's conclusion that Arizona had important regulatory interests for requiring precinct-based voting. The court found that this precinct system serves an important planning function by allowing counties to estimate the number of voters who may be expected at any particular precinct, allowing for better allocation of resources and personnel. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *24. A well-run election increases voter confidence and reduces wait times. Id . Second, the precinct voting system ensures that each voter receives a ballot reflecting only the races for which that person is entitled to vote, which "promotes voting for local candidates and issues and helps make ballots less confusing by not providing voters with ballots that include races for which they are not eligible to vote." Id .

The court concluded that the OOP policy was sufficiently justified by Arizona's important interests in light of the minimal burdens it imposes, and held that Arizona's practice did not need to be the narrowest means of enforcement. Id . at –––– – ––––, 2018 WL 2191664, at *24–26. The court therefore rejected DNC's arguments that Arizona should be required to adopt a more narrowly tailored rule and partially count ballots that were cast out-of-precinct, i.e., "counting only the offices for which the OOP voter is eligible to vote." Id . at ––––, 2018 WL 2191664, at *25. Moreover, the court concluded that such a requirement would have significant impacts. If Arizona no longer enforced in-precinct voting, the court reasoned, people would "have far less incentive to vote in their assigned precincts and might decide to vote elsewhere." Id . at ––––, 2018 WL 2191664, at *25. Voters could also "be nefariously directed to vote elsewhere," id ., as detailed in N.C. State Conference of NAACP v. McCrory , 182 F.Supp.3d 320, 461 (M.D.N.C. 2016), rev'd on other grounds , 831 F.3d 204 (4th Cir. 2016). Further, partially counting ballots would burden candidates for local office, who would have to persuade voters to vote in-precinct. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *25. Finally, it would "impose a significant financial and administrative burden on Maricopa and Pima Counties because of their high populations." Id . Accordingly, the court concluded that Arizona's rejection of ballots cast out-of-precinct does not violate the First and Fourteenth Amendments.

We agree with the district court's analysis. The interests served by precinct-based voting are well recognized. As the Sixth Circuit has explained:

The advantages of the precinct system are significant and numerous: it caps the number of voters attempting to vote in the same place on election day; it allows each precinct ballot to list all of the votes a citizen may cast for all pertinent federal, state, and local elections, referenda, initiatives, and levies; it allows each precinct ballot to list only those votes a citizen may cast, making ballots less confusing; it makes it easier for election officials to monitor votes and prevent election fraud; and it generally puts polling places in closer proximity to voter residences.

Sandusky Cty. Democratic Party , 387 F.3d at 569.

DNC does not dispute these legitimate interests, but argues that the OOP policy is not justified because it is administratively feasible to count ballots cast out-of-precinct, pointing to 20 other states which partially count out-of-precinct ballots. But restrictions such as the OOP policy that impose minimal burdens on voters' rights need not be narrowly tailored, see Burdick , 504 U.S. at 433, 112 S.Ct. 2059, and thus Arizona is not required to show that its electoral system "is the one best tailored to achieve its purposes." Dudum , 640 F.3d at 1114. Moreover, as the district court pointed out, DNC's "requested relief essentially would transform Arizona's precinct-based counties, including its two most populous, into quasi-vote-center counties." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *26. The mere fact that a minority of jurisdictions adopt a different system does not mean that Arizona's choice is unjustified. Where, as here, the plaintiff "effectively ask[s] the court to choose between electoral systems," we ordinarily reject such challenges. See Dudum , 640 F.3d at 1115. "[A]bsent a truly serious burden on voting rights," we have held that we must have "respect for governmental choices in running elections," particularly where "the challenge is to an electoral system, as opposed to a discrete election rule (e.g., voter ID laws, candidacy filing deadlines, or restrictions on what information can be included on ballots)." Id. at 1114–15 (emphasis omitted). As we have recognized, such variations are "the product of our democratic federalism, a system that permits states to serve 'as laboratories for experimentation to devise various solutions where the best solution is far from clear.' " Pub. Integrity All. , 836 F.3d at 1028 (quoting Ariz. State Legislature v. Ariz. Indep. Redistricting Comm'n , ––– U.S. ––––, 135 S.Ct. 2652, 2673, 192 L.Ed.2d 704 (2015) ).

DNC also contends that there is insufficient evidence that more voters will vote out-of-precinct if Arizona began partially counting out-of-precinct ballots. But just as with fraud prevention, Arizona does not need to produce "elaborate, empirical verification of the weightiness of [its] asserted justifications." Timmons v. Twin Cities Area New Party , 520 U.S. 351, 364, 117 S.Ct. 1364, 137 L.Ed.2d 589 (1997) ; see also Munro , 479 U.S. at 195, 107 S.Ct. 533 ("To require States to prove actual voter confusion, ballot overcrowding, or the presence of frivolous candidacies as a predicate to the imposition of reasonable ballot access restrictions would invariably lead to endless court battles over the sufficiency of the 'evidence' marshaled by a State to prove the predicate."). Courts wisely do not require "that a State's political system sustain some level of damage" before it can impose "reasonable restrictions" on the electoral process.³⁰ Munro , 479 U.S. at 195, 107 S.Ct. 533. Therefore, we affirm the district court's holding that the OOP policy is valid under the Anderson / Burdick framework.

30 The dissent also challenges the wisdom of Arizona's OOP policy, labeling as "illogical" Arizona's concern that without the policy voters may not have an incentive to identify and vote in their correct precinct. Dissent at 745–46. In reaching this conclusion, the dissent relies only on its own view of proper policy, a view that contradicts a majority of states, which each adopt the same approach as Arizona. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *8. We therefore reject this argument.

B

Finally, we address DNC's claim that the OOP policy violates § 2 of the VRA.

As noted above, at the first step, DNC must carry its burden of showing that the challenged practice (here Arizona's requirement that in-person voters vote in the correct precinct) gives members of a protected class less opportunity than other members of the electorate both "to participate in the political process and to elect representatives of their choice." Chisom , 501 U.S. at 397, 111 S.Ct. 2354 (quoting 52 U.S.C. § 10301(b) ).

The district court held that DNC did not carry its burden at the first step of its § 2 claim. Although finding that "minorities are over-represented among the small number of voters casting OOP ballots,"³¹ the court also found that out-of-precinct "ballots represent ... a small and ever-decreasing fraction of the overall votes cast in any given election." Reagan , ––– F.Supp.3d at –––– – ––––, 2018 WL 2191664, at *34–35. As noted above, only 3,970 out of 2,661,497 total votes, or 0.15 percent, were cast in the wrong precinct during the 2016 general election. Id . at ––––, 2018 WL 2191664, at *35. Further, as in its Anderson / Burdick analysis, the court found that the burden of identifying the correct precinct was minimal. The court noted that DNC had not challenged "the manner in which Arizona counties allocate and assign polling places or Arizona's requirement that voters re-register to vote when they move." Id . Nor had DNC claimed that there was "evidence of a systemic or pervasive history" of disproportionately giving minority voters misinformation as to precinct locations, or evidence "that precincts tended to be located in areas where it would be more difficult for minority voters to find them, as compared to non-minority voters." Id . Because the number of votes cast out of precinct by any voters was small and decreasing, and because the burden of finding the correct precinct was minimal (and the state had not made the burden more difficult for minorities), the district court concluded that the OOP policy did not give minority voters less opportunity than the rest of the electorate to participate in the political process and elect their preferred representatives. Id . at ––––, 2018 WL 2191664, at *36. Therefore, the court concluded that DNC had failed to carry its burden at the first step of § 2.³²

31 For example, among all counties that reported out-of-precinct ballots in the 2016 general election, roughly 99 percent of Hispanic, African American, and Native American voters cast ballots in the correct precinct, while the other 1 percent voted in the wrong precinct. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *34. By comparison, 99.5 percent of non-minority voters voted in the correct precinct, with 0.5 percent casting out-of-precinct ballots. Id . While this data shows, as Arizona notes, that minority voters were "twice as likely" to cast OOP ballots as non-minority voters, the relative percentages of voters in each group who vote in the correct and incorrect precincts are far more meaningful. See Frank , 768 F.3d at 752 n.3.

32 Having reached this conclusion, the district court did not need to reach step two, but nonetheless analyzed both challenged election practices together and found that, although some of the Senate Factors were present, DNC's causation theory was too tenuous to meet its burden. Reagan , ––– F.Supp.3d at –––– – ––––, 2018 WL 2191664, at *36–40. These findings are not clearly erroneous. In arguing to the contrary, the dissent again engages in appellate fact-finding, emphasizing some parts of the extensive record and ignoring others. For example, the district court found that DNC did not carry its burden of proving that "there is a significant lack of responsiveness on the part of elected officials to the particularized needs of the members of the minority groups." Id . at ––––, 2018 WL 2191664, at *27. This conclusion is supported by substantial evidence in the record, including evidence of outreach efforts by the Arizona Citizens Clean Elections Commission to increase minority voter education and participation, and evidence that Arizona had the sixteenth-highest minority representation ratio in the country. Although the dissent points to other evidence in the record, e.g., evidence that Arizona has the fourth-poorest health insurance coverage for children, and is ranked second-lowest overall per-pupil spending for Fiscal Year 2014, Dissent at 702–03, our proper role is to determine whether "the district court's account of the evidence is plausible in light of the record viewed in its entirety," Bessemer City , 470 U.S. at 574, 105 S.Ct. 1504, not to substitute our own evaluation of the record. Here, the district court's view of the evidence was clearly permissible, and we therefore disregard the dissent's impermissible reweighing of the evidence.

The district court did not clearly err in reaching this conclusion. Although DNC argues that minorities are more likely to cast out-of-precinct ballots, and that there have been close elections where out-of-precinct ballots could have made a difference, the fact that a practice falls more heavily on minorities is not sufficient to make out a § 2 violation. Salt River , 109 F.3d at 595. Rather, there must be a showing that the challenged practice causes a material impact on the opportunity provided to minorities to participate in the political process and to elect representatives of their choice. "[U]nless minority group members experience substantial difficulty electing representatives of their choice, they cannot prove that a challenged electoral mechanism impairs their ability 'to elect.' " Gingles , at 48 n.15, 106 S.Ct. 2752 (quoting 52 U.S.C. § 10301(b) ). A precinct voting system, by itself, does not have such a causal effect. Such a common electoral practice is a minimum requirement, like the practice of registration, that does not impose anything beyond "the usual burdens of voting." Crawford , 553 U.S. at 198, 128 S.Ct. 1610. As with other laws that impose such minimal burdens, a court can reasonably conclude that this background requirement, on its own, does not cause any particular group to have less opportunity to "influence the outcome of an election." Chisom , 501 U.S. at 397, 111 S.Ct. 2354. Indeed, DNC has not adduced any evidence to the contrary.

In arguing that the district court erred, the dissent relies primarily on its erroneous view that any disparate impact on minorities constitutes a violation of step one of § 2. See supra at 714–15 n.18. Based on this misunderstanding, the dissent argues that "the district court legally erred in determining that a critical mass of minority voters must be disenfranchised before § 2 is triggered."³³ Dissent at 735. But it is the dissent that errs in arguing that evidence that an election rule has any disparate impact on minorities is sufficient to succeed on a § 2 claim. Dissent at 735–36. As the Supreme Court pointed out, to meet the language of § 2, "all such claims must allege an abridgement of the opportunity to participate in the political process and to elect representatives of one's choice," Chisom , 501 U.S. at 398, 111 S.Ct. 2354, and must prove more than "the mere loss of an occasional election." Gingles , 478 U.S. at 51, 106 S.Ct. 2752. Here, the district court was faithful to the language of § 2. 52 U.S.C. § 10301 (b).³⁴

33 Of course, as explained above, supra at 724–25 n.27, an election rule requiring voters to identify their correct precinct in order to have their ballots counted does not constitute a "disenfranchisement" of voters.

34 In the alternative, the dissent argues that "in this instance, a critical mass has been shown." Dissent at 735 n.2. The record provides no support for this statement. Rather, the evidence shows that approximately 99 percent of Hispanic, African American, and Native American voters cast ballots in their correct precinct. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *34. In 2016 only 3,970 votes were cast out of precinct—0.15 percent of the total votes cast—and the record is silent on what number of those ballots were cast by minority voters. Reagan , ––– F.Supp.3d at –––– – ––––, 2018 WL 2191664, at *34–35. The dissent's only support for its claim is its brief reference to the dissent in Feldman II , 842 F.3d at 634, which in turn references two close primary elections in Arizona (one Republican, one Democrat) in 2012 and 2014, and five other close races over the course of the past 100 years (from 1916 to 2012). Dissent at 735–36 n.2. This certainly does not compel a conclusion that the district court's view of the relevant evidence was clearly erroneous.

This is not to say that plaintiffs could never carry their burden of showing a precinct-based voting system gave minority voters less opportunity. For instance, it is possible that a state could implement such a system in a manner that makes it more difficult for a significant number of members of a protected group to discover the correct precinct in order to cast a ballot. This could occur, for instance, if the state did not provide necessary information in the language best understood by a language minority. But here, the district court found that DNC did not present any evidence of this sort of practice. Reagan , ––– F.Supp.3d at –––– – ––––, 2018 WL 2191664, at *23–24. DNC does not contest this finding on appeal, nor does it challenge any other elements of Arizona's precinct voting system, such as individual counties' location of polling places, as unlawful.

Therefore, the district court correctly determined that the precinct voter system did not lessen the opportunities of minorities to participate in the political process and to elect representatives of their choice, and did not clearly err in rejecting DNC's argument that it need not provide evidence of this factor so long as there is evidence of some disparity in out-of-precinct voting.

V

After an exhaustive ten-day bench trial involving the testimony of 51 witnesses and over 230 exhibits, the district court made two key factual findings. First, it found that neither Arizona's precinct voter system nor H.B. 2023 imposed more than a minimal burden on voters or increased the ordinary burdens traditionally associated with voting. Second, it found that the Arizona state legislature was not motivated by a discriminatory purpose in enacting H.B. 2023. These findings, which were not clearly erroneous, effectively preclude DNC's claims. The finding that Arizona's two election practices place only the most minimal burden on voters necessarily leads to the conclusion that the practices did not result in less opportunity for minority voters to participate in the political process and elect representatives of their choice for purposes of § 2 of the VRA. Further, in light of the court's finding that the burden imposed on voters by the two election practices was minimal, Arizona easily carried its burden under the Anderson / Burdick test to show that its election practices were reasonably tailored to achieve the State's important regulatory interests. Finally, the court's finding that the legislature had no discriminatory purpose in enacting H.B. 2023 effectively eviscerates DNC's Fifteenth Amendment claim. Accordingly, we affirm the district court's determination that Arizona's election practices did not violate the First and Fourteenth Amendments or § 2 of the VRA, and H.B. 2023 did not violate the Fifteenth Amendment.

AFFIRMED .

THOMAS, Chief Judge, dissenting:

"No right is more precious in a free country than that of having a voice in the election of those who make the laws under which, as good citizens, we must live." Wesberry v. Sanders , 376 U.S. 1, 17, 84 S.Ct. 526, 11 L.Ed.2d 481 (1964). Our right to vote benefits government as much as it benefits us: a representative democracy requires participation, and the people require representatives accountable to them. Arizona's electoral scheme impedes this ideal and has the effect of disenfranchising Arizonans of African American, Hispanic, and Native American descent.

Arizona's policy of wholly discarding—rather than partially counting—votes cast out-of-precinct has a disproportionate effect on racial and ethnic minority groups. It violates § 2 of the Voting Rights Act ("VRA"), and it unconstitutionally burdens the right to vote guaranteed by the First Amendment and incorporated against the states under the Fourteenth Amendment.

H.B. 2023, which criminalizes most ballot collection, serves no purpose aside from making voting more difficult, and keeping more African American, Hispanic, and Native American voters from the polls than white voters.

I respectfully dissent.

I

No state rejects more out-of-precinct ("OOP") votes than Arizona. As the district court recognized, Arizona voters are far likelier to vote OOP than voters of other states. Democratic Nat'l Comm. v. Reagan , No. CV-16-01065-PHX-DLR, ––– F.Supp.3d ––––, ––––, 2018 WL 2191664, at *21 (D. Ariz. May 10, 2018) (hereinafter Reagan ). Indeed, "[i]n 2012 alone more than one in every five Arizona in-person voters was asked to cast a provisional ballot, and over 33,000 of these—more than 5 percent of all in-person ballots cast—were rejected." Id . (internal quotation marks and alterations omitted). The following graph compares the rate at which Arizona rejects OOP ballots to that of other states, showing just how much of an outlier Arizona is:

Arizona voters are likely to vote OOP for a constellation of reasons, the most striking of which is the frequency with which polling locations change, particularly in the highly populated urban areas. Id . at ––––, 2018 WL 2191664, at *22. Between 2006 and 2008, at least 43 percent of all polling places in Maricopa County—where approximately two-thirds of Arizona's registered voters live—changed locations, and 40 percent moved again between 2010 and 2012. Id. In 2016, Maricopa County went from 60 vote centers for the presidential preference election to 122 polling locations for the May special election to over 700 assigned polling locations in the August primary and November general elections. Id. In other words, the paths to polling places in the Phoenix area is much like the changing stairways at Hogwarts, constantly moving and sending everyone to the wrong place. The effect? Voters whose polling location changed were forty percent likelier to vote OOP. Id. Additionally, polling locations are often counterintuitive, further driving up OOP rates. Polls are likely to be placed on the edge of the precinct, and they are frequently clustered together—sometimes even in the same building. Unsurprisingly, voters who live further from their assigned polling location than from a location nearest to them or who are close to more than one location are likelier to end up casting a discarded ballot. Indeed, one-quarter of OOP voters cast their ballots in locations closer than their assigned polling place to their homes.

Worse, voters left confused by Arizona's labyrinthian system often miss out on the opportunity to cast a ballot in their assigned location, where it will be counted. At trial, all but one of the affected witnesses testified that they were never informed that they were voting OOP and that their ballot would not be counted. And the one witness who was given this crucial information was nonetheless unable to vote; he could not make it to his assigned location before the polls closed.

There is no question that Arizona's practice of discarding OOP ballots is also a practice of disproportionately discarding ballots cast by minority voters. The district court recognized as much. Id. at ––––, ––––, 2018 WL 2191664, at *4, *34. Indeed, although rates of OOP voting decreased in the last election, the disparity between white and minority voters remains constant. In the 2016 general election, Hispanic, African American, and Native American voters were twice as likely as white voters to vote OOP. Id. at ––––, 2018 WL 2191664, at *34.

Race and ethnicity intersect with the socioeconomic conditions that drive up OOP voting. It is frequently more difficult for minority voters to locate and vote in their assigned polling locations. As the district court noted, "OOP voting is concentrated in relatively dense precincts that are disproportionately populated with renters and those who move frequently. These groups, in turn, are disproportionately composed of minorities." Id. at ––––, 2018 WL 2191664, at *35.

Moreover, minority voters are far likelier to face significant barriers in traveling to the polls, barriers that compound the difficulty faced by the voter who is informed that she is in the wrong location and therefore needs to travel to a different precinct. The evidence showed that African American, Hispanic, and Native American voters in Arizona are more likely to work multiple jobs and to lack reliable transportation and childcare resources. Id. at ––––, 2018 WL 2191664, at *31. Given that voters may wait as long as five hours in line just to cast a ballot, it is not difficult to see how socioeconomic conditions may increase the significance of barriers to ballot access.

Native American voters, many of whom live on sovereign lands, face unique challenges. Navajo voters in Northern Apache County, for example, are not assigned standard addresses; their polling locations are assigned according to "guesswork." Id. at ––––, 2018 WL 2191664, at *35. And they often have different polling locations for tribal elections and state and federal elections. Id.

Despite these startling indicators, the district court concluded that Arizona's policy of discarding OOP ballots violates neither § 2 of the VRA nor the First Amendment, applicable to the states pursuant to the Fourteenth Amendment. I respectfully disagree on both counts.

II

Arizona's practice of discarding OOP ballots violates § 2 of the VRA. The practice "results in a denial or abridgement of the right of any citizen of the United States to vote on account of race or color," 52 U.S.C. § 10301(a), and, "based on the totality of circumstances," members of protected classes "have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice," id. § 10301(b).

The VRA "should be interpreted in a manner that provides 'the broadest possible scope' in combating racial discrimination." Chisom v. Roemer , 501 U.S. 380, 403, 111 S.Ct. 2354, 115 L.Ed.2d 348 (1991) (quoting Allen v. State Bd. of Elections , 393 U.S. 544, 567, 89 S.Ct. 817, 22 L.Ed.2d 1 (1969) ). There are two routes to vindication of a § 2 claim—a plaintiff may satisfy either the "intent test" or the "results test." Thornburg v. Gingles , 478 U.S. 30, 35, 44, 106 S.Ct. 2752, 92 L.Ed.2d 25 (1986). DNC has not alleged that the challenged practice was initiated for a discriminatory purpose, as required to satisfy the intent test. Rogers v. Lodge , 458 U.S. 613, 618, 102 S.Ct. 3272, 73 L.Ed.2d 1012 (1982) (requiring a showing of "invidious discriminatory purpose").

Thus, the operative question is whether, under "the totality of circumstances," members of a racial or ethnic minority "have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice," 52 U.S.C. § 10301(b).¹ Under the results test, a challenged law or practice violates § 2 of the VRA if: (1) it "impose[s] a discriminatory burden on members of a protected class, meaning that members of the protected class have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice"; and (2) that burden is "in part ... caused by or linked to social and historical conditions that have or currently produce discrimination against members of the protected class." League of Women Voters of N.C. v. North Carolina , 769 F.3d 224, 240 (4th Cir. 2014) (internal quotation marks omitted) (quoting Ohio State Conf. of the NAACP v. Husted , 768 F.3d 524, 553 (6th Cir. 2014) ); accord Veasey v. Abbott , 830 F.3d 216, 244 (5th Cir. 2016) ; Ohio Democratic Party v. Husted , 834 F.3d 620, 637 (6th Cir. 2016).

1 The use of the conjunction "and" in the quoted language did not create a new and more rigorous two-part test, as the majority's reading of Chisom v. Roemer , 501 U.S. 380, 111 S.Ct. 2354, 115 L.Ed.2d 348 (1991) suggests. See Op. 712–15. Rather, in Chisom , the Court explained why it rejected the notion that voters could not bring a vote dilution claim for judicial elections. Chisom , 501 U.S. at 396–97, 111 S.Ct. 2354. The Court clearly understood that the VRA does not demand a showing that the challenged provision may be outcome-determinative: "Any abridgment of the opportunity of members of a protected class to participate in the political process inevitably impairs their ability to influence the outcome of an election." Id. at 397, 111 S.Ct. 2354. Indeed, the Court wrote that it was a relatively "mere[ ]" thing to show that voters are denied the ability to influence an election's outcome; the greater hurdle is to show that voters are not allowed to fully participate. Id. at 396–97, 111 S.Ct. 2354 (rejecting the position that "a ... practice ... which has a disparate impact on black voters' opportunity to cast their ballots under § 2, may be challenged even if a different practice that merely affects their opportunity to elect representatives of their choice to a judicial office may not.").

Our responsibility is to interpret the law in accordance with Congress's "broad remedial purpose of 'ridding the country of racial discrimination in voting,' " Chisom , 501 U.S. at 403, 111 S.Ct. 2354 (alteration omitted) (quoting South Carolina v. Katzenbach , 383 U.S. 301, 315, 86 S.Ct. 803, 15 L.Ed.2d 769 (1966) ). Here, we know that African American, Hispanic, and Native American Arizonan voters are twice as likely as white voters to be disenfranchised by Arizona's OOP policy, and we know that the problem could be easily remedied. I would hold the challenged practice in violation of § 2 and enjoin Arizona from wholly discarding OOP ballots.

A

As the district court recognized, DNC "provided quantitative and statistical evidence of disparities in OOP voting." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *34. That evidence was "credible and shows that minorities are over-represented among the small number of voters casting OOP ballots." Id. Indeed, in 2016, whites were half as likely to vote OOP as African Americans, Hispanics, or Native Americans, a pattern displayed in all counties save one, which is predominately white. Id. The analysis at step one of the § 2 results test ought to end at this point, as DNC clearly met its burden of demonstrating that Arizona's practice of discarding OOP ballots places a "discriminatory burden" on African Americans, Hispanics, and Native Americans. League of Women Voters , 769 F.3d at 240.

The district court discredited this disparity, writing: "Considering OOP ballots represent such a small and ever-decreasing fraction of the overall votes cast in any given election, OOP ballot rejection has no meaningfully disparate impact on the opportunities of minority voters to elect their preferred representatives." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *35. However, this consideration is irrelevant to step one of § 2's results test, which focuses solely on the differences in opportunity and effect enjoyed by groups of voters. 52 U.S.C. § 10301. Thus, the district court legally erred in determining that a critical mass of minority voters must be disenfranchised before § 2 is triggered.² See Chisom , 501 U.S. at 397, 111 S.Ct. 2354 ("Any abridgment of the opportunity of members of a protected class to participate in the political process inevitably impairs their ability to influence the outcome of an election.").

2 What is more, in this instance, a critical mass has been shown. As I wrote when this case was last before us, regarding DNC's request for a preliminary injunction, the record demonstrates vote margins as thin as 27 votes in a 2016 partisan primary and about 10,000 votes in the 2002 gubernatorial general election. Feldman v. Ariz. Sec'y of State's Office , 842 F.3d 613, 634 (9th Cir. 2016).

The district court also determined that, "as a practical matter, the disparity between the proportion of minorities who vote at the wrong precinct and the proportion of nonminorities who vote at the wrong precinct does not result in minorities having unequal access to the political process." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *35. But when, as a result, proportionately fewer of the ballots cast by minorities are counted than those cast by whites, that is precisely what it means.

Under the standard applied by the district court, a poll tax or literacy test—facially neutral, evenly applied across racial and ethnic lines—could withstand scrutiny. After all, regardless of race, individuals who pay the tax or pass the test get to vote. However, the § 2 results test rejects this line of thinking. Gingles , 478 U.S. at 44, 106 S.Ct. 2752 (quoting S. Rep. No. 97-417, at 28 (1982), as reprinted in 1982 U.S.C.C.A.N. 177, 206) ("The 'right' question, ... is whether 'as a result of the challenged practice or structure plaintiffs do not have an equal opportunity to participate in the political processes and to elect candidates of their choice.' ").

Similarly, it is inappropriate to require, as the district court did, that DNC demonstrate a causal connection between Arizona's policy of not counting OOP ballots and the disparate rates of OOP voting. Reagan , ––– F.Supp.3d at –––– – ––––, 2018 WL 2191664, at *35–36. The district court misstated the burden by concluding that DNC is challenging the voters' own behavior rather than the state's policy of not counting OOP ballots. Because the challenged practice is Arizona's wholesale rejection of OOP ballots, it does not matter whether such rejection increases the rates of OOP voting.³

3 For the same reason, I disagree that we must be more deferential to the State on the grounds that "the challenge is to an electoral system , as opposed to a discrete election rule ." Op. 703 (quoting Dudum v. Arntz , 640 F.3d 1098, 1114 (9th Cir. 2011) ).

Moreover, the VRA does not demand the causal connection required by the district court. Rather, it is violated by a law that "impose[s] a discriminatory burden on members of a protected class" when that burden is "in part ... caused by or linked to" discriminatory conditions. League of Women Voters , 769 F.3d at 240. The district court flipped the requisite connection between the burden alleged and the conditions of discrimination by demanding DNC to show that the burden of having votes go uncounted leads to the socioeconomic disparities that in turn lead to OOP voting.

Applying the appropriate causation requirement leads to a different conclusion. The evidence showed the existence of a "causal connection between the challenged voting practice and [a] prohibited discriminatory result." Smith v. Salt River Project Agr. Imp. & Power Dist. , 109 F.3d 586, 595 (9th Cir. 1997) (quoting Ortiz v. City of Phila. Office of City Comm'rs Voter Registration Div. , 28 F.3d 306, 312 (3d Cir. 1994) ); see also id. at 595 ("Only a voting practice that results in discrimination gives rise to § 2 liability.") (emphasis added). Here, the challenged practice—not counting OOP ballots—results in "a prohibited discriminatory result"; a substantially higher percentage of minority votes than white votes are discarded. Id. at 586.

The district court recognized that socioeconomic disparities between whites and minorities increase the likelihood of OOP voting. In the district court's words, "OOP voting is concentrated in relatively dense precincts that are disproportionately populated with renters and those who move frequently. These groups, in turn, are disproportionately composed of minorities." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *35. It also recognized that "Hispanics, Native Americans, and African Americans ... are significantly less likely than non-minorities to own a vehicle, more likely to rely upon public transportation, [and] more likely to have inflexible work schedules." Id. at ––––, 2018 WL 2191664, at *32.

I cannot accept the proposition that, under § 2, the State is absolved of any responsibility to correct disparities if they can be attributed to socioeconomic factors. See Gingles , 478 U.S. at 63, 106 S.Ct. 2752 ("[T]he reasons black and white voters vote differently have no relevance to the central inquiry of § 2."). When we look at the evidence through this lens, the district court's findings give rise to certain logical inferences. For one, when a polling location is situated on one end of a precinct—as often occurs—it is disproportionately difficult for minorities to get to that location. And, in the event that a poll worker informs the voter that she is in the wrong precinct and her ballot will be uncounted, she is likelier to have the opportunity to successfully travel to and vote at her assigned polling location if she is white. The district court erred by requiring DNC to show that "Arizona's policy to not count OOP ballots is ... the cause of the disparities in OOP voting." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *35. The VRA imposes no such requirement.

The district court also erred by discounting the significance of its determination that "[p]olling place locations present additional challenges for Native American voters." Id. As the trial court itself noted:

Navajo voters in Northern Apache County lack standard addresses, and their precinct assignments for state and county elections are based upon guesswork, leading to confusion about the voter's correct polling place. Additionally, boundaries for purposes of tribal elections and Apache County precincts are not the same. As a result, a voter's polling place for tribal elections often differs from the voter's polling place for state and county elections. Inadequate transportation access also can make travelling to an assigned polling place difficult.

Id. Remedying the legal error committed by the trial court in imposing an overly onerous burden on the plaintiffs, the court's own findings demonstrate that African American, Hispanic, and Native American voters are far likelier than white voters to vote OOP and see their votes go uncounted.

In sum, I take no issue with the district court's findings of fact. Rather, I disagree with the application of law to the facts, and the conclusions drawn from them. In particularly, I respectfully disagree with the conclusion that the findings—which conclusively demonstrate the existence of disparate burdens on African American, Hispanic, and Native American voters—can be discounted on the grounds that there are not enough disenfranchised voters to matter. See Salt River Project , 109 F.3d at 591 (citation and internal quotation marks omitted) (noting "the [court's] power to correct errors of law, including those that may infect a so-called mixed finding of law and fact, or a finding of fact that is predicated on a misunderstanding of the governing rule of law").

B

As required at step two of the results test, DNC has shown that, under the "totality of circumstances," 52 U.S.C. § 10301(b), the disparate burden of disenfranchisement is "in part ... caused by or linked to social and historical conditions that have or currently produce discrimination against members of the protected class," League of Women Voters , 769 F.3d at 240 (citation and internal quotation marks omitted). This step "provides the requisite causal link between the burden on voting rights and the fact that this burden affects minorities disparately because it interacts with social and historical conditions that have produced discrimination against minorities currently, in the past, or both." Veasey , 830 F.3d at 244. "[T]he second step asks not just whether social and historical conditions 'result in' a disparate impact, but whether the challenged voting standard or practice causes the discriminatory impact as it interacts with social and historical conditions." Husted , 834 F.3d at 638 (emphasis removed).

In 1982, Congress amended the VRA in response to Mobile v. Bolden , 446 U.S. 55, 100 S.Ct. 1519, 64 L.Ed.2d 47 (1980), in which the Supreme Court held that the VRA—like the Civil Rights Amendments—was indifferent to laws with a disparate impact on minority voters. Gingles , 478 U.S. at 35, 106 S.Ct. 2752. Consistent with Congress's intent, courts consider a non-exhaustive list of factors outlined in the Senate Report accompanying the 1982 amendments. Id. As relevant here, courts consider: (1) the history of official discrimination connected to voting; (2) racially polarized voting patterns; (3) whether systemicdiscrimination disproportionately affects minority group's access to the polls; (4) racial appeals in political campaigns; (5) the number of minorities in public office; (6) officials' responsiveness to the needs of minority groups; and (7) the importance of the policy underlying the challenged restriction. Id. at 36–37, 106 S.Ct. 2752 (citing S. Rep. No. 97-417, at 28–29).

Here, each of the listed factors weigh in DNC's favor.

1

Courts are to consider "the extent of any history of official discrimination in the state ... that touched the right of the members of the minority group to register, to vote, or otherwise to participate in the democratic process." Gingles , 478 U.S. at 36–37, 106 S.Ct. 2752 (1986) (quoting S. Rep. No. 97-417, at 28–29 ). The district court classified this factor as a "mixed bag," but the evidence—even as it was described by the court—points overwhelmingly in the DNC's favor.

The district court recognized Arizona's "history of discrimination against Native Americans, Hispanics, and African Americans" throughout the entirety of its statehood. Reagan , ––– F.Supp.3d at –––– – ––––, 2018 WL 2191664, at *36–38. For example, Native Americans could not legally vote until 1948, when the Arizona Supreme Court held the disenfranchisement of Native Americans unconstitutional. Id. at ––––, 2018 WL 2191664, at *36. From the state's inception until Congress passed the VRA, literacy tests enacted specifically to limit "the ignorant Mexican vote" prevented Hispanics, Native Americans, and African Americans from full participation in the electoral franchise. Id. The state discriminates against minorities in other ways which ultimately limit voting participation, too, particularly by undereducating nonwhite residents and refusing to offer appropriate Spanish translations, practices that continue into the present day and likely serve to widen the racial and ethnic gaps in OOP voting. Id. at ––––, 2018 WL 2191664, at *37.

The district court noted that "discrimination against minorities in Arizona has not been linear." Id. However, the fact that "[d]iscriminatory action has been more pronounced in some periods of state history than others ... [and] each party (not just one party) has led the charge in discriminating against minorities over the years" does not support the district court's conclusion that this factor is inconclusive. Id. at ––––, 2018 WL 2191664, at *38. Rather, despite some advancements, most of which were mandated by courts or Congress, Arizona's history is marred by discrimination. What is more, while evidence of sustained improvement must be considered, "sporadic[ ] and serendipitous[ ]" indicators of improvement are not grounds for discounting a long history of discrimination. Gingles , 478 U.S. at 76, 106 S.Ct. 2752.

Additionally, the district court discounted some evidence on the grounds that "[m]uch of the discrimination that has been evidenced may well have in fact been the unintended consequence of a political culture that simply ignores the needs of minorities." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *38. The results test avoids such a chicken-or-the-egg inquiry. Gingles , 478 U.S. at 63, 106 S.Ct. 2752. When Congress amended the VRA in 1982, it did so in recognition that discrimination need not be intentional to disenfranchise minority groups.

2

Courts are also tasked with considering "the extent to which voting in the elections of the state ... is racially polarized." Gingles , 478 U.S. at 37, 106 S.Ct. 2752 (quoting S. Rep. No. 97-417, at 28–29). The district court correctly concluded that "Arizona has a history of racially polarized voting, which continues today." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *38. This factor was never in dispute.

However, it bears mentioning the degree to which Arizona politics are racially polarized. In reasonably contested elections, 59% of white Arizonans vote Republican, in contrast to 35% of Hispanic Arizonans and an undetermined minority of African American and Native American voters. Arizona politics are even more polarized along the lines of the candidate's ethnicity; in non-landslide district-level contests between a Hispanic Democratic candidate and a white Republican candidate, 84% of Hispanic voters, 77% of Native American voters, 52% of African American voters, and only 30% of white voters select the Hispanic candidate.

3

Similarly, there is no dispute that "members of the minority groups[s] in the state ... bear the effects of discrimination in such areas as education, employment and health, which hinder their ability to participate effectively in the political process[.]" Gingles , 478 U.S. at 37, 106 S.Ct. 2752 (quoting S. Rep. No. 97-417, at 28–29). As the district court noted, "[r]acial disparities between minorities and non-minorities in socioeconomic standing, income, employment, education, health, housing, transportation, criminal justice, and electoral representation have persisted in Arizona." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *38. Although the district court's order only briefly mentions this factor, the evidence is overwhelming. Indeed, compared to white Arizonans, black Arizonans are over twice as likely to live in poverty, Hispanic Arizonans are nearly three times as likely, and Native Americans are almost four times as likely. Id. at ––––, 2018 WL 2191664, at *31.

4

Arizona politicians have a long history of making "overt or subtle racial appeals," and that history extends to the present day. Gingles , 478 U.S. at 37, 106 S.Ct. 2752 (quoting S. Rep. No. 97-417, at 28–29). As the district court noted, candidates have relied on racial appeals since the 1970s. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *38. For example, during Raul Castro's successful gubernatorial run in the 1970s, his opponent's supporters called on the electorate to choose the candidate who "looked like a governor," and a newspaper printed Fidel Castro's face below a headline reading, "Running for governor of Arizona." Id.

More recently, too, during his winning campaign for State Superintendent of Public Office, John Huppenthal, a white candidate running against a Hispanic competitor, ran an ad touting that he was "one of us," that he was opposed to bilingual education, and that he "will stop La Raza," an influential Hispanic civil rights organization. Id. And when former Maricopa County Attorney Andrew Thomas ran for governor, one of his ads included an image of the Mexican flag with a red line striking through it. Id. Moreover, as I discuss at length below, racial appeals were made specifically in regard to H.B. 2023. These racial appeals "lessen to some degree the opportunity of [minorities] to participate effectively in the political processes and to elect candidates of their choice." Gingles , 478 U.S. at 40, 106 S.Ct. 2752.

5

Also relevant is "the extent to which members of the minority group[s] have been elected to public office in the jurisdiction." Gingles , 478 U.S. at 37, 106 S.Ct. 2752 (quoting S. Rep. No. 97-417, at 28–29). The district court noted that "the disparity in the number of minority elected officials in Arizona has declined." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *39. However, a "decline" does not translate to equity. Gingles , 478 U.S. at 76, 106 S.Ct. 2752. While nonwhites compose 44% of Arizona's total population, only two minority statespersons—one Hispanic governor in 1974 and one African American Corporation Commissioner in 2008—have been elected to statewide positions in the last 50 years. Id. There are currently no minorities in statewide office. Minorities hold only 22% of state congressional seats and 9% of judgeships.

Minorities are seriously underrepresented in public office in Arizona, and the problem is most severe at the statewide level. Significantly, because Arizona could not be required to count votes for which an OOP voter is not qualified to vote, Arizona's practice of wholly discarding OOP ballots only has an effect on top-of-the-ticket races, where representation is at its lowest.

6

A § 2 claim is likelier to succeed where "there is a significant lack of responsiveness on the part of elected officials to the particularized needs of the members of the minority group[s]." Gingles , 478 U.S. at 37, 106 S.Ct. 2752 (quoting S. Rep. No. 97-417, at 28–29). The district court found that DNC's evidence was "insufficient to establish a lack of responsiveness on the part of elected officials to particularized needs of minority groups." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *39. It bolstered its conclusion with evidence that the Arizona Citizens Clean Elections Commission engages in outreach with minority populations, but engagement by one entity is not conclusive, especially in the face of overwhelming evidence of government nonresponsiveness.

The district court ignored evidence that Arizona underserves minority populations. For example, it failed to recognize that Arizona was the last state in the nation to join the Children's Health Insurance Program, which may explain, in part, why forty-six states have better health insurance coverage for children. Similarly, it ignored evidence that Arizona's public schools are drastically underfunded; in fact, in 2016 Arizona ranked 50th among the states and the District of Columbia in per pupil spending on public elementary and secondary education. Given the well-documented evidence that minorities are likelier to depend on public services—evidence generally credited by the district court—Arizona's refusal to provide adequate state services demonstrates its nonresponsiveness to minority needs.⁴

4 Rather than discuss the evidence supporting DNC, the district court simply discredited the testimony of one of DNC's experts, Dr. Allan Lichtman, on the grounds that he "ignored various topics that are relevant to whether elected officials have shown responsiveness, and he did not conduct research on the issues in Arizona." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *39. However, the court also found "Dr. Lichtman's underlying sources, research, and statistical information [to be] useful." Id. at ––––, 2018 WL 2191664, at *2. Thus, my analysis incorporates only Dr. Lichtman's "underlying sources, research, and statistical information."

Indeed, the district court's finding is directly contradicted by its recognition, later in its order, that Arizona has a "history of advancing partisan objectives with the unintended consequence of ignoring minority interests." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *43. And, as I discuss below, there is significant specific evidence of the legislature's disregard for minority needs in the legislative history leading to the passage of H.B. 2023. The district court failed to consider important facts and overstated the significance of one minor item of evidence. It clearly erred in finding that this factor does not support DNC. See, e.g. , Myers v. United States , 652 F.3d 1021, 1036 (9th Cir. 2011) (holding that the district court clearly erred when it ignored evidence contradicting its findings).

7

Courts may also consider "whether the policy underlying the state ... practice ... is tenuous." Gingles , 478 U.S. at 37, 106 S.Ct. 2752 (quoting S. Rep. No. 97-417, at 28). In its analysis of this factor, the district court erroneously misstated the inquiry as whether the precinct-based system—rather than the practice of wholly discarding OOP votes—is justified. Finding the precinct-based system well-supported, the district court determined only that "Arizona's policy to not count OOP ballots is one mechanism by which it strictly enforces this system to ensure that precinct-based counties maximize the system's benefits." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *39. However, the district court attempted no further explanation, fully adopting the state's explanation for its practice of discarding votes without considering its logic.

Arizona's OOP policy does not serve any purpose beyond administrative ease. Simply put, it takes fewer resources to count fewer ballots. There is no indication that there is any correlation between the precinct-based model and the OOP policy. Because the analysis of this factor is essentially no different than the analysis under step two of the Anderson / Burdick test, I will not discuss it at length here. Because it misstated DNC's challenge, the district court clearly erred in its finding regarding the justifications for the OOP policy. There is no indication that the precinct-based electoral scheme runs more effectively because Arizona refuses to count OOP votes.

8

Summing up its analysis, the district court found that "[some] of the germane Senate Factors ... are present in Arizona and others are not." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *40. Because DNC showed that each of the relevant factors was satisfied, the district court's characterization of the evidence was clearly erroneous.

Further, the district court took issue with the Senate Factors themselves, writing that DNC's "causation theory is too tenuous to support [its] VRA claim because, taken to its logical conclusion, virtually any aspect of a state's election regime would be suspect as nearly all costs of voting fall heavier on socioeconomically disadvantaged voters." Id. However, the results test was not on trial here; Congress specifically amended the VRA in response to such concerns. Gingles , 478 U.S. at 43–44, 106 S.Ct. 2752 ("The Senate Report which accompanied the 1982 amendments ... dispositively rejects the position of the plurality in Mobile v. Bolden , 446 U.S. 55, 100 S.Ct. 1519, 64 L.Ed.2d 47 (1980), which required proof that the contested electoral practice or mechanism was adopted or maintained with the intent to discriminate against minority voters.").

DNC demonstrated that Arizona's practice of not counting OOP ballots "results in a denial or abridgement of the right of any citizen of the United States to vote on account of race or color," 52 U.S.C. § 10301(a), and that, "based on the totality of circumstances," members of protected classes "have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice," id. § 10301(b).

III

Arizona's practice of wholly discarding OOP votes also violates the First Amendment, which applies to the states pursuant to the Fourteenth Amendment. In deciding otherwise, the district court made several legal errors, discussed below. Upon correcting the district court's errors and applying the Anderson / Burdick test to the uncontested facts, the record compels a contrary conclusion. See United States v. Silverman , 861 F.2d 571, 576 (9th Cir. 1988) (citation omitted) (clear error standard met when appellate court is left with the "definite and firm conviction" that a mistake was made). Arizona unconstitutionally infringes upon the right to vote by disenfranchising voters unable to find or travel to the correct precinct, even as to those contests for which the voter is qualified to vote.

The First and Fourteenth Amendments protect individual voting rights by limiting state interference with those rights. Reynolds v. Sims , 377 U.S. 533, 554–55, 84 S.Ct. 1362, 12 L.Ed.2d 506 (1964) ; Tashjian v. Republican Party of Conn. , 479 U.S. 208, 217, 107 S.Ct. 544, 93 L.Ed.2d 514 (1986). While "the right[s] to vote in any manner and ... to associate for political purposes" are not "absolute," Burdick v. Takushi , 504 U.S. 428, 433, 112 S.Ct. 2059, 119 L.Ed.2d 245 (1992), neither is the state's constitutionally designated authority to regulate the "Times, Places and Manner of holding Elections for Senators and Representatives," U.S. Const. art. I, § 4, cl. 1 ; Williams v. Rhodes , 393 U.S. 23, 29, 89 S.Ct. 5, 21 L.Ed.2d 24 (1968) (a state's power to regulate elections is "subject to the limitation that [it] may not be exercised in a way that violates other ... provisions of the Constitution."). Thus, "[t]he power to regulate the time, place, and manner of elections does not justify, without more, the abridgment of fundamental rights, such as the right to vote." Tashjian , 479 U.S. at 217, 107 S.Ct. 544.

Courts apply the Anderson / Burdick test, a "flexible" balancing test, to determine whether a voting regulation runs afoul of the First Amendment right to associate. Burdick , 504 U.S. at 434, 112 S.Ct. 2059. The Court must "weigh 'the character and magnitude of the asserted injury to the rights ... that the plaintiff seeks to vindicate' against 'the precise interests put forward by the State as justifications for the burden imposed by its rule,' taking into consideration 'the extent to which those interests make it necessary to burden the plaintiff's rights.' " Id. (quoting Anderson v. Celebrezze , 460 U.S. 780, 789, 103 S.Ct. 1564, 75 L.Ed.2d 547 (1983) ). There is no substitute for the "balancing and means-end fit framework" required under Anderson / Burdick ; even if a burden is minimal, it must be justified. Pub. Integrity All., Inc. v. City of Tucson , 836 F.3d 1019, 1025 (9th Cir. 2016) (en banc).

A

The burden imposed by Arizona's refusal to count OOP votes is severe. The district court and the majority mischaracterize that burden as the burden of complying with the State's general requirement that individuals vote in their assigned precinct. However, the burden here is the burden of disenfranchisement suffered by those voters whose votes are discarded even as to those elections in which the voter is qualified to vote. DNC brought suit alleging that Arizona's practice of discarding OOP ballots unconstitutionally infringes upon individual voting rights. They sought an injunction barring Arizona from continuing that practice. They did not challenge Arizona's precinct-based system in its entirety.

1

The defendants and intervenors rely on semantics, casting the discarding of OOP ballots as the "consequence" of Arizona's precinct system. However, wholly discarding OOP ballots is not a fundamental requirement of—or even a logical corollary to—a precinct-based model. Instead, Arizona's practice of discarding such ballots is exactly that—a practice. And it can change.⁵

5 Indeed, the district court determined in its analysis of standing, which has not been contested on appeal, that the alleged injury—not counting OOP ballots—is redressable. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *10.

The district court legally erred when it restated the burden along the lines urged by the defendants and intervenors.⁶ Concluding that the burden was that of voting in the correct precinct, the district court determined that Arizona's voters are themselves partially responsible for any burden because they are so likely to change residences and to rent rather than own their homes. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *22. However, if such a consideration were permissible, a poll tax could be upheld on the grounds that poor voters could simply earn more money or spend the money that they do earn differently—propositions that have, thankfully, been rejected. See Harper v. Virginia State Board of Elections , 383 U.S. 663, 86 S.Ct. 1079, 16 L.Ed.2d 169 (1966).

6 I respectfully disagree with the majority that the district court rightly restated DNC's challenge because "under DNC's theory, a state could not enforce even a rule requiring registration, because the state's failure to count the vote of a non-registered voter would 'disenfranchise' the noncompliant voter." Op. 724. The Anderson / Burdick test is a balancing test. If a basic registration requirement imposes a burden on voters—and it does—it will still be upheld if that burden is justified—and it is. DNC has merely asked us to apply the Anderson / Burdick framework to its challenge; it has not asked for a per se rule striking any policy or law under which votes go uncounted.

The court also rejected DNC's challenge because "there is no evidence that it will be easier for voters to identify their correct precincts if Arizona eliminated its prohibition on counting OOP ballots." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *23. But the problem is not with the voters, who are dealing with a system insensitive to their needs; the problem is with an electoral system that refuses to acknowledge and respond to the needs of the State's voting population. A democracy functions only to the degree that it fosters participation.

The district court also legally erred when it equated Arizona's policy of discarding OOP votes with similar policies in other states, policies which were not on trial in this lawsuit. Voting rights claims demand an "intensely local appraisal." Gingles , 478 U.S. at 78, 106 S.Ct. 2752 (quoting White v. Regester , 412 U.S. 755, 769, 93 S.Ct. 2332, 37 L.Ed.2d 314 (1973) ). What is more, the constitutionality of these other states' policies has not been affirmatively decided. Thus, the fact that those other states also have policies of not counting votes cast OOP is not indicative of the constitutionality of Arizona's policy.

Thus, the district court erred as a matter of law in determining that "[t]hough the consequence of voting OOP might make it more imperative for voters to correctly identify their precincts, it does not increase the burdens associated with doing so." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *22. The burden identified by DNC and faced by the voter is disenfranchisement.

2

The burden is severe. Because the district court misstated the burden, it also miscalculated its severity. For example, the district court determined that the burden is slight based on its finding that "there is no evidence that it will be easier for voters to identify their correct precincts if Arizona eliminated its prohibition on counting OOP ballots." Id. at ––––, 2018 WL 2191664, at *23. But that reasoning turns the appropriate legal framework on its head.

Under the first prong of the Anderson / Burdick test, the issue is the severity of the burden faced by voters whose ballots are discarded because they voted OOP. Pub. Integrity Alliance , 836 F.3d at 1024 n.2 ("[C]ourts may consider not only a given law's impact on the electorate in general, but also its impact on subgroups, for whom the burden, when considered in context, may be more severe."). Perhaps Arizona's electoral scheme justifies that burden, no matter its severity. If so, however, that determination comes in under step two of the Anderson / Burdick analysis.

For those whose votes go uncounted, "there can be no do-over and no redress." League of Women Voters , 769 F.3d at 247. To determine the burden, the Court looks not to the voters unaffected by the practice, as the district court did, Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *21 ("Arizona's rejection of OOP ballots ... has no impact on the vast majority of Arizona voters."), but to those who suffer the burden, Crawford v. Marion Cty. Election Bd. , 553 U.S. 181, 186, 128 S.Ct. 1610, 170 L.Ed.2d 574 (2008) (plurality opinion); Pub. Integrity All. , 836 F.3d at 1024 n.2. And those voters are effectively rendered unable to vote in elections for which they are qualified and in which they cast otherwise legitimate ballots. There is no burden more severe in the voting rights context.

However, even if the district court had properly stated the burden alleged, its ultimate finding would be clearly erroneous. The district court found that Arizona makes it easy for voters to find their precincts. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *23. The district court's finding is inconsistent with the evidence presented and generally credited by the court.

The government bears responsibility for the high rate of OOP voting. First, precincts appear to change polling locations and practices even more often than residents change homes. Id. at ––––, 2018 WL 2191664, at *22 ("[I]n Maricopa County, between 2006 and 2008 at least 43 percent of polling locations changed from year to the next[.]"). Second, polling places are often in counterintuitive locations, far from some residents' homes. Id. And third, the district court noted (and did not discredit) evidence that election workers fail to inform voters that they are in the wrong precinct and that a provisional ballot will not be counted. Id. Thus, the district court clearly erred in determining that Arizona does all it should to prevent OOP voting.

B

The severe burden faced by OOP voters is not outweighed by a sufficiently important government interest. Pub. Integrity All. , 836 F.3d at 1024. Because the district court misstated the burden, it also overstated the government interest by focusing on the "numerous and significant advantages" of a precinct-based voting model. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *24. The inquiry should instead be whether the state can justify the interests served by the challenged practice of not counting OOP ballots. It cannot.

As the district court itself found, "[c]ounting OOP ballots is administratively feasible." Id. at ––––, 2018 WL 2191664, at *25. This is demonstrated by: (1) the methods used by the 20 states that use a precinct-based system and nonetheless count OOP ballots; and (2) Arizona's readily transferable method "to process certain types of ballots that cannot be read by an optical scan voting machine" and "some provisional ballots cast by voters who are eligible to vote in federal elections, but whom Arizona does not permit to vote in state elections." Id. Certainly, Arizona can count the votes cast by all qualified voters.

The district court determined that, although OOP votes could be counted, Arizona nonetheless could justify its policy on the basis of assumptions regarding what could happen if the state counted all of the ballots that it received. Voters may "decide to vote" out of precinct or "incorrectly believe that they can vote at any location and receive the correct ballot." Id. Worse, they could "be nefariously directed to vote elsewhere." Id. This reasoning is illogical and unsupported by the facts. There is no demonstrated increase in OOP voting in states where those votes are counted than in Arizona (where, of course, OOP voting is at its highest level). And "nefarious" interests would be far better served by misdirecting voters if their out-of-precinct vote would not be counted at all than if it were partially tallied.⁷

7 Under the current system, for example, a Democrat could conceivably misdirect likely Republican voters to the wrong precinct in order to render their ballots null. However, if OOP ballots counted, the Democrat would have less incentive, as the Republicans' choices for statewide and federal office would still register.

Arizona's interest in administrative ease does not justify the severe burden of disenfranchisement. I would hold Arizona's practice of discarding OOP ballots unconstitutional.

IV

Next, DNC challenges a recently enacted law, H.B. 2023, which criminalizes most ballot collection. Under the law, a person who collects another's ballot commits a felony unless the collector is an official engaged in official duties or the voter's family member, household member, or caregiver. Ariz. Rev. Stat. § 16-1005(H)–(I).

H.B. 2023 was not Arizona's first attempt to limit ballot collection. Prior to Shelby County v. Holder , 570 U.S. 529, 133 S.Ct. 2612, 186 L.Ed.2d 651 (2013), Arizona was subject to the VRA's § 5 preclearance requirements. In 2011, Arizona passed S.B. 1412, which criminalized the collection of more than ten ballots by any one individual. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *42. Arizona submitted the bill to the DOJ for preclearance, and the DOJ "precleared all provisions except for the provision regulating ballot collection," about which the DOJ requested further information in order to ensure that the provision had neither the purpose nor the effect of limiting minority participation in voting. Id. Arizona did not proffer the requested information, instead withdrawing the provision before formally repealing the law. Id. With good reason: the State Elections Director, who helped draft the bill, told the DOJ that the law was "targeted at voting practices ... in predominantly Hispanic areas" and that state officials were expecting § 5 review. Withdrawing a provision was not standard procedure for Arizona, which fully or partially withdrew only 6 of its 773 preclearance provisions. Id. In 2013, the legislature tried a new approach. It passed H.B. 2305 "along nearly straight party lines in the waning hours of the legislative session." Id. The law "banned partisan ballot collection and required other ballot collectors to complete an affidavit stating that they had returned the ballot." Id. The public outcry was immediate, with "citizen groups organiz[ing] a referendum effort and collect[ing] more than 140,000 signatures to place H.B. 2305 on the ballot for a straight up-or-down vote" in the next election. Id. "Rather than face a referendum," which would have barred further related legislation without a supermajority vote, "Republican legislators again repealed their own legislation along party lines." Id. At the time, then-State Senator Michele Reagan (now Secretary of State and defendant to this action), who sponsored the bill, stated that the legislature would reintroduce the bill, but in smaller fragments. Id.

As the district court noted, H.B. 2023 was passed not only "on the heels of" these earlier bills, but also "in the context of racially polarized voting" and "increased use of ballot collection as a Democratic [get-out-the-vote] strategy in ... minority communities." Id. at ––––, 2018 WL 2191664, at *41. Legislators supporting the bill were particularly motivated by two items of evidence: the wildly irrational testimony of then-State Senator Don Shooter, and a racist video prepared by former Maricopa Republican Party Chair A.J. LaFaro, in which LaFaro claims that a Hispanic man engaged in a lawful get-out-the-vote ballot collection effort is a "thug" breaking the law. Id. at –––– – ––––, ––––, 2018 WL 2191664, at *38–39, *41.

DNC brings three challenges to H.B. 2023. It argues that the provision was motivated by racial animus, in violation of the Fourteenth and Fifteenth Amendments and § 2 of the VRA. It claims that it has a discriminatory effect, also in violation of § 2. And, finally, it contends that the law unreasonably burdens voters' First Amendment rights. I agree on all counts and would hold the provision invalid under the VRA and the United States Constitution.

V

H.B. 2023 was enacted for the purpose of suppressing minority votes, in violation of § 2 of the VRA and the Fourteenth and Fifteenth Amendments. Although lawmakers were also motivated by partisanship, their intent to reduce the total number of Democratic votes does not render the law constitutional.

Under the Fourteenth and Fifteenth Amendments and § 2 of the VRA, a law passed with the intent to discriminate against racial or ethnic minorities cannot stand. The law imposes a high burden on plaintiffs, who must show "[p]roof of racially discriminatory intent or purpose." Vill. of Arlington Heights v. Metro. Hous. Dev. Corp. , 429 U.S. 252, 265, 97 S.Ct. 555, 50 L.Ed.2d 450 (1977). Voting regulations are unconstitutional when they are " 'conceived or operated as purposeful devices to further racial discrimination' by minimizing, cancelling out or diluting the voting strength of racial elements in the voting population." Rogers , 458 U.S. at 617, 102 S.Ct. 3272 (quoting Whitcomb v. Chavis , 403 U.S. 124, 149, 91 S.Ct. 1858, 29 L.Ed.2d 363 (1971) ). A plaintiff need not show that officials acted solely to further a racially motivated agenda, Arlington Heights , 429 U.S. at 265, 97 S.Ct. 555, but the ultimate issue is whether "the legislature enact[ed] a law 'because of,' and not 'in spite of,' its discriminatory effect," N.C. State Conf. of NAACP v. McCrory , 831 F.3d 204, 220 (4th Cir.2016) (quoting Pers. Adm'r of Mass. v. Feeney , 442 U.S. 256, 279, 99 S.Ct. 2282, 60 L.Ed.2d 870 (1979) )."Necessarily, an invidious discriminatory purpose may often be inferred from the totality of the relevant facts...." Rogers , 458 U.S. at 618, 102 S.Ct. 3272 (quoting Washington v. Davis , 426 U.S. 229, 242, 96 S.Ct. 2040, 48 L.Ed.2d 597 (1976) ). "Thus determining the existence of a discriminatory purpose 'demands a sensitive inquiry into such circumstantial and direct evidence of intent as may be available.' " Id. (quoting Arlington Heights , 429 U.S. at 266, 97 S.Ct. 555 ). Courts consider the Arlington Heights factors, a non-exhaustive list of considerations, to determine whether a law was enacted to satisfy a motive to discriminate: (1) the historical background and sequence of events leading to enactment; (2) substantive or procedural departures from the normal legislative process; (3) relevant legislative history; and (4) the impact of the law on a particular racial group. Arlington Heights , 429 U.S. at 266–68, 97 S.Ct. 555.

Here, all four factors weigh in favor of DNC.

A

The historical background of a challenged provision is an important evidentiary source, "particularly if it reveals a series of official actions taken for invidious purposes." Id. at 267, 97 S.Ct. 555. As the district court recognized, "H.B. 2023 emerged in the context of racially polarized voting, increased use of ballot collection as a Democratic [get-out-the-vote] strategy in low-efficacy minority communities, and on the heels of several prior efforts to restrict ballot collection." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *41. And as discussed below, in my analysis of § 2's results test, a longer view of history similarly weighs in favor of DNC. Quite simply, the historical background suggests that the restriction was enacted in order to prevent minority ballots from being counted.

The fact that the minority votes would help Democratic candidates does not alter the analysis. See id. (suggesting that because "some individual legislators and proponents were motivated in part by partisan interests"⁸ they were not motivated by racially discriminatory interests). Indeed, if that were the case, consideration for racially polarized voting patterns—a constant in VRA and constitutional voting regulation challenges—would be impermissible or weigh in favor of upholding a regulation. By nature of the political process, an unconstitutionally discriminatory voting regulation is a law enacted by the political party in power in order to maintain power by preventing minorities from voting, assuredly because they belong to the other political party.

8 The majority concludes that the district court "did not err in giving little weight to evidence that 'some individual legislators and proponents were motivated in part by partisan interests.' " Op. 720 (quoting Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *43 ). But the court did not discredit this evidence. Rather it relied on it to show proof of nondiscrimination.

The first Arlington Heights factor suggests discriminatory motive.

B

Under Arlington Heights courts consider "the defendant's departures from its normal procedures or substantive conclusions." Pac. Shores Props., LLC v. City of Newport Beach , 730 F.3d 1142, 1159 (9th Cir. 2013) (citing Arlington Heights , 429 U.S. at 266–68, 97 S.Ct. 555 ). The district court recognized that "the circumstances surrounding" H.B. 2023 were "somewhat suspicious." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *42. This is an understatement. H.B. 2023 flowed directly out of the Arizona legislature's two prior attempts to limit ballot collection.⁹ The law enacted does not cure the intent to discriminate demonstrated by its precursors; rather, H.B. 2023 was part of the same general strategy of limiting the minority vote by limiting ballot collection.

9 While it is true that discriminatory intent as to an earlier law does not necessarily carry through to any other provision on the subject, Op. 721–22, we do not have to suspend common sense. The recency of the earlier provisions, coupled with relevant public statements and the weak legislative record supporting H.B. 2023, places H.B. 2023 on one end of an unbroken line beginning just a few years earlier with S.B. 1412.

This Arlington Heights factor suggests discriminatory motive.

C

"The legislative ... history may be highly relevant, especially where there are contemporary statements by members of the decisionmaking body...." Arlington Heights , 429 U.S. at 267, 97 S.Ct. 555. The district court found evidence of racial animus in the legislative history but discounted its significance, suggesting that any initial discriminatory motive was cured because some legislators acted either out of self-interest or an unfounded but sincere belief that voter fraud was likely.

The district court's reasoning is clearly erroneous. First, partisan self-interest cannot absolve discriminatory intent. If we were to allow racially motivated voting schemes whenever those schemes serve partisan interests, the exception would swallow the rule, and there would be no prohibition on enacting laws in order to discriminate. Second, the sincerity of the legislators' belief in a wholly theoretical risk of voter fraud is—as the district court itself suggested—indicative of discriminatory intent. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *41 (describing legislators' motives as "perhaps implicitly informed by racial biases").

Moreover, the district court's own specific factual findings belie its ultimate conclusion on the third Arlington Heights factor. The district court determined that the proponents of H.B. 2023 voted for the bill in response to two pieces of evidence: (1) the "demonstrably false," "unfounded and often farfetched allegations of ballot collection fraud" made by former Arizona State Senator Don Shooter; and (2) a "racially-tinged" video created by Maricopa County Republican Chair A.J. LaFaro (the "LaFaro Video"). Id. Because there was "no direct evidence of ballot collection fraud ... presented to the legislature or at trial," the district court understood that Shooter's allegations and the LaFaro Video were the reasons the bill passed. Id. ("Shooter's allegations and the LaFaro Video were successful in convincing H.B. 2023's proponents that ballot collection presented opportunities for fraud that did not exist for in-person voting....").

Both of these evidentiary items demonstrate racial animus. As the district court made clear, Senator Shooter's testimony regarding the existence and prevalence of voter fraud was not only incorrect but in fact "unfounded and often farfetched." Id. If Senator Shooter was sincere, his distorted view of reality is explainable only by what the district court downplayed as being "implicitly informed by racial biases,"—or, in starker terms, by racism. Id. An unfounded and exploited fear that members of minority groups are "engage[d] in nefarious activities," id. , supports a finding of racial animus. And if Senator Shooter was insincere, he purposefully distorted facts in order to prevent Hispanics—who generally preferred his opponent—from voting. Id. ("Due to the high degree of racial polarization in his district, Shooter was in part motivated by a desire to eliminate what had become an effective Democratic [get-out-the-vote] strategy. ... Indeed, Shooter's 2010 election was close: he won with 53 percent of the total vote, receiving 83 percent of the non-minority vote but only 20 percent of the Hispanic vote.").

The LaFaro Video is even more damning. The video shows a Hispanic man, a volunteer with a get-out-the-vote organization, delivering early ballots to the polls. The video is itself wholly mundane; it is eight soundless minutes of a man moving completed ballots from a cardboard box to the ballot box. It markedly "did not show any obviously illegal activity." Id. at ––––, 2018 WL 2191664, at *39. However, LaFaro provided a voice-over narration, "includ[ing] statements that the man was acting to stuff the ballot box; that LaFaro did not know if the person was an illegal alien, a dreamer, or citizen, but knew that he was a thug; and that LaFaro did not follow him out to the parking lot to take down his tag number because he feared for his life." Id. at ––––, 2018 WL 2191664, at *38. It is LaFaro's narration—not the dull raw material showing a Hispanic man dropping off ballots—that "became quite prominent in the debates over H.B. 2023." Id. at ––––, 2018 WL 2191664, at *39. As the district court recognized, the LaFaro Video evidences racial animus.

After recognizing the existence of discriminatory intent, the district court seems to have determined that intent was later cured because the bill "found support among some minority officials and organizations" and because some lawmakers opposed H.B. 2023 for reasons other than that it being grounded in racial discrimination. Id. at ––––, 2018 WL 2191664, at *41. The district court's reasoning is incorrect. As the Supreme Court has stated, there is no room for judicial deference "[w]hen there is ... proof that a discriminatory purpose has been a motivating factor in the decision." Arlington Heights , 429 U.S. at 265–66, 97 S.Ct. 555.

Moreover, the district court was wrong to determine that a law is not racially motivated if any people of color support it. Rather, the evidence that particular Hispanic and African American Arizonans supported H.B. 2023 simply demonstrates that people of color have diverse interests, some of which may outweigh potential concerns that a law was enacted with the intent to discriminate. And although one lawmaker "testified that she has no reason to believe H.B. 2023 was enacted with the intent to suppress Hispanic voting," the district court also recognized that "some Democratic lawmakers accused their Republican counterparts of harboring partisan or racially discriminatory motives." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *41. Again, a diversity of perspectives is neither surprising nor particularly telling, especially when the operative legal test recognizes that a law may be unconstitutionally discriminatory even if it is not driven solely by racial animus: "legislators ... are properly concerned with balancing numerous competing considerations." Arlington Heights , 429 U.S. at 265, 97 S.Ct. 555.

The district court's concerns were also assuaged because Shooter's "demonstrably false" allegations and "the racially-tinged LaFaro Video ... spurred a larger debate in the legislature about the security of early mail voting as compared to in-person voting." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *41. The court's finding is neither here nor there. The legislature did not act to limit all early voting, but it targeted a specific practice known to be popular among minority voters, despite the absence of any evidence that ballot collection was less secure than other early voting methods.

This Arlington Heights factor weighs in favor of DNC.

D

"The impact of the official action whether it 'bears more heavily on one race than another' " is "important" to the analysis of whether a law was enacted to serve a discriminatory motive. Arlington Heights , 429 U.S. at 266, 97 S.Ct. 555 (quoting Davis , 426 U.S. at 242, 96 S.Ct. 2040.) The district court wholly failed to measure H.B. 2023's impact on minority voters in its discussion of Arlington Heights . Rather, it counterintuitively concluded that concerns about the law's effect on minority groups "show[ ] only that the legislature enacted H.B. 2023 in spite of its impact on minority [get-out-the-vote] efforts, not because of that impact." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *43. The district court's determination is not only illogical but also out of place in its discussion of the fourth Arlington Heights factor. As I will discuss in my analysis of the § 2 results test, H.B. 2023 disproportionately affects minority voters.

Like the first three factors considered, the fourth and final factor supports a conclusion that the law is motivated by racial animus. Thus, under the purpose test of § 2 of the VRA and the Fourteenth and Fifteenth Amendments, H.B. 2023 cannot survive.

VI

Like Arizona's practice of discarding OOP votes, H.B. 2023 imposes an unlawful discriminatory burden on minority voters. As discussed above, § 2 of the VRA provides that "[n]o voting ... standard, practice, or procedure shall be imposed or applied ... 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." 52 U.S.C. § 10301(a).

Under the results test, "[t]he essence of a § 2 claim is that a certain electoral law, practice, or structure interacts with social and historical conditions to cause an inequality in the opportunities enjoyed by [minority] and white voters to elect their preferred representatives." Gingles , 478 U.S. at 47, 106 S.Ct. 2752. The test is one of the "totality of circumstances." 52 U.S.C. § 10301(b) ; Gingles , 478 U.S. at 43, 106 S.Ct. 2752. In this instance, the totality of the circumstances conclusively demonstrates that H.B. 2023 disproportionately burdens minority voters, and that burden can be traced directly to historical and social conditions of discrimination. League of Women Voters , 769 F.3d at 240.

A

The first prong of the results test "inquires about the nature of the burden imposed and whether it creates a disparate effect." Veasey , 830 F.3d at 244.

The district court suggested that DNC's challenge ought to fail at step one because of a lack of quantitative evidence, but it ultimately based its disposition on its determination that "Plaintiffs' circumstantial and anecdotal evidence is insufficient to establish a cognizable disparity under § 2." Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *31. The district court erred as a matter of law when it determined that although, "prior to H.B. 2023's enactment minorities generally were more likely than non-minorities to give their early ballots to third parties," id. , it could not find for DNC because it could not "speak in more specific or precise terms than 'more' or 'less.' " Id. at ––––, 2018 WL 2191664, at *33.While it is true that a plaintiff bears the burden of demonstrating the existence and extent of a disparity, Gonzalez v. Arizona , 677 F.3d 383, 406 (9th Cir. 2012) (en banc), it is not true that the plaintiff is required to do so with statistical evidence, 52 U.S.C. § 10301(b) (providing that relevant inquiry is into "the totality of circumstances"). The question is simply whether members of the affected ethnic and racial minority groups "have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice." Id. § 10301(b).

The evidence presented at trial weighed overwhelmingly in DNC's favor. For political and socioeconomic reasons, H.B. 2023 is far likelier to affect African American, Hispanic, and Native American Arizonan voters than white voters. As the district court recognized, minority voters used ballot collection services more than white voters. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *31. The disparity is not caused solely by geography, as the socioeconomic conditions leading minority voters to depend on ballot collection "exist in both urban and rural areas." Id. at ––––, 2018 WL 2191664, at *32.

The witnesses with direct experience in collecting ballots, without exception, testified at trial that racial and ethnic minority voters were far likelier to vote with the help of ballot collection services. For example, one individual who worked in several ballot collection groups testified that "the overwhelming majority" of voters with whom he worked were Hispanic or African American. Another stated that the "vast majority of the ballot pickups" done by the Maricopa County Democratic Party are in "[m]ajority-minority districts." Democratic State Senator Martin Quezada described requests for ballot collection, testifying that "[t]he large majority of those requests came from the lower income and the neighborhoods that were a larger percentage Latino than others."

No one had a clear statistical analysis of the disparity. Nor could anyone, as the state would be the only entity in a position to collect such evidence, and it has not done so. However, one ballot collector testified as to what she termed a "case study" showing the extent of the disparity. In 2010, she and her fellow organizers collected "somewhere south of 50 ballots" in one particular district. The area was redistricted before the next election to add a heavily Hispanic neighborhood, Sunnyslope, and in 2012, the organization "pulled in hundreds of ballots, vast majority from that Sunnyslope area."

Not only is there no evidence in the record of any significant reliance on ballot collection by white voters, but the evidence is also replete with evidence explaining why a disparity is natural. For example, in rural Somerton and San Luis, both of which are over 95% Hispanic, voters lack home mail service and are unlikely to have access to reliable transportation. Id. at ––––, 2018 WL 2191664, at *32. In urban areas, too, Hispanic voters are less likely to have access to mail services and, due to mail theft, less likely to trust mail-in voting. Id.

As the district court rightfully noted, the "problems are particularly acute in Arizona's Native American communities." Id. Indeed, uncontroverted expert testimony showed that "the majority of Native Americans in non-metropolitan Arizona do not have home mail delivery" and that non-Hispanic white voters are 350% more likely to have home mail service than Native American voters. Id. In fact, only 18% of Native Americans outside of Pima and Maricopa Counties have home mail service—in contrast to 86% of non-Hispanic whites. And residents of sovereign nations often must travel 45 minutes to 2 hours just to get to a mailbox. In the district court's words, "for many Native Americans living in rural locations, especially on reservations, voting is an activity that requires the active assistance of friends and neighbors." Id.

In contrast, none of the evidence discussed by the district court suggested that there was no disparate burden or that any such disparity was minor. In short, the district court summarized the overwhelming evidence showing a disparate burden and then concluded that because it couldn't pin down the difference with exactitude, it could not find for DNC.

The district court also suggested that it could not find for DNC because too few voters rely on ballot collection for a restriction on ballot collection to matter. Id. at –––– – ––––, 2018 WL 2191664, at *33–34. To the degree that this finding matters, it is a consideration under the Anderson / Burdick analysis, not under step one of the VRA analysis. Moreover, the district court's analysis ignores that the VRA exists to protect minority groups—those groups least likely to have their voices heard. Thus, the precise number of affected voters is not particularly helpful.

Because it misstated the legal requirements for establishing a disparity, the district court clearly erred in concluding that DNC failed to meet their burden. I would hold that H.B. 2023 imposes a disparate burden on members of protected classes.

B

As detailed earlier, within my application of the § 2 results test to the OOP policy, the Senate Factors demonstrate the existence of social and historical conditions of discrimination in Arizona. Those determinations have equal force here, and I will not belabor the point by repeating my analysis here. Instead, I will focus on the ways in which H.B. 2023 is directly connected to those conditions of discrimination.

For example, one of the Senate Factors considers the state's history of racial discrimination. Gingles , 478 U.S. at 36–37, 106 S.Ct. 2752. Not only does Arizona have a history of official discrimination, as I have discussed, but the history of H.B. 2023—passed after one provision was rejected under § 5 of the VRA and after the people of Arizona demonstrated concern with another—powerfully links the statute to that history. Similarly, as to racially polarized voting patterns, as the district court noted, one of the most vocal proponents for criminalizing ballot collection, Senator Shooter, did so in part because he was facing a close election in which Hispanic voters were highly unlikely to vote for him.

Perhaps most significantly, there is direct evidence of racial appeals being made in the context of this very issue. Gingles , 478 U.S. at 36–37, 106 S.Ct. 2752. In the LaFaro video, a Hispanic get-out-the-vote volunteer gives no indication that he is violating election law but is nonetheless described as a "thug" likely to physically harm a white political figure. Reagan , ––– F.Supp.3d at –––– – ––––, 2018 WL 2191664, at *38–39. That video figured "prominently" in public debates about voter fraud and ballot collection, even though it showed no illegal activity. Id. at ––––, 2018 WL 2191664, at *39. The Senate Factors clarify that even "subtle" racial appeals are significant under the § 2 analysis, but the subtext of the LaFaro video does not demand decoding. Gingles , 478 U.S. at 37, 106 S.Ct. 2752 (1986) (quoting S. Rep. No. 97-417, at 28–29 ).

Additionally, the legislative record demonstrates a "significant lack of responsiveness on the part of elected officials to the particularized needs of the members of the minority group[s]." Gingles , 478 U.S. at 37, 106 S.Ct. 2752 (1986) (quoting S. Rep. No. 97-417, at 28–29 ). Legislators were apprised of concerns that H.B. 2023 would place an especial burden on minority voters. Their response? In the words of the bill's sponsor: "not my problem." And in those of another state senator supporting the measure, "I don't know why we have to spoon-fe[e]d and baby them over their vote."

H.B. 2023 "interacts with social and historical conditions to cause an inequality in the opportunities enjoyed by [minority] and white voters to elect their preferred representatives." Gingles , 478 U.S. at 47, 106 S.Ct. 2752. DNC has conclusively met its burden of showing that H.B. 2023 limits African American, Hispanic, and Native American Arizonan voters' ability to fully participate in the political process and to elect representatives of their choice.

VII

Finally, H.B. 2023 cannot be reconciled with the First Amendment, which applies to the states under the Fourteenth Amendment and which guarantees that the right to vote will not be unreasonably burdened. Burdick , 504 U.S. at 434, 112 S.Ct. 2059.

A

The burden is identified by looking to those affected by the challenged provision. Crawford , 553 U.S. at 198, 128 S.Ct. 1610 ("The burdens that are relevant to the issue before us are those imposed on persons who are eligible to vote but do not possess a current photo identification that complies with the requirements."). Here, then, the relevant burden is that faced by individuals who vote with the assistance of others who are not family members, household members, or caregivers.

"[C]ourts may consider not only a given law's impact on the electorate in general, but also its impact on subgroups, for whom the burden, when considered in context, may be more severe." Pub. Integrity All. , 836 F.3d at 1024 n.2. And, indeed, the Court recognized this principle in Crawford by noting that "a somewhat heavier burden may be placed on a limited number of persons." 553 U.S. at 199, 128 S.Ct. 1610. A determination of the severity of that burden takes into account socioeconomic situations. Id. (considering "persons who because of economic or other personal limitations may find it difficult either to secure a copy of their birth certificate or to assemble the other required documentation to obtain a state-issued identification").

Here, there is a heavy burden on, at minimum, Native Americans living in rural Arizona, 82% of whom lack home mail service. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *32. Many of these individuals without home mail access may have serious difficulties getting to the post office due to distance, socioeconomic conditions, and lack of reliable transportation. Id. Additionally, as the district court recognized, the State's definition of a family relationship, codified in H.B. 2023, does not track with family relationships in Indian Country. Id. at ––––, 2018 WL 2191664, at *33.

The district court erred by failing to consider a significant body of evidence demonstrating the burdens faced by voters. The district court wrote that it "ha[d] insufficient evidence from which to measure the burdens on discrete subsets of voters" because it could not determine a precise number of voters that had relied on ballot collection in the past or predict a likely number in the future. Id. at ––––, 2018 WL 2191664, at *14. Its reliance on Crawford for this assertion is legally erroneous. In Crawford , the Court did not set forth a rigorous evidentiary standard requiring the production of quantifiable evidence; instead, the Court simply said that DNC did not produce anything sufficiently reliable to demonstrate who would be burdened or to what degree. 553 U.S. at 200–02, 128 S.Ct. 1610.

DNC presented a much better case than the plaintiffs in Crawford . First, here, unlike in Crawford , the district court did not reject the plaintiff's evidence as "utterly incredible and unreliable." Crawford , 553 U.S. at 200, 128 S.Ct. 1610. Second, also distinguishable from Crawford , here, there is evidence that some will be unable to vote under H.B. 2023. For example, an individual who collected ballots for the Maricopa County Democratic Party testified that even though the organization only collected ballots for voters with "no other option," she nonetheless witnessed its collection of 1,200 to 1,500 ballots. Here, there was no evidentiary failure.

That said, even if the district court properly classified the burden as minimal at step one of the Anderson / Burdick analysis, H.B. 2023 nonetheless fails at step two.

B

H.B. 2023 was and is not supported by the "adequate justification" of "reduc[ing] opportunities for early ballot loss or destruction," Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *40, or of "maintain[ing] public confidence in election integrity," id. at ––––, 2018 WL 2191664, at *18. Rather, the legislative history uncontrovertedly indicates that the best justification offered by the legislators voting for the measure was a generic concern regarding voter fraud—a solution in search of a problem. Even after the bill was passed and a trial was held, the trial court could find "no direct evidence that the type of ballot collection fraud the law is intended to prevent or deter has occurred." Id. ¹⁰ H.B. 2023's foundation is not only shaky, it's illusory.

10 Nor was there any suggestion that legislators had reason to believe that public faith in the system had been shaken, as the district court notes. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *18.
--------

Even if the district court had been correct to classify the burden imposed by H.B. 2023 as minimal, the law does not withstand scrutiny under the First Amendment. "However slight [a] burden may appear, ... it must be justified by relevant and legitimate state interests 'sufficiently weighty to justify the limitation.' " Crawford , 553 U.S. at 191, 128 S.Ct. 1610 (quoting Norman v. Reed , 502 U.S. 279, 288–89, 112 S.Ct. 698, 116 L.Ed.2d 711 (1992) ). " '[E]venhanded restrictions that protect the integrity and reliability of the electoral process itself are not invidious and satisfy the standard.' " Crawford , 553 U.S. 181, 189–90, 128 S.Ct. 1610, 170 L.Ed.2d 574 (quoting Anderson , 460 U.S. at 788, 103 S.Ct. 1564 ). Here, no legitimate interest justifies H.B. 2023.

Crawford is not a blank check for legislators seeking to restrict voting rights with baseless cries of "voter fraud." In Crawford , the Court held that the state's interest in deterring voter fraud was legitimate despite the record's absence of "evidence of any [in-person] fraud actually occurring ... at any time in its history," but the case is distinguishable for at least two reasons. Id. at 194, 128 S.Ct. 1610. First, the voter I.D. restriction considered in Crawford was tied to "the State's interest in counting only the votes of eligible voters," particularly given the extreme disorganization of Indiana's voter rolls. Id. at 196, 128 S.Ct. 1610. On the other hand, the nature of the relationship between the voter and the person submitting a ballot has no similar logical connection to that interest. The same safeguards—e.g., "tamper evident envelopes and a rigorous voter signature verification procedure"—are in place for voters who give their ballots to their sister as for those who participate in a get-out-the-vote effort. Reagan , ––– F.Supp.3d at ––––, 2018 WL 2191664, at *19.

Second, the Court in Crawford was untroubled by its determination that the legislature was motivated by partisanship because it determined that the legislature was also motivated by legitimate concerns. Crawford , 553 U.S. at 204, 128 S.Ct. 1610 ("[I]f a nondiscriminatory law is supported by valid neutral justifications, those justifications should not be disregarded simply because partisan interests may have provided one motivation for the votes of individual legislators."). Here, however, the legislature was motivated by discriminatory intent, as I have discussed.

Moreover, even in the absence of discriminatory intent, given the precision of H.B. 2023 toward Democratic get-out-the-vote operations, "partisan considerations" did not simply "play[ ] a significant role in the decision to enact [the law]" but rather "provided the only justification for [the restriction on ballot collection]." Id. at 203, 128 S.Ct. 1610. In Crawford , the plurality "assume[d]" that such a law would be held unconstitutional. Id. The Court's assumption was based in Harper v. Virginia State Board of Elections , 383 U.S. 663, 86 S.Ct. 1079, in which the Court struck a poll tax requirement. Harper is instructive. There, the Court wrote that "the interest of the State, when it comes to voting, is limited to the power to fix qualifications." Id. at 668, 86 S.Ct. 1079. Just as "[w]ealth, like race, creed, or color, is not germane to one's ability to participate intelligently in the electoral process[,]" neither is political affiliation. Id. at 668, 86 S.Ct. 1079.

VIII

As I said in the previous appeal in this case, voting should be easy in America. It is not in Arizona, and the burden falls most heavily on minority voters. In my view, the district court should have granted an injunction as to both of DNC's challenges. Arizona's practice of discarding OOP votes violates § 2 of the VRA and the First and Fourteenth Amendments. And H.B. 2023 cannot withstand scrutiny under § 2 and the First, Fourteenth, and Fifteenth Amendments.

I respectfully dissent.