OPINION

KAREN NELSON MOORE, Circuit Judge.

Defendants Jon Husted, the Ohio Secretary of State, and Mike DeWine, the Ohio Attorney General, appeal from the district court's order granting Plaintiffs' motion for a preliminary injunction. The district court enjoined the enforcement of Senate Bill 238 ("SB 238") and Secretary of State Directive 2014-17, and ordered the restoration of additional early in-person ("EIP") voting hours as set forth below on the basis that SB 238 and Directive 2014-17 violate the Equal Protection Clause of the Fourteenth Amendment and Section 2 of the Voting Rights Act of 1965. For the reasons set forth below, we AFFIRM the district court's judgment granting the preliminary injunction.

I. BACKGROUND

A. Procedural History

Plaintiffs, Ohio State Conference of the National Association for the Advancement of Colored People et al. ("NAACP"), filed a complaint in the United States District Court for the Southern District of Ohio on May 1, 2014, pursuant to 42 U.S.C. § 1983 and 42 U.S.C. § 1973 challenging the constitutionality and legality of SB 238 and Directive 2014-17. In their complaint for declaratory and injunctive relief, Plaintiffs allege that SB 238 and Directive 2014-06 (now Directive 2014-17) (1) violate the Equal Protection Clause of the Fourteenth Amendment by burdening the fundamental right to vote; and (2) violate Section 2 of the Voting Rights Act of 1965 by disproportionately burdening African American voters' ability to participate effectively in the political process.

On June 30, 2014, Plaintiffs moved for a preliminary injunction to "enjoin the enforcement of ... Senate Bill 238 ... and require Defendant Husted to set uniform and suitable in-person early voting hours for all eligible voters that includes multiple Sundays and weekday evening hours." R. 17 (Pls.' Mot. Prelim. Inj. at 61) (Page ID # 152). Following a hearing on August 11, 2014, the district court granted Plaintiffs' motion for a preliminary injunction on September 4, 2014. R. 72 (D. Ct. Op. and Order at 70) (Page ID # 5917). The district court's order provided as follows:

Id. at 70-71 (Page ID # 5917-18) (footnote omitted).

Defendants timely appealed the district court's order granting a preliminary injunction to Plaintiffs and moved this court to expedite that appeal. After the district court denied Defendants' motion for a stay of that order, Defendants moved this court to stay the order pending appeal. We granted Defendants' motion to expedite the appeal on September 11, 2014, and denied their motion for a stay of the order granting a preliminary injunction to Plaintiffs on September 12, 2014.

The Ohio General Assembly ("General Assembly") filed a motion with the district court on July 11, 2014 to intervene in this case, which the district court denied on July 30. On August 1, the General Assembly filed a notice of appeal of that decision, which appeal is pending under case number 14-3756.

After the district court granted Plaintiffs' motion for a preliminary injunction and after Defendants filed their notice of appeal in the instant case (appeal number 14-3877), the district court granted the General Assembly's renewed motion to intervene, stating that the motion was granted "for the purpose of appeal only." R. 75 (D. Ct. Order Granting General Assembly's Intervention for Appeal) (Page ID # 5954). The General Assembly then filed a notice of appeal that is docketed as appeal number 14-3881. The General Assembly has filed a brief in appeal 14-3881. It has also filed a motion to file a brief instanter in 14-3877, which included an accompanying brief supporting Defendants' appeal in this case. We do not address in this appeal whether the district court's intervention decisions were proper, and we do not resolve appeals numbers 14-3756 and 14-3881. Nevertheless, we consider the arguments the General Assembly presented in the brief filed in 14-3881 as if it were filed as an amicus curiae brief in this case. Moreover, we also consider the arguments presented by amici curiae United States and Cuyahoga County in briefs filed in this case.

B. Factual Background

Ohio established early in-person voting largely in response to well-documented problems in administering the 2004 general election. As we explained in Obama for America v. Husted, 697 F.3d 423 (6th Cir. 2012), "[d]uring that election, Ohio voters faced long lines and wait-times that, at some polling places, stretched into the early morning of the following day." Id. at 426. In League of Women Voters of Ohio v. Brunner, 548 F.3d 463 (6th Cir.2008), we summarized the problems the League of Women Voters of Ohio reported voters faced as follows:

Id. at 477-78. In sum, many voters in the 2004 general election were effectively disenfranchised and unable to vote.

In 2005, the Ohio General Assembly passed Substitute House Bill 234 to remedy these problems. 2005 Ohio Laws 40 (Sub.H.B.234). HB 234 instituted no-fault early voting, eliminating the requirement that Ohio voters had to provide an excuse for not being able to vote on Election Day in order to vote early. Early voting is done via an "absentee ballot," which may be cast either early in-person ("EIP") at the voter's Board of Elections' ("BOE") designated voting location or by mailing the ballot to the BOE. Ohio Rev.Code § 3509.05(A). Each county has one BOE, which is permitted to operate only one location for EIP voting. Id. § 3501.10(C). Under the 2005 early-voting scheme, the BOEs were required to make absentee ballots available for voters — either for EIP voting or by mail voting — no later than 35 days before the election. Id. § 3509.01(B)(2) (2014) (as amended Feb. 25, 2014). Ohio law requires voters to be registered at least 30 days prior to an election. Ohio Const. § 5.01; Ohio Rev. Code § 3503.01(A). Therefore, Ohio voters could register and vote on the same day for a five-day period that Plaintiffs refer to as "Golden Week."

Until 2012, Ohio law gave each of the BOEs for Ohio's eighty-eight counties the discretion to set their own EIP voting hours. R. 62 (Parties' Statement Undisputed Facts ¶ 6) (Page ID # 3307). Thus, for the 2008 and 2010 elections each BOE set its own EIP voting hours. Id. Several counties, including six counties with the highest African American populations in Ohio, offered early voting during the evenings and on multiple Sundays. Id. ¶ 8 (Page ID #3307); R. 65-3 (2010 Early Voting Days & Times) (Page ID # 4576-84); R. 72 (D. Ct. Op. and Order at 9) (Page ID # 5856).

On August 15, 2012, Secretary Husted issued Directive 2012-35, which established uniform EIP voting hours for all BOEs for the 2012 general election. R. 62 (Parties' Statement Undisputed Facts ¶ 15) (Page ID # 3308). Directive 2012-35 eliminated all weekend EIP voting hours. R. 18-34 (Directive 2012-35) (Page ID # 527-28). It did provide for some evening EIP voting hours on ten weekdays in the last two weeks before Election Day. Id. Directive 2012-35 was challenged in separate litigation as violating the Equal Protection Clause because it allowed only military voters to vote EIP during the last three days before the election. In Obama for America, we upheld the district court's issuance of a preliminary injunction enjoining the enforcement of the Directive regarding the last three days of EIP voting before the election; under the preliminary injunction local BOEs had discretion to set EIP voting hours for those days so long as those hours applied to all voters, not just military voters. 697 F.3d at 437.

In the 2008, 2010, and 2012 elections, many Ohio voters took advantage of early voting. As we noted in Obama for America, in 2008 "approximately 1.7 million Ohioans cast their ballots before election day, amounting to 20.7% of registered voters and 29.7% of the total votes cast.... In 2010, approximately 1 million Ohioans voted early, and 17.8% of them chose to cast their ballots in person." 697 F.3d at 426. In the 2012 election, roughly 32% of Ohioans voted early. R. 18-1 (Smith Rep. at 6) (Page ID # 167). Thousands of voters also registered or updated their registration and voted during Golden Week. R. 62 (Parties' Statement Undisputed Facts ¶¶ 10, 14) (Page ID # 3308).

The General Assembly passed SB 238 on February 19, 2014, and it went into effect on June 1, 2014. SB 238 amended the Ohio Code to make the first permitted day of early voting the day after the close of voter registration. Ohio Rev.Code § 3509.01(B)(2)-(3). Thus, SB 238 reduces the total number of EIP voting days by eliminating Golden Week. SB 238 largely mirrored the recommendations in a report by the Ohio Association of Election Officials ("OAEO"). R. 18-33 (OAEO Rep.) (Page ID # 521-26).

On February 25, 2014, Secretary Husted issued Directive 2014-06, which set EIP voting hours for the 2014 primary and general elections. For the general elections, the Directive did not include EIP voting hours for the following times: (1) the Sunday and Monday immediately before Election Day; (2) Tuesday, September 30th through Monday, October 6th, the last day for voter registration (Golden Week); (3) Saturday, October 11th, Sunday, October 19th, or Sunday, October 26th; and (4) evening EIP voting hours after 5 p.m. on all weekdays or after 4 p.m. on Saturday, October 25th and Saturday, November 1st. R. 18-36 (Directive 2014-06 at 2) (Page ID # 531).

On June 11, 2014, however, the U.S. District Court for the Southern District of Ohio issued a permanent injunction in Obama for America v. Husted, No. 2:12-CV-636, 2014 WL 2611316 (S.D.Ohio June 11, 2014). The district court "require[d] Secretary of State Husted to set uniform and suitable in-person early voting hours for all eligible voters for the three days preceding all future elections." Id. at *5. No party appealed this final judgment.

To comply with the permanent injunction, Secretary Husted issued Directive 2014-17 on June 17, 2014. R. 18-37 (Directive 2014-17) (Page ID # 532-33). Directive 2014-17 sets uniform EIP voting hours for all future elections in three categories: (1) Presidential General Elections; (2) Presidential Primary Elections and Gubernatorial General Elections; and (3) Regular Municipal Elections, Primary Elections, and Special Elections. Id. As required by the permanent injunction, Directive 2014-17 restores EIP voting hours for the Gubernatorial General Elections (the only election relevant to the 2014 general election) on the Sunday and Monday immediately prior to Election Day on November 4, 2014. Id. at 2 (Page ID # 533). In all other respects Directive 2014-17 sets the same EIP voting hours that Directive 2014-06 set for the Gubernatorial General Election. Id.

II. DISCUSSION

A. Standard of Review

"[F]our factors ... must [be] balance[d] when considering a motion for preliminary injunction: (1) whether the movant has a strong likelihood of success on the merits; (2) whether the movant would suffer irreparable injury without the injunction; (3) whether issuance of the injunction would cause substantial harm to others; and (4) whether the public interest would be served by issuance of the injunction." City of Pontiac Retired Employees Ass'n v. Schimmel, 751 F.3d 427, 430 (6th Cir.2014) (internal quotation marks omitted). "We review a district court's grant of a preliminary injunction for an abuse of discretion." Obama for America, 697 F.3d at 428. However, we review de novo the district court's legal conclusions, and we review its factual findings for clear error. Id. Thus, "[t]he district court's determination will be disturbed only if the district court relied upon clearly erroneous findings of fact, improperly applied the governing law, or used an erroneous legal standard." McNeilly v. Land, 684 F.3d 611, 614 (6th Cir.2012) (internal quotation marks omitted). Moreover, "the `determination of whether the movant is likely to succeed on the merits is a question of law and is accordingly reviewed de novo.'" Liberty Coins, LLC v. Goodman, 748 F.3d 682, 689 (6th Cir.2014) (quoting Bays v. City of Fairborn, 668 F.3d 814, 819 (6th Cir.2012)).

B. The District Court's Factual Findings Are Not Clearly Erroneous

The district court analyzed the record evidence and made a number of factual findings in granting Plaintiffs' motion for a preliminary injunction. In particular, the district court's opinion carefully considered the conclusions of Plaintiffs' four expert witnesses (Smith, Roscigno, Burden, and Gronke) and Defendants' three expert witnesses (Trende, McCarty, and Brunell). See R. 72 (D. Ct. Op. and Order at 26-45) (Page ID #5873-92). After assessing each, the district court credited Smith's conclusion that, based on his statistical analysis, African Americans will be disproportionately and negatively affected by the reductions in early voting in SB 238 and Directive 2014-17. Id. at 44-45 (Page ID # 5891-92). The district court also accepted Roscigno's "undisputed" findings that disparities in employment and in residential, transportation, and childcare options between African American and white voters significantly increased the cost of casting a vote for African American voters. Id. at 45 (Page ID # 5892). The district court then relied on Smith's statistical findings and conclusions in its Equal Protection analysis, and it relied on both Smith's and Roscigno's findings in its Voting Rights Act analysis.

Defendants have not challenged the admissibility of any of Plaintiffs' experts' conclusions under Daubert, either at the district court or on appeal. See Daubert v. Merrell Dow Pharm., Inc., 509 U.S. 579, 113 S.Ct. 2786, 125 L.Ed.2d 469 (1993). Indeed, Defendants' brief makes little mention of the district court's factual findings or its decision to credit the conclusions of Plaintiffs' experts. Thus, whether or not the district court properly considered these expert findings is not before us. In any event, while they do not dispute Roscigno's, Burden's, or Gronke's conclusions, to the extent Defendants and the General Assembly believe the district court improperly credited Smith's findings over the conclusions offered by Defendants' experts, they are mistaken.

We review a district court's factual findings for clear error. Williamson v. Recovery Ltd. P'ship, 731 F.3d 608, 627 (6th Cir.2013). "When reviewing for clear error, we cannot substitute our judgment for that of the lower court but rather must uphold the lower court's account of the evidence if it `is plausible in light of the record viewed in its entirety.'" Pledger v. United States, 236 F.3d 315, 320 (6th Cir. 2000) (quoting Anderson v. City of Bessemer City, 470 U.S. 564, 574, 105 S.Ct. 1504, 84 L.Ed.2d 518 (1985)). We will thus reverse the district court's interpretation of the evidence "`only where we are left with a definite and firm conviction that [the district court] committed a clear error of judgment.'" In re Scrap Metal Antitrust Litig., 527 F.3d 517, 528 (6th Cir. 2008) (quoting Conwood Co., L.P. v. U.S. Tobacco Co., 290 F.3d 768, 781 (6th Cir. 2002)). Consequently, "[i]f the district court interprets the evidence in a manner consistent with the record, we are required to uphold its decision even if we would have reached the opposite conclusion." United States v. Darwich, 337 F.3d 645, 663 (6th Cir.2003).

Defendants and the General Assembly are unable to show that the district court clearly erred by crediting Smith's statistical conclusions. First, contrary to the General Assembly's claim, the record does not support a finding that the district court erred because Smith's conclusions are based on faulty data. General Assembly Br. at 48-51. The district court recognized at the outset of its analysis "that some significant limitations exist regarding the available election data," including different election management systems and policies for tabulating absentee votes among the counties, which made statewide comparisons difficult. R. 72 (D. Ct. Op. and Order at 26-27) (Page ID # 5873-74). Indeed, Smith suggested as much in his expert reports. R. 18-1 (Smith Rep. at 12) (Page ID # 173); 53-11 (Smith Rebuttal Rep. at 1-2, 25) (Page ID # 1628-29, 1652).

Recognizing these limitations, Smith utilized several techniques based on entirely different statistical methods and data sources to determine whether the propensity of African Americans to cast EIP ballots in Ohio is greater than whites. See 53-11 (Smith Supp. Rep. at 1-2, 25) (Page ID # 1628-29, 1652). In particular, Smith utilized a "triangulation" method, which relied on data from the U.S. Census Bureau, Ohio, and county Boards of Elections, and included three different "standard ecological inference techniques" to analyze voting trends in the 2010 midterm and 2012 presidential elections. Id. at 1 (Page ID # 1628). Then, in a separate analysis, Smith examined data from the Current Population Voting and Registration Supplement to determine whether African American voters in Ohio were disproportionately more likely to cast EIP ballots in the 2012 and 2008 elections based on this data. Id. at 1-2 (Page ID # 1628-29). As the district court found, by varying degrees, each of these examinations supports Smith's conclusion that African American voters in Ohio utilize EIP voting at higher rates than white voters in recent elections.1

Smith then supplemented his findings by citing additional studies indicating that African American and indigent voters utilized early voting more than white and affluent voters and would be negatively impacted by restrictions on early voting. See, e.g., R. 18-1 (Smith Rep. at 7, 17, 29-30) (Page ID # 168, 178, 190-91); R. 72 (D. Ct. Op. and Order at 45-46) (Page ID # 5892-93). These findings are further supported by expert reports submitted by Gronke and Burden and studies attached to Plaintiffs' briefing to the district court.2 See R. 72 (D. Ct. Op. and Order at 45-46) (Page ID #5892-93); 53-5 (Gronke Rep. at 6-12) (Page ID # 1563-69) (noting research indicating that African Americans disproportionately use early voting in many states and shortening the early-vote period negatively impacted turnout among African Americans); R. 53-4 (Burden Rep. at 3) (Page ID # 1555) (citing research noting that "restrictions on early voting in Florida finds that it deterred participation of black voters"). Although, standing alone, any one analysis may not have proven dispositive, when reading them together and as properly supported by other record evidence, the district court did not clearly err by crediting Smith's analysis despite the possibility of flaws in the data.

Second, for the same reason, limitations in Smith's analysis of the 2010 election do not demonstrate that the district court clearly erred by relying on Smith's findings. General Assembly Br. at 51-52. Defendants cite the 2010 analysis to show that the early-voting days that were eliminated were among the days with the lowest African American voting rates. Appellants Br. at 56. And it is true that Smith's 2010 analysis considered only data from five of eighty-eight Ohio counties. But, here, it is Defendants who attempt to cherry-pick the findings. Again, Smith's analysis of the 2010 mid-term election was one of a number of studies cited by the district court in support of its conclusion that African Americans would be disproportionately impacted by restrictions in EIP voting. Moreover, the five counties analyzed by Smith in his findings based on the 2010 election make up one-third of Ohio's population and nearly seventy-three percent of all African Americans living in Ohio, and the findings overall indicate that African Americans participated in EIP at a higher rate than white voters in these counties. R. 18-1 (Smith Rep. at 10) (Page ID # 171); R. 53-11 (Smith Rebuttal Rep. at 22) (Page ID #1649). Thus, the 2010 analysis is certainly relevant to whether African American voters utilized early voting more than white voters, and the district court properly considered this finding along with the other evidence in the record in reaching its conclusion.

Third, the General Assembly's suggestion that Smith's analyses relating to the 2012 and 2008 elections are not probative here because these were presidential elections and the 2014 election is an off-year election is not well-taken. General Assembly Br. at 52-53. Plaintiffs' complaint does not limit its challenge to the 2014 midterm elections. See R. 1 (Complaint) (Page ID # 1). Indeed, SB 238 is the law in Ohio and will apply to all elections moving forward, and nothing in the record suggests that the restrictions on early voting in Directive 2014-17 will be limited to the 2014 election. In fact, Directive 2014-17 expressly applies to "Presidential General Elections." R. 18-37 (Directive 2014-17 at 1) (Page ID # 532). Thus, any attempt to diminish the probative value of Smith's 2012 and 2008 election analyses for this reason has no merit. Similarly, attempts to disregard voter turnout among African Americans in the 2012 and 2008 elections because African American voters were targeted by an African American presidential candidate are equally meritless. General Assembly Br. at 53. The suggestion is that African American voters in Ohio — a "battleground" state, central to any presidential candidate's chance of winning an election — will not be as heavily targeted in future elections. Id. But this claim is both unsupported by record evidence and, given the continued importance of Ohio in national elections, contrary to common sense. See R. 41-3 (Trende Rep. at 31) (Page ID # 1041) (noting that African American voter turnout has risen in Ohio and the United States as a whole since 2004); see also Florida v. United States, 885 F.Supp.2d 299, 326 (D.D.C. 2012) ("[W]e cannot ignore elections in which minority candidates make breakthroughs in winning elected office on the assumption that future elections will revert to the status quo.").

Fourth, Defendants and the General Assembly suggest that the district court should not have credited Smith's analysis because Defendants' expert, Sean Trende, performed a statistical analysis that produced different results. See Appellants Br. at 56; General Assembly Br. at 52. While we acknowledge that a Daubert issue is not before us, it remains true that district courts play the role of "gatekeeper" and are charged "with evaluating the relevance and reliability of proffered expert testimony with heighted care." Surles ex rel. Johnson v. Greyhound Lines, Inc., 474 F.3d 288, 295 (6th Cir. 2007). For this reason, we generally defer to the district court's decision to credit one expert over another. In re Scrap Metal, 527 F.3d at 528 (recognizing the deference afforded a district court's assessment of expert testimony). Moreover, "[w]here there are two permissible views of the evidence, the factfinder's choice between them cannot be clearly erroneous." Anderson, 470 U.S. at 574, 105 S.Ct. 1504.

Here, Trende analyzed EIP turnout from the 2010 elections and found that, contrary to Smith's conclusion, "it is difficult to conclude that early voting enhances African-American turnout." R. 41-3 (Trende Rep. at 42) (Page ID # 1052). He acknowledged, however, that the "strength of the relationship tested depends on the judgment call that is made about the different variables." Id. Indeed, Trende asserted that much of his analysis — as was undoubtedly the case for Smith's analysis — reflected judgment calls that could "reasonably be argued either way." Id. at 34 (Page ID # 1044). The district court's decision in assessing the evidence to then credit Smith's findings — an academic in the area of electoral processes and election issues, R. 18-1 (Smith Rep. at 2-3) (Page ID # 163-64) — and the judgment calls inherent in the same, over Trende's — an elections analyst for the political website RealClearPolitics, who apparently has not conducted a peer-reviewed analysis similar to the one at issue here, R. 41-3 (Trende Rep. at 3) (Page ID # 1013); R. 53-6 (7/30/14 Trende Dep. at 281) (Page ID # 1576) — is afforded deference. In re Scrap Metal, 527 F.3d at 528. Given this, along with the multiple methods and data sources used by Smith and other record evidence corroborating his findings, we conclude that the district court did not clearly err by crediting Smith's findings over Trende's.

Finally, the General Assembly's claim that the district court erred because Smith's methodology is flawed also fails. The General Assembly asserts that Smith's findings are unreliable because factors other than race could explain the results of Smith's census block analysis of the 2012 election, General Assembly Br. at 54-55; however, no evidence is offered supporting this. Moreover, the other record evidence suggesting that African American voters utilize EIP voting at higher rates than white voters indicates that race, rather than some other variable, helps explain Smith's findings in his 2012 census block analysis. At the least, the district court did not clearly err in so finding. See Surles, 474 F.3d at 295 (noting the "broad discretion" district courts possess to assess the reliability of expert findings); see also McLean v. 988011 Ontario, Ltd., 224 F.3d 797, 801 (6th Cir.2000) ("[M]ere `weaknesses in the factual basis of an expert witness' opinion ... bear on the weight of the evidence rather than on its admissibility.'") (quoting United States v. L.E. Cooke Co., 991 F.2d 336, 342 (6th Cir.1993)).

Similarly, Defendants' expert, Dr. Nolan McCarty, and Smith quibble over whether Smith should have conducted his census block analysis at the county level rather than the precinct level — McCarty claims the results are more accurate at the county level while Smith asserts that "aggregat[ing] up ... dramatically worsens the problem of aggregation bias." Compare R. 53-11 (Smith Rebuttal Rep. at 5-6) (Page ID # 1632-33), with General Assembly Br. at 55-56. And the General Assembly claims that the district court gave too much weight to the "direction of the relationship" between African American voters and EIP voting in the 2012 and 2010 analyses and ignored the "degree of the relationship," which it claims is small. General Assembly Br. at 56-57. But neither argument supports reversal — again, given the other record evidence supporting Smith's conclusion and the deference afforded the district court, the district court's position is plausible based on the record as a whole, and so there is no clear error. King v. Zamiara, 680 F.3d 686, 694 (6th Cir.2012), cert. denied, ___ U.S. ___, 133 S.Ct. 985, 184 L.Ed.2d 773 (2013) ("If the district court's account is `plausible in light of the record viewed in its entirety, the court of appeals may not reverse.'") (quoting Anderson, 470 U.S. at 574, 105 S.Ct. 1504).

Consequently, Defendants and the General Assembly have failed to show that the district court clearly erred in crediting Smith's statistical conclusions.

C. Equal Protection Clause Claim

The right to vote is a "fundamental" right. Harper v. Virginia State Bd. of Elections, 383 U.S. 663, 670, 86 S.Ct. 1079, 16 L.Ed.2d 169 (1966); Burdick v. Takushi, 504 U.S. 428, 433, 112 S.Ct. 2059, 119 L.Ed.2d 245 (1992) ("It is beyond cavil that `voting is of the most fundamental significance under our constitutional structure.'") (quoting Illinois Bd. of Elections v. Socialist Workers Party, 440 U.S. 173, 184, 99 S.Ct. 983, 59 L.Ed.2d 230 (1979)). "Other rights, even the most basic, are illusory if the right to vote is undermined." Wesberry v. Sanders, 376 U.S. 1, 17, 84 S.Ct. 526, 11 L.Ed.2d 481 (1964). Moreover, "[t]he right to vote is protected in more than the initial allocation of the franchise. Equal protection applies as well to the manner of its exercise." Bush v. Gore, 531 U.S. 98, 104, 121 S.Ct. 525, 148 L.Ed.2d 388 (2000); League of Women Voters of Ohio, 548 F.3d at 476 (quoting the same). Two aspects of "the manner of its exercise" warrant special attention: "[t]he Equal Protection Clause applies when a state either classifies voters in disparate ways or places restrictions on the right to vote." Obama for America, 697 F.3d at 428 (emphasis added) (internal citations omitted).

Of course, "the Constitution provides that States may prescribe" `[t]he Times, Places and Manner of holding Elections for Senators and Representatives,' Art. I § 4, cl. 1, and the Court therefore has recognized that States retain the power to regulate their own elections." Burdick, 504 U.S. at 433, 112 S.Ct. 2059; Ne. Ohio Coal. for the Homeless v. Husted, 696 F.3d 580, 592 (6th Cir.2012). And practically, "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).

"When equal protection challenges ask us to resolve these competing interests, we calibrate the equal protection standard to `[t]he precise character of the state's action and the nature of the burden on voters.'" Ne. Ohio Coal. for the Homeless, 696 F.3d at 592 (quoting Obama for America, 697 F.3d at 428). State regulations that do not treat similarly situated voters differently and do not burden the fundamental right to vote are assessed through rational basis review. Obama for America, 697 F.3d at 429; Ne. Ohio Coal. for the Homeless, 696 F.3d at 592. On the other end of the spectrum, strict scrutiny applies to state regulations that impose "severe" burdens on the fundamental right to vote. Obama for America, 697 F.3d at 429 (citing Harper, 383 U.S. at 670, 86 S.Ct. 1079, and Burdick, 504 U.S. at 434, 112 S.Ct. 2059).

"For the majority of cases falling between these extremes, we apply the `flexible' Anderson-Burdick balancing test." Ne. Ohio Coal. for the Homeless, 696 F.3d at 592 (quoting Obama for America, 697 F.3d at 429). The Anderson-Burdick test provides as follows:

Burdick, 504 U.S. at 434, 112 S.Ct. 2059 (quoting Anderson v. Celebrezze, 460 U.S. 780, 789, 103 S.Ct. 1564, 75 L.Ed.2d 547 (1983)). "There is no `litmus test' to separate valid from invalid voting regulations; courts must weigh the burden on voters against the state's asserted justifications and `make the "hard judgment" that our adversary system demands.'" Obama for America, 697 F.3d at 429 (quoting Crawford v. Marion Cnty. Election Bd., 553 U.S. 181, 190, 128 S.Ct. 1610, 170 L.Ed.2d 574 (2008) (Stevens, J., announcing the judgment of the Court)). Even a minimal burden "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)).

The district court "characterize[d] the overall degree of burden on voting imposed by SB 238 and Directive 2014-17 as significant although not severe." R. 72 (D. Ct. Op. and Order at 53) (Page ID # 5900). Focusing on SB 238, the district court found that its elimination of "Golden Week" burdened African American and low-income voters in two ways. Id. at 50 (Page ID # 5897). First, SB 238 in conjunction with Directive 2014-17 reduced the overall number of EIP voting days from 35 to 28 days. The district court noted evidence in the record that 67,408 Ohioans voted in 2008 during Golden Week; 26,230 did so in 2010; and 89,224 voters did so in 2012. Id. The district court also credited statistical and survey analysis by Plaintiffs' expert Smith that African American voters in Ohio have higher EIP voting rates than white voters, and that African American voters in the 2008, 2010, and 2012 elections "disproportionately cast EIP absentee ballots on days that would have been eliminated by SB 238 and Directive 2014-06." R. 72 (D. Ct. Op. and Order at 32, 50) (Page ID # 5879, 5897); R. 18-1 (Smith Rep. at 31) (Page ID # 192). The district court cited four other statistical studies in the record on racial early-voting patterns in Ohio that it found supported Smith's conclusions. Id. at 45-46 (Page ID # 5892-93).

Second, the district court concluded that the elimination of Golden Week "burdens the voting rights of lower income and homeless individuals" because the record reflected that such individuals "move frequently" as well as "lack access to transportation," which combine to make it harder for such individuals to maintain accurate registration. R. 72 (D. Ct. Op. and Order at 51) (Page ID # 5898). Thus, the ability to register and vote on the same day "can make the difference between being able to exercise the fundamental right to vote and not being able to do so." Id. The court pointed to evidence in the record that 12,842 voters used Golden Week to register or update their registration and vote in 2008; 1,651 voters did so in 2010; and 5,844 voters did so in 2012. Id.

Turning to Directive 2014-17, the district court found that it burdened African American and lower-income voters by eliminating all evening voting hours for non-presidential elections and by providing only one Sunday of EIP voting, the Sunday before Election Day. Id. at 51-53 (Page ID #5898-5900). The district court noted that the record reflected that lower-income voters are "more likely to rely on public transportation and work wage-based jobs wherein they are less likely" to be able to vote between 8 a.m. and 5 p.m. at the one early-voting location permitted in each county, which might be a great distance away. Id. at 53 (Page ID # 5900). Regarding the elimination of all but one Sunday of EIP voting, the court pointed to evidence in the record that since the instatement of EIP voting, African Americans have come to rely on Sunday voting through "Souls to the Polls initiatives," in which churches have leveraged the transportation they already provide to and from church to bring voters to EIP voting locations. Id. at 52 (Page ID # 5899). Souls to the Polls organizers reported that, during the one permitted day of Sunday voting during the 2012 general election, there were long lines of mainly African American voters. Id. While the district court acknowledged that Souls to the Polls organizers could switch to the two Saturdays that are still designated EIP voting days under Directive 2014-17, the court concluded that this would still impose "some burden" because churches are already organized to provide transportation on Sundays. Id. at 53 (Page ID # 5900).

Therefore, because the district court found that the burden imposed on Plaintiffs was "significant," it proceeded to apply the Anderson-Burdick test to SB 238 and Directive 2014-17. R. 72 (D. Ct. Op. and Order at 55) (Page ID # 5902).

Defendants argue that rational basis review, rather than the Anderson-Burdick test, is the proper standard of review for two reasons. First, they argue that "[t]he `right to vote' has never included the `right to receive absentee ballots.'" Appellants Br. at 18 (quoting McDonald v. Bd. of Election Com'rs of Chicago, 394 U.S. 802, 807, 89 S.Ct. 1404, 22 L.Ed.2d 739 (1969)). Second, Defendants argue that when a facially neutral voting law is at issue, as SB 238 and Directive 2014-17 are, the Supreme Court in Crawford held that Anderson-Burdick applies only if the law "severely burdens the right to vote of the general class of state voters." Id. at 19. Otherwise, Defendants assert that traditional Equal Protection Clause principles govern — which require proof of discriminatory intent — and Plaintiffs have not established that either SB 238 or Directive 2014-17 was adopted with discriminatory intent. Id. at 19, 27. Finally, if Anderson-Burdick review does apply, Defendants argue that the district court improperly determined that the burden imposed on voters represented by Plaintiffs is "significant." Id. at 31.

1. The District Court Properly Applied Anderson-Burdick Review

We addressed Defendants' first argument regarding McDonald in Obama for America, 697 F.3d 423. In McDonald, the Supreme Court did not apply rational basis review to the challenged Illinois statute allowing only certain categories of voters to receive absentee ballots solely because absentee ballots were at issue. Rather,

Obama for America, 697 F.3d at 431 (emphasis added).

Thus, in Obama for America, we held that the district court properly applied the Anderson-Burdick balancing test, rather than rational basis review, to evaluate whether the challenged Directive's elimination of early in-person voting for the three days immediately preceding Election Day violated the Equal Protection Clause. Id. Unlike the plaintiffs in McDonald, we noted that "Plaintiffs introduced extensive evidence that a significant number of Ohio voters will in fact be precluded from voting without the additional three days of in-person early voting." Id. This evidence included "statistical studies that estimated approximately 100,000 Ohio voters would choose to vote during the three-day period before Election Day, and that these voters are disproportionately `women, older, and of lower income and education attainment,'" groups which the plaintiffs represented. Id. (internal citation omitted). The defendants in that case also argued that the plaintiffs would not actually be precluded from voting as required by McDonald because they had "ample" other means of voting, including by mail, voting EIP at other times, or on Election Day. Id. However, we held not clearly erroneous the district court's conclusion that early voters would not be able to exercise their right to vote in person because the challenged Directive also eliminated all evening and weekend hours of EIP voting, times during which early voters would likely have voted in the past because they tend to have lower incomes and less education than election day voters. Id.

We did not read McDonald to require proof that there was no possibility that the plaintiffs would find a way to adjust and vote through the remaining options. We acknowledged that the challenged law "does not absolutely prohibit early voters from voting," but focused on the evidence in the record that the plaintiffs' "ability to cast a ballot is impeded by Ohio's statutory scheme." Id. at 433. To the extent that McDonald spoke in terms of "precluding" an individual from voting, which might imply the necessity of such proof, we note that McDonald was decided before the development of the Anderson-Burdick test. Thus, the McDonald Court applied a two-tier test for evaluating restrictions on the right to vote, rational basis review for no burdens and strict scrutiny for "severe" burdens, a threshold that more clearly invites consideration of "preclusion." However, as noted above, that two-tier test has evolved into the Anderson-Burdick framework, under which burdens that fall between those two extremes can still be found to violate the Equal Protection Clause. In more recent cases, the Supreme Court has not required absolute certainty in predicting how many voters would be prevented from voting by laws that impose burdens on the right to vote. See, e.g., Crawford, 553 U.S. at 221, 128 S.Ct. 1610 (Souter, J., dissenting) (stating that "Petitioners, to be sure, failed to nail down precisely how great the cohort of discouraged and totally deterred voters will be, but empirical precision beyond the foregoing numbers has never been demanded for raising a voting-rights claim." and citing Washington State Grange v. Washington State Republican Party, 552 U.S. 442, 461-62, 128 S.Ct. 1184, 170 L.Ed.2d 151 (2008) (Roberts, C. J., concurring) ("Nothing in my analysis requires the parties to produce studies regarding voter perceptions on this score"); Dunn v. Blumstein, 405 U.S. 330, 335 n. 5, 92 S.Ct. 995, 31 L.Ed.2d 274 (1972) ("[I]t would be difficult to determine precisely how many would-be voters throughout the country cannot vote because of durational residence requirements."); and Bullock v. Carter, 405 U.S. 134, 144, 92 S.Ct. 849, 31 L.Ed.2d 92 (1972) (taking account of "the obvious likelihood" that candidate filing fees would "fall more heavily on the less affluent segment of the community, whose favorites may be unable to pay the large costs")). Thus, in this case the district court properly held that whether voters might adjust to vote during a different time in EIP voting such that overall turnout might not be affected "is not determinative of the Equal Protection analysis." R. 72 (D. Ct. Op. and Order at 50) (Page ID # 5897).3

Like the plaintiffs in Obama for America, Plaintiffs in this case presented ample evidence that African American, lower-in-come, and homeless voters disproportionately have used in past elections the EIP voting times that Directive 2014-17 and SB 238 eliminated, and that the number of individuals who had previously voted during these periods was not insignificant. For example, the number of voters in Golden Week alone ranged from 26,230 in 2010 to 89,224 in 2012. R. 72 (D. Ct. Op. and Order at 50) (Page ID # 5897). And the regulations at issue in the case reduce the overall time for EIP voting more than the three days that had been eliminated in Obama for America. Moreover, unlike the plaintiffs in Obama for America, Plaintiffs also presented evidence to show that voting by mail is not actually a viable "alternative means of access to the ballot" for the groups they represent. Cf. Obama for America, 697 F.3d at 440 (White, J., concurring) (noting that the study in the record "did not consider the extent to which these voters would or could avail themselves of other voting options, either by mail ballot or in-person absentee ballot at other times, or in-person voting on election day.") (emphasis added). The district court noted that "the record is undisputed that African Americans, lower-income individuals, and the homeless are distrustful of the mail and/or voting by mail." R. 72 (D. Ct. Op. and Order at 54) (Page ID #5901). Additionally, the district court considered the fact that "[t]he associated costs and more complex mechanics of voting by mail, coupled with other information in the record concerning the enumerated groups including homelessness, lower educational attainment, more limited financial resources, reliance on public transportation, and transience" to bolster its conclusion that "voting by mail may not be a suitable alternative for many voters." Id. The record also reflected that lower-income voters, because of their reliance on public transportation and higher likelihood of working in wage-based jobs, would face substantial difficulties in voting between 8 a.m. and 5 p.m. Id. at 53 (Page ID # 5900). Under Obama for America, then, the district court properly concluded that Plaintiffs had presented sufficient evidence that the groups they represent are in fact significantly burdened by Directive 2014-17 and SB 238 such that McDonald's rational basis standard does not apply.4

2. Crawford Does Not Foreclose Applying Anderson-Burdick Review

In Northeast Ohio Coalition for the Homeless, we squarely addressed the applicability of Anderson-Burdick to facially neutral restrictions on voting. 696 F.3d 580. The State defendant in that case argued that the challenged practice — "Ohio's automatic disqualification rule for wrong-precinct ballots" — "treats all voters equally and therefore does not involve any classification that could violate the equal protection standard." Id. at 592 (internal quotation marks omitted). In responding to this argument, we explained that

Id. Because the plaintiffs in that case had "`demonstrated that their right to vote is... burdened by Ohio's law that rejects wrong-precinct ballots regardless of poll-worker error," we held that "[t]he Anderson-Burdick standard ... applies." Id. (quoting Obama for America, 697 F.3d at 430); see also Obama for America, 697 F.3d at 428-29 (stating that the "[t]he Equal Protection Clause applies when a state either classifies voters in disparate ways or places restrictions on the right to vote") (emphasis added) (internal citations omitted). However, as the plaintiffs in Northeast Ohio Coalition for the Homeless asserted that the law at issue created a burden on provisional voters generally, rather than on a subclass of provisional voters, we did not address Defendants' more specific argument here that Anderson-Burdick requires a showing of a burden on voters generally.

Contrary to Defendants' assertion, a majority of the Court in Crawford did not expressly hold that a challenger must demonstrate that a voting restriction burdens voters generally in order to trigger scrutiny under Anderson-Burdick. The opinion authored by Justice Stevens announcing the judgment of the Court, which gained only two other votes, did not explicitly reject the petitioners' argument that the middle level of scrutiny under the Anderson-Burdick balancing test could be triggered by evidence of burdens on a subgroup of voters, instead of all voters; rather, the Court held that "on the basis of the evidence in the record it is 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 is fully justified." Crawford, 553 U.S. at 200, 128 S.Ct. 1610; see also id. at 202, 128 S.Ct. 1610 ("In sum, on the basis of the record that has been made in this litigation, we cannot conclude that the statute imposes `excessively burdensome requirements' on any class of voters.") (emphasis added) (citation omitted). Thus, Justice Stevens weighed the evidence of minimal burdens of the law on voters generally, the vast majority of whom had IDs, and found that burden justified by the state's interests. Id. at 202, 128 S.Ct. 1610. In contrast, Justice Scalia's concurrence, which two other Justices joined, expressly would have required the plaintiffs to demonstrate that voters generally were burdened for the Anderson-Burdick balancing test to apply. Id. at 205-06, 128 S.Ct. 1610 (Scalia, J., concurring). Justice Scalia cited decisions of the Court outside of the elections context holding that a generally applicable law does not violate the Equal Protection Clause when it merely disproportionately burdens a subgroup of people absent evidence of discriminatory intent. Id. at 207-08, 128 S.Ct. 1610. Justice Souter's dissent, joined by Justice Ginsburg, assumed that Anderson-Burdick balancing could be triggered by burdens on subgroups, and disagreed with Justice Stevens that the petitioners had not presented sufficient evidence of a more than minimal burden on the subgroup of voters they represented. Id. at 237, 128 S.Ct. 1610 (Souter, J., dissenting).

Thus, a majority of the justices in Crawford either did not expressly reject or in fact endorsed the idea that a burden on only a subgroup of voters could trigger balancing review under Anderson-Burdick. Alternatively, as the narrowest basis of the judgment of the Court, Justice Stevens's opinion may be viewed as "the holding of the Court" given the "fragmented" Crawford opinions. See, e.g., Marks v. United States, 430 U.S. 188, 193, 97 S.Ct. 990, 51 L.Ed.2d 260 (1977) ("When a fragmented Court decides a case and no single rationale explaining the result enjoys the assent of five Justices, the holding of the Court may be viewed as that position taken by those Members who concurred in the judgments on the narrowest grounds.") (internal quotation marks omitted). Both Justice Stevens and Justice Scalia held that the Indiana law was constitutional, but Justice Scalia reached further and held that challengers to a voting restriction must show that it burdens voters generally. Thus, Justice Stevens's opinion is narrower. See Frank v. Walker, ___ F.Supp.3d ___, ___, No. 11-CV-01128, 2014 WL 1775432, at *4 (E.D.Wis. Apr. 29, 2014) (holding that Justice Stevens's opinion in Crawford is the narrower ground for the opinion under Marks). Finally, it is worth noting that in Anderson, the Supreme Court in fact assessed the burden imposed by the challenged law by looking to its impact on a subgroup of voters. Anderson, 460 U.S. at 792, 103 S.Ct. 1564 (holding that "[i]t is clear, then, that the March filing deadline places a particular burden on an identifiable segment of Ohio's independent-minded voters," specifically Anderson's supporters, but not assessing whether the deadline burdens all voters).

Therefore, that Plaintiffs presented evidence only of SB 238 and Directive 2014-17's burdens on African American, lower-income, and homeless voters does not automatically mean that only rational basis review or standard Equal Protection Clause analysis applies. Crawford merely stands for the proposition that Plaintiffs must present more evidence than the petitioners did in that case to show that the subgroups of voters they represent are more than minimally burdened.

As discussed previously, Plaintiffs presented extensive statistical, survey, and anecdotal evidence that SB 238 and Directive 2014-17 will disproportionately burden the ability of African American, lower-income, and homeless individuals to vote. The petitioners in Crawford had not presented any evidence in the record that even estimated the number of individuals who lacked identification cards. Crawford, 553 U.S. at 200, 128 S.Ct. 1610. Nor did the affidavits or depositions in the record of lower-income individuals or elderly voters in Crawford substantiate that they in fact faced difficulties in obtaining identification cards. Id. at 201, 128 S.Ct. 1610. In contrast, Plaintiffs here presented statistical and survey evidence that indicated that thousands of individuals whom they represent had voted in past elections during the times that have been eliminated by SB 238 and Directive 2014-17, as well as numerous depositions, affidavits, and expert testimony documenting that the groups Plaintiffs represent have relied on the eliminated EIP voting times and would face difficulties in voting without them.5

In sum, we hold that the district court's characterization of the overall burden imposed by SB 238 and Directive 2014-17 as significant, but not severe, was not clearly erroneous given the extensive evidence in the record of the burdens African American, lower-income, and homeless voters will face in voting, absent the times eliminated by SB 238 and Directive 2014-17.6 It therefore properly applied the Anderson-Burdick balancing test. We next turn to the district court's evaluation of Defendants' asserted justifications for SB 238 and Directive 2014-17.

3. The State's Justifications Do Not Outweigh the Significant Burden on Voters

Once a court has determined that a law burdens voters, under Anderson-Burdick those burdens must be weighed 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." Anderson, 460 U.S. at 789, 103 S.Ct. 1564 (emphasis added). Put differently, the state must articulate specific, rather than abstract state interests, and explain why the particular restriction imposed is actually necessary, meaning it actually addresses, the interest put forth. See Obama for America, 697 F.3d at 433-34 (assessing under Anderson-Burdick whether the state had presented actual evidence to support the justifications it provided for the challenged law).

Before even articulating these interests, Defendants appear to argue that we held in Obama for America that a law such as SB 238 or Directive 2014-17 would automatically survive this scrutiny of state interests. Appellants Br. at 21 (quoting our statement in Obama for America, 697 F.3d at 433-34, that "If the State had enacted a generally applicable, nondiscriminatory voting regulation that limited in-person early voting for all Ohio voters, its `important regulatory interests' would likely be sufficient to justify the restriction.'"). That statement, of course, was not central to our holding in that case and does not control the present case. Moreover, the quoted language in that sentence is in fact from Burdick, 504 U.S. at 434, 112 S.Ct. 2059 (stating that "when [a regulation] imposes only `reasonable, nondiscriminatory restrictions' upon [voting rights], the State's important regulatory interests are generally sufficient to justify the restrictions") (quoting Anderson, 460 U.S. at 788, 103 S.Ct. 1564). The key in that statement is the word generally; the Burdick Court was merely making clear that not all restrictions on voting will be struck down simply because they impose any kind of burden, as states do have the power to regulate elections generally. Burdick itself involved a nondiscriminatory restriction on write-in voting, and the Court still probed the state's asserted justifications for the restriction in the manner required by Anderson. Burdick, 504 U.S. at 434, 112 S.Ct. 2059. Indeed, in a more recent case, the Supreme Court has tied this statement's applicability to situations in which the burden imposed is modest. Washington State Grange v. Washington State Republican Party, 552 U.S. 442, 452, 128 S.Ct. 1184, 170 L.Ed.2d 151 (2008) ("If a statute imposes only modest burdens, however, then `the State's important regulatory interests are generally sufficient to justify reasonable, nondiscriminatory restrictions' on election procedures.") (emphasis added) (quoting Anderson, 460 U.S. at 788, 103 S.Ct. 1564).

We also note that how Ohio's early-voting system compares to that of other states is not relevant under the Anderson-Burdick balancing test. The test directs courts to weigh the burdens imposed on voters in a particular state against the justifications that that state has proffered for the challenged law or practice that imposes those burdens. Early voting does not necessarily play the same role in all jurisdictions in ensuring that certain groups of voters are actually able to vote. Thus, the same law may impose a significant burden in one state and only a minimal burden in another. Similarly, a particular state may have stronger justifications for a law that burdens voters than other states with the same law.

Thus, we will examine in turn each of Defendants' asserted justifications-preventing voter fraud and containing costs for SB 238, and uniformity for Directive 2014-17-under the Anderson-Burdick balancing framework.

i. Fraud

Regarding SB 238, Defendants argue that it is necessary as a measure to reduce fraud arising from same-day registration and voting during Golden Week. Appellants Br. at 26; General Assembly Br. at 38. Defendants point to the testimony by the OAEO Director that the "`registration deadline' exists so officials `can confirm that a voter is who they say they are before they cast a ballot.'" Id. at 26 (citing R. 54-4 (Keeran Decl. at 7) (Page ID # 1851)). Defendants assert that "[w]hen the deadline is later than the start of voting, votes might be counted even though cast `by people who fraudulently registered during this period, because the election officials could not confirm their registration status before Election Day.'" Id. The Ohio General Assembly points to declarations of individual county election officials that "historically voter fraud was most likely to occur during Golden Week," R. 68-2 (Ward Decl. ¶ 4) (Page ID # 5123), or that it is difficult to verify an individual's residence when someone registers and votes on the same day, R. 68-3 (Cuckler Decl. ¶ 9) (Page ID # 5511), or that some individuals were able to cast absentee ballots in one county and then register and cast an EIP Ballot during Golden Week in another county, id. ¶ 8 (Page ID # 5510-11). They argue that the Crawford Court considered similar evidence regarding fraud to hold that the state's interest in that case was sufficiently important to outweigh burdens on voters. Finally, the General Assembly argues that the district court "trivialized or ignored this unrefuted evidence" and improperly substituted its own judgment about how fraud from absentee voting could be "best" handled. General Assembly Br. at 41.

Weighing the state's asserted interest in preventing voter fraud against the significant burden the elimination of Golden Week places on Plaintiffs, we conclude that Defendants have not met their burden to establish that their interests outweigh these burdens. To be sure, "[t]here is no question about the legitimacy or importance of the State's interest in counting only the votes of eligible voters.... While the most effective method of preventing election fraud may well be debatable, the propriety of doing so is perfectly clear." Crawford, 553 U.S. at 196, 128 S.Ct. 1610. This does not mean, however, that the State can, by merely asserting an interest in preventing voter fraud, establish that that interest outweighs a significant burden on voters. Defendants did not provide more than a handful of actual examples of voter fraud, and their general testimony regarding the difficulties of verifying voter registration before counting ballots did not clearly pertain to problems with Golden Week specifically. The district court properly identified that the specific concern Defendants expressed regarding voter fraud — that the vote of an EIP voter would be counted before his or her registration could be verified — was not logically linked to concerns with voting and registering on the same day, but rather "has more to do with the registration process and verification of absentee ballots" generally. R. 72 (D. Ct. Op. and Order at 56) (Page ID #5903). The court explained that, since Ohio law requires that officials segregate absentee ballots and not count them until registration is verified, see, e.g., R. 53-10 (Directive 2012-36) (Page ID # 1625-26); R. 58-16 (Clyde Dec. ¶ 16) (Page ID # 2169), there is no reason to think that the registration of voters who registered and voted on the same day during Golden Week would be any harder to verify than an individual who registered on the last permissible day and then voted the next day, or for that matter than someone who voted very close to the election. R. 72 (D. Ct. Op. and Order at 56) (Page ID #5903). Defendants did not explain why it is harder to segregate and count later the absentee ballots of individuals who vote and register on the same day as opposed to segregating absentee ballots that are returned a different way, particularly given that officials would have at least 30 days to verify the registration of those who register and vote during Golden Week. Thus, the district court properly concluded that Defendants did not meet their burden of explaining why eliminating Golden Week serves to prevent a "precise" problem of voter fraud in a way that is "necessary" to burden the voters Plaintiffs' represent, as opposed to a measure that might more directly target the asserted problem without burdening voters. Anderson, 460 U.S. at 789, 103 S.Ct. 1564.7

Moreover, the General Assembly's argument that Crawford suggests it sufficiently met its burden to demonstrate that its interest in fraud prevention outweighs the significant burden on Plaintiffs is misplaced. The Crawford Court did not hold that scattered historical examples of voter fraud necessarily establish a sufficient state interest to overcome any burden imposed on voters; given the simply minimal burden that the petitioners had shown on voters in that case, the Court essentially only engaged in a rational basis review of the state's asserted interest in preventing voter fraud, not the more piercing scrutiny that a greater burden would require under Anderson-Burdick. Crawford, 553 U.S. at 194-96, 202, 128 S.Ct. 1610. Here, in contrast, the district court concluded that the burden on Plaintiffs was significant. Thus, its more searching review of Defendants' asserted justification in preventing voter fraud was warranted.

ii. Cost

Defendants next argue that SB 238 and Directive 2014-17 are necessary as cost containment measures. They argue that increasing EIP voting would "increase the costs and administrative burdens" on Election Boards, such as requiring additional staff. Appellants Br. at 25. Defendants also assert that they determined the days and times to eliminate strategically so as "to be more efficient with our tax payer dollars" because "[a] relatively smaller proportion of voters... voted during the times Ohio eliminated, compared to the times it kept." Id. at 25. The Ohio General Assembly adds that the district court improperly dismissed the specific cost estimates of retaining EIP voting provided by several local election officials simply because they "lack[ed] a frame of reference" and ignored less easily quantifiable burdens of Election Boards "having to divert manpower" away from other tasks to administering EIP voting. General Assembly Br. at 45. They argue that they did not need to show that they could not handle the costs of the old system, as the district court suggested. Id.

We also conclude that Defendants' asserted interest in reducing costs does not adequately justify the burdens SB 238 and Directive 2014-17 place on voters. In Obama for America, we held that the State's asserted interest in reducing costs and administrative burdens did not justify the burdens on voters because there was "no evidence that local boards of elections have struggled to cope with early voting in the past, no evidence that they may struggle to do so during the November 2012 election," and because at least one local board, Cuyahoga County, said it had budgeted for EIP voting. 697 F.3d at 433-34. Our focus on whether local boards would "struggle" to handle costs in Obama for America makes clear that it is not enough merely to assert that a restriction on voting saves costs. Arguably some cost-saving rationale could be identified in most voting restrictions. Rather, where more than minimal burdens on voters are established, the State must demonstrate that such costs would actually be burdensome.

The district court thus properly concluded that Defendants had not demonstrated that they would "struggle" with the costs of maintaining EIP voting. While Defendants presented specific cost estimates from a handful of election boards for reinstating Golden Week, the district court properly noted that those figures "lack[ed] a frame of reference" in that Defendants did not indicate whether or how those costs would be burdensome overall. R. 72 (D. Ct. Op. and Order at 57) (Page ID # 5904). Cuyahoga County again filed an amicus brief saying it had already budgeted money for Golden Week and the additional weekend voting days for the 2014 election. R. 28 (Am. Cur. Br. Cuyahoga Cnty. at 8-9) (Page ID #600-01). Nor did Defendants present specific estimates of the costs of maintaining the eliminated days of weekend voting. The district court also pointed to evidence in the record that election boards are required by law to be open during Golden Week, and only five of the eighty-eight counties statewide ran EIP voting at sites other than their regular offices during the 2008 and 2010 elections. R. 72 (D. Ct. Op. and Order at 59) (Page ID # 5906). Thus, the court reasonably concluded that most election boards would not have to bear substantial extra costs associated with maintaining an offsite location for that period. Id. Moreover, like the State in Obama for America, Defendants did not present evidence that the old EIP voting schedule, which included Golden Week and weekend voting, "created undue or burdensome costs." Id. at 58 (Page ID # 5905).

iii. Uniformity

Finally, regarding Directive 2014-17, Defendants justify it as necessary to promote uniformity. Appellants Br. at 24. Defendants argue that "uniformity makes it easier for the State to educate voters about election days and hours." Id. It also ensures fairness by making sure that all voters can vote during the same times across counties. Id. And Defendants argue it reduces litigation risks by ensuring equal treatment of voters. Id.

Again, we conclude this asserted interest fails to outweigh the burdens on Plaintiffs. Defendants present only an abstract interest in uniformity that is not tied to the necessity of SB 238 and Directive 2014-17 specifically. Uniformity can be important to make voter education easier, but Defendants do not explain why a uniform EIP voting schedule could not also include Golden Week and the other eliminated EIP voting times. As the district court explained, "uniformity, standing alone," is not an interest important enough to significantly burden Plaintiffs' ability to vote. R. 72 (D. Ct. Op. and Order at 60) (Page ID # 5907); see also Obama for America, 697 F.3d at 442 (White, J., concurring) ("The desire for uniformity has little to do with the elimination of all weekend and after-hours in-person voting. Defendants offer no explanation for curtailing hours other than on the final weekend, and uniformity without some underlying reason for the chosen rule is not a justification in and of itself. Nor is there a showing that eliminating all weekend and after-hours voting will in fact produce uniform access, as opposed to uniform hours.").

In sum, because we have concluded that none of the interests put forth by Defendants sufficiently justify the significant burden that the district court found SB 238 and Directive 2014-17 place on the voters whom Plaintiffs represent, we find that Plaintiffs are likely to succeed on their Equal Protection Clause claim.

D. Voting Rights Act Section 2 Claim

Section 2 of the Voting Rights Act of 1965, 42 U.S.C. § 1973,8 provides that "[n]o 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...." 42 U.S.C. § 1973(a). In 1982, Congress amended the Voting Rights Act to make clear that "Section 2, unlike other federal legislation that prohibits racial discrimination, does not require proof of discriminatory intent. Instead, a plaintiff need show only that the challenged action or requirement has a discriminatory effect on members of a protected group." Moore v. Detroit Sch. Reform Bd., 293 F.3d 352, 363 (6th Cir.2002); see also Mixon v. Ohio, 193 F.3d 389, 407 (6th Cir.1999) ("Section 2 of the Voting Rights Act requires only a showing of discriminatory effect."). Section 2(b) sets out the test for determining whether a challenged practice violates Section 2(a):

42 U.S.C. § 1973(b).

In Thornburg v. Gingles, the Supreme Court endorsed nine factors, first listed in the Senate Judiciary Committee report for the 1982 amendments to Section 2 (referred to as the "Senate factors"), as relevant to assessing "the totality of the circumstances" in Section 2(b):

478 U.S. 30, 36-37, 106 S.Ct. 2752, 92 L.Ed.2d 25 (1986) (quoting S.REP. NO. 97-417 at 28-29). The Court added, however, that the Senate Report makes clear that "this list of typical factors is neither comprehensive nor exclusive" and that "`there is no requirement that any particular number of factors be proved, or that a majority of them point one way or the other.'" Id. at 45, 106 S.Ct. 2752 (quoting S. REP. at 29).

In finding that Plaintiffs are likely to succeed on their Section 2 claim, the district court first pointed to evidence in the record of a discriminatory effect on African American voters — that the EIP voting hours and days eliminated by SB 238 and Directive 2014-17 would "disproportionately impact African American voters resulting in less opportunity to participate in the political process than other voters." R. 72 (D. Ct. Op. and Order at 61) (Page ID # 5908). The court pointed to the statistical evidence that African Americans use EIP voting at higher rates than others; to the expert testimony of Professor Roscigno and other evidence in the record that African Americans "tend to disproportionately make up the groups that benefit the most from same-day registration: the poor and the homeless"; that the provision of only one Sunday of EIP voting "burdens the voting rights of African Americans by arbitrarily limiting Souls to the Polls voting initiatives"; and that, because African Americans are more likely to be of lower-socioeconomic status, they "tend to work hourly jobs and can find it difficult to find time to vote during normal business hours." Id. at 65-66 (Page ID # 5912-13).

The district court also credited the testimony of Professor Roscigno as establishing Senate factors one, two, three, five, six, seven, and nine such that "SB 238 and Directive 2014-17 interact with the historical and social conditions facing African Americans in Ohio to reduce the opportunity to participate in the political process relative to other groups of voters." R. 72 (D. Ct. Op. and Order at 65) (Page ID # 5912). While again conceding that Plaintiffs had not established that voter turnout would necessarily be decreased overall, the district court explained that "by its plain terms, § 2 is not necessarily about voter turnout but about opportunity to participate in the political process compared to other groups." Id. at 67 (Page ID #5914). And Plaintiffs had, the district court concluded, "demonstrated a strong likelihood of establishing that the combined effects of SB 238 and Directive 2014-17 result in fewer opportunities for African Americans to participate in the electoral process." Id. at 68 (Page ID # 5915) (emphasis in original).

Defendants make four arguments as to why the district court's analysis is incorrect. First, they argue that the district court improperly used the EIP voting system as it existed before SB 238 and Directive 2014-17 as its benchmark against which to measure the discriminatory effect on African American voters. Appellants Br. at 39. In so doing, Defendants assert, the district court used a "retrogression" standard, which is permissible only under Section 5 of the Voting Rights Act. Id. at 39-42. Second, Defendants contend that the canon of constitutional avoidance and federalism counsel in favor of not reading Section 2 to reach changes to early-voting systems at all. Id. at 44. Third, Defendants argue that the district court "should not have looked at [the Senate] factors at all to resolve this vote-denial claim." Id. at 53 (emphasis in original). Finally, focusing on the 2010 election data, Defendants contend that "Plaintiffs' own evidence shows that it cannot prove causation." Id. at 55.

1. The Test for a Section 2 Vote Denial Claim

Section 2 applies to any discriminatory "standard, practice, or procedure... which results in a denial or abridgement" of the right to vote. 42 U.S.C. § 1973(a) (emphasis added). Abridgement's "core meaning is `shorten,'" Reno v. Bossier Parish School Bd. ("Bossier II"), 528 U.S. 320, 333-34, 120 S.Ct. 866, 145 L.Ed.2d 845 (2000), or "[t]o reduce or diminish," Black's Law Dictionary (9th ed.2009). Similarly, Section 2(b) directs courts to consider whether members of a protected class have "less opportunity" to exercise their right to vote than other groups of voters, not simply whether protected voters only have no opportunity to vote. 42 U.S.C. § 1973(b) (emphasis added). In other words, Section 2 applies to any "standard, practice, or procedure" that makes it harder for an eligible voter to cast a ballot, not just those that actually prevent individuals from voting. Cf. Gingles, 478 U.S. at 45 n. 10, 106 S.Ct. 2752 ("Section 2 prohibits all forms of voting discrimination, not just vote dilution."); S.REP. NO. 97-417 at 30 ("Section 2 remains the major statutory prohibition of all voting rights discrimination.").

Courts have therefore found a range of "standard[s], practice[s], or procedure[s]" that make it harder, but not necessarily impossible, for eligible voters to vote to fall within Section 2. See, e.g., Mississippi State Chapter, Operation Push v. Allain, 674 F.Supp. 1245 (N.D.Miss.1987), aff'd sub nom. Mississippi State Chapter, Operation Push, Inc. v. Mabus, 932 F.2d 400 (5th Cir.1991) (restrictions on voter registration opportunities); Spirit Lake Tribe v. Benson Cnty., No. 2:10-cv-095, 2010 WL 4226614 (D.N.D. Oct. 21, 2010) (location of polling places); Brown v. Dean, 555 F.Supp. 502 (D.R.I.1982) (same); Brooks v. Gant, No. Civ. 12-5003, 2012 WL 4482984 (D.S.D. Sept. 27, 2012) (number and location of early voting sites); Harris v. Graddick, 615 F.Supp. 239 (M.D.Ala. 1985) (number of minority poll officials).

Thus, Plaintiffs' claim that SB 238 and Directive 2014-17 disproportionately place burdens on African American voters that make it harder for them to exercise their right to vote than other groups of voters is encompassed within Section 2. It does not matter that Plaintiffs do not argue that they are completely prevented from voting.

We also find unconvincing Defendants' argument that Section 2 does not cover challenges to early-voting systems, or that the canon of constitutional avoidance or federalism concerns compel such a conclusion. We first note that Defendants have raised these arguments for the first time on appeal. Generally, arguments raised for the first time on appeal are forfeited. See, e.g., Armstrong v. City of Melvindale, 432 F.3d 695, 700 (6th Cir. 2006) ("[T]he failure to present an issue to the district court forfeits the right to have the argument addressed on appeal."). Nevertheless, we briefly address both arguments for the sake of completeness.

As discussed above, the plain language of Section 2 does not exempt early-voting systems from its coverage. Section 2 applies to any discriminatory "standard, practice, or procedure ... which results in a denial or abridgement" of the right to vote. 42 U.S.C. § 1973(a). It does not specify that certain "standard[s], practice[s], or procedure[s]" are included within its scope and others excluded. The Voting Rights Act contains a similarly broad definition of the right to vote:

42 U.S.C. § 1973l (c)(1). This definition specifically notes that the right to vote "includ[es], but [is] not limited to," the specified examples. Id.

Nor has any court held that the Voting Rights Act does not apply to early-voting systems. The Supreme Court has in fact held that the Voting Rights Act should be interpreted broadly:

Chisom v. Roemer, 501 U.S. 380, 403-04, 111 S.Ct. 2354, 115 L.Ed.2d 348 (1991) (emphasis added) (internal citations omitted). Johnson v. Governor of Florida, 405 F.3d 1214 (11th Cir.2005), in which the Eleventh Circuit held that the Voting Rights Act does not apply to felon disenfranchisement laws, is easily distinguishable because the court determined that the Fourteenth Amendment in fact permits such laws and that the "Senate and House reports" for the Voting Rights Act "strongly suggest ... that Congress did not intend Section 2 of the Voting Rights Act to cover felon disenfranchisement provisions." Id. at 1228-29, 1232-33. Defendants do not point to a textual basis in the Fourteenth Amendment or specific statements in the legislative history of the Voting Rights Act that would compel a similar conclusion here.

Regarding Defendants' federalism arguments, we find Congress's statement in Section 2 that it applies to any discriminatory "standard, practice, or procedure" provides a sufficiently clear statement of its intention to change the federal-state balance to encompass a "standard, practice, or procedure" related to early voting if it produces discriminatory results in a way prohibited by Section 2. "Congress considered the [results] test of [Section 2] `necessary and appropriate to ensure full protection of the Fourteenth and Fifteenth Amendments rights.'" Bush v. Vera, 517 U.S. 952, 992, 116 S.Ct. 1941, 135 L.Ed.2d 248 (1996) (O'Connor, J., concurring) (quoting S.REP. NO. 97-417 at 27). The Supreme Court has elsewhere repeatedly held that "the Reconstruction Amendments by their nature contemplate some intrusion into areas traditionally reserved to the States." Lopez v. Monterey Cnty., 525 U.S. 266, 282, 119 S.Ct. 693, 142 L.Ed.2d 728 (1999). Finally, we note that the plain text of the National Voter Registration Act does not contain a clear statement to the opposite effect; it specifically states that "[n]othing in this chapter authorizes or requires conduct that is prohibited by the Voting Rights Act of 1965." 52 U.S.C. § 20510(d)(2).

Turning to the case law on Section 2, then, we note that the vast majority of cases have concerned a different kind of claim — vote dilution. Vote dilution claims involve challenges to methods of electing representatives — like redistricting or at-large districts — as having the effect of diminishing minorities' voting strength. Unsurprisingly, then, the case law has developed to suit the particular challenges of vote dilution claims. A clear test for Section 2 vote denial claims — generally used to refer to any claim that is not a vote dilution claim — has yet to emerge. See, e.g., Janai S. Nelson, The Causal Context of Disparate Vote Denial, 54 B.C. L. REV. 579, 595 (2013) ("[T]he legal contours of vote denial claims remain woefully underdeveloped as compared to vote dilution claims.").

We read the text of Section 2 and the limited relevant case law as requiring proof of two elements for a vote denial claim. First, as the text of Section 2(b) indicates, the 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." 42 U.S.C. § 1973(a)-(b). Second, the Supreme Court has indicated that that 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.9 Gingles, 478 U.S. at 47, 106 S.Ct. 2752 ("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."). In assessing both elements, courts should consider "the totality of circumstances." 42 U.S.C. § 1973(b).

Despite Defendants' assertions to the contrary, we see no reason why the Senate factors cannot be considered in assessing the "totality of the circumstances" in a vote denial claim, particularly with regard to the second element. While the Court has noted that the Senate Report indicates that "the enumerated factors will often be pertinent to certain types of § 2 violations, particularly to vote dilution claims," neither the Report nor the Court suggested that the factors can be considered only in vote dilution claims. Gingles, 478 U.S. at 45, 106 S.Ct. 2752. And several of the few Circuit court decisions to address vote denial claims have expressly stated that the Senate factors are relevant to vote denial claims. See, e.g., Gonzalez v. Arizona, 677 F.3d 383, 405-06 (9th Cir.2012) (en banc) (considering the Senate factors in evaluating a Section 2 challenge to Arizona's voter ID law), aff'd on other grounds sub nom. Arizona v. Inter Tribal Council of Arizona, ___ U.S. ___, 133 S.Ct. 2247, 186 L.Ed.2d 239 (2013); Johnson, 405 F.3d at 1227 n. 26 (explaining that, in a vote denial claim, "courts consider a non-exclusive list of objective factors (the `Senate factors') detailed in a Senate Report accompanying the 1982 amendments" as part of evaluating whether, under the "totality of the circumstances," "the political processes... are not equally open to participation by [members of a protected class]"); Smith, 109 F.3d at 596 n. 8 ("Appellants suggest that the `Senate factors' apply only to `vote dilution' claims. To the contrary, the `totality of the circumstances' test established in § 2(b) was initially applied only in `vote denial' claims such as this.").

We find Senate factors one, three, five, and nine particularly relevant to a vote denial claim in that they specifically focus on how historical or current patterns of discrimination "hinder [minorities'] ability to participate effectively in the political process." Gingles, 478 U.S. at 37, 106 S.Ct. 2752 (quoting Senate factor five). All of the factors, however can still provide helpful background context to minorities' overall ability to engage effectively on an equal basis with other voters in the political process.

2. Assessment of Plaintiffs' Section 2 Claim

We conclude that the district court properly found that Plaintiffs are likely to succeed on their Section 2 claim. Plaintiffs demonstrated that SB 238 and Directive 2014-17 will disproportionately burden African American voters and that this burden means that they will have a harder time voting than other members of the electorate. As previously discussed in Sections II and III, the district court did not clearly err in crediting the statistical and survey analysis of Plaintiffs' expert Smith and other studies in the record as demonstrating that African Americans vote EIP at higher rates than other groups, including on the eliminated EIP voting days. Nor did the district court clearly err in considering data from the elections in 2008, 2010, and 2012 in reaching this conclusion, rather than focusing solely on 2010 as Defendants' causation argument implicitly urges. Defendants' argument that Plaintiffs did not establish causation is therefore without merit. And the fact that African Americans are more likely to be of lower-socioeconomic status in Ohio — as Professor Roscigno's undisputed report establishes — and both distrust voting by mail and face obstacles doing so, means that the remaining EIP voting times are not sufficient to ensure African Americans have a truly equal opportunity to vote as other groups of voters. More specifically, African Americans are more likely to vote on Sundays through the Souls to the Polls initiatives because of the free transportation church groups can provide. Lower-income individuals face difficulties in voting during the day because they are more likely to work in hourly-wage jobs with little flexibility. Lower-income individuals, often because they are more likely to move and/or have difficulty accessing transportation, also most need same-day registration. R. 18-2 (Roscigno Rep. at 16-19) (Page ID # 266-69). Thus, the disproportionate burdens SB 238 and Directive 2014-17 place on African Americans, combined with their lower-socioeconomic status in Ohio, operate to give African American voters "less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice." 42 U.S.C. § 1973(b).

Defendants' argument that the district court failed to identify an objective benchmark against which to assess the burdens SB 238 and Directive 2014-17 place on African American voters is unpersuasive. The case law Defendants cite on the need for objective benchmarks involve vote dilution claims. See, e.g., Holder v. Hall, 512 U.S. 874, 876, 114 S.Ct. 2581, 129 L.Ed.2d 687 (1994) ("This case presents the question whether the size of a governing authority is subject to a vote dilution challenge under § 2 of the Voting Rights Act"); Bossier II, 528 U.S. 320, 120 S.Ct. 866 (considering whether Section 5 prohibits preclearance of a dilutive, but nonretrogressive redistricting plan).10 A vote dilution claim requires courts to make the difficult judgment of whether a challenged practice impermissibly dilutes minorities' voting strength, or whether minorities' lack of electoral success in fact simply stems from "mere ... political defeat at the polls." Whitcomb v. Chavis, 403 U.S. 124, 153, 91 S.Ct. 1858, 29 L.Ed.2d 363 (1971). Thus, determining what an undiluted benchmark should be can be challenging, particularly because Section 2 expressly 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." 42 U.S.C. § 1973(b).

In contrast, Section 2 vote denial claims inherently provide a clear, workable benchmark. Again, under Section 2(b), the relevant inquiry is whether minority voters "have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice." 42 U.S.C. § 1973(b) (emphasis added). The benchmark is thus quite straightforward — under the challenged law or practice, how do minorities fare in their ability "to participate in the political process" as compared to other groups of voters ?

Thus, the district court properly assessed whether, under the system established by SB 238 and Directive 2014-17, African American voters would have less opportunity than other Ohio voters to cast their ballots. The disproportionate impact on African American voters as compared to white voters of cutting Sunday and evening EIP voting as well as Golden Week is clearly relevant to this inquiry. So, too, is the relative ability of African American voters to vote through the remaining options available under SB 238 and Directive 2014-17.

Turning to the second element of a Section 2 vote denial claim, the district court also properly found that Plaintiffs had sufficiently demonstrated that the disproportionate burdens SB 238 and Directive 2014-17 place on African Americans are in part caused by or linked to "social and historical conditions" of discrimination. Professor Roscigno's undisputed report regarding Senate factor five is particularly on point. He explained that African Americans in Ohio tend to be of lower-socioeconomic status because of "stark and persistent racial inequalities ... [in] work, housing, education and health," inequalities that stem from "both historical and contemporary discriminatory practices." R. 18-2 (Roscigno Rep. at 3) (Page ID #253). As support, Professor Roscigno pointed to "[s]ubstantial bodies of social science research ... [that] investigate the root causes of ... occupational inequalities, often concluding that contemporary institutional practices and discrimination play a significant role, especially when the disparities are as large as they are in Ohio." Id. at 7 (Page ID # 257). "Racial occupational inequalities are easily linked to racial disparities in, for instance, family income and poverty status as well as residential and schooling options and racial health disparities." Id. at 10 (Page ID # 260). As previously discussed, African Americans' lower-socioeconomic status in turn plays a key role in explaining why the disproportionate impact of SB 238 and Directive 2014-17 burdens African Americans' voting opportunities.

The remaining Senate factors considered by the district court support its finding that the burdens SB 238 and Directive 2014-17 place on African American voters are in part caused by or linked to "social and historical conditions" that have produced or currently produce discrimination against African Americans in Ohio. Given that Plaintiffs' claim focuses on reductions in voting opportunities, past and current discrimination linked to the electoral system itself is relevant. Professor Roscigno's report demonstrated that, per Senate factor one, historically "[o]fficial voting-related discrimination against racial/ethnic minorities was a cornerstone in Ohio." Id. at 26 (Page ID # 276). More recently, Ohio has implemented "voting practices that suppress minority political participation," such as poll watchers and voter ID laws-practices that fall within the ambit of Senate factor three. Id. at 28-30 (Page ID # 278-80). Under Senate factor nine, it is also relevant that, as explained in the Equal Protection Clause analysis, the policy justifications for SB 238 and Directive 2014-17 are "tenuous." Gingles, 478 U.S. at 37, 106 S.Ct. 2752. The evidence Professor Roscigno provided regarding Senate factors two, six, and seven — while not as clearly linked to Plaintiffs' claim — further contextualizes the role race still plays in Ohio elections.

The district court did not improperly engage in a retrogression analysis in considering the opportunities available to African Americans to vote EIP under the prior law as part of the "totality of circumstances" inquiry. To be sure, Congress intended — and the Court has read — Section 2 and Section 5 not to have exactly the same scope. Procedurally, Section 5 requires that covered states obtain preclearance from the Attorney General or the District Court for the District of Columbia before they change a voting "qualification, prerequisite, standard, practice, or procedure." 42 U.S.C. § 1973c. Section 2 applies to all states and includes no preclearance requirement. "[T]he purpose of § 5 has always been to insure that no voting-procedure changes would be made that would lead to a retrogression in the position of racial minorities with respect to their effective exercise of the electoral franchise." Beer v. United States, 425 U.S. 130, 141, 96 S.Ct. 1357, 47 L.Ed.2d 629 (1976). In other words, "§ 5 prevents nothing but backsliding," whereas Section 2 is aimed at combatting "discrimination more generally." Bossier II, 528 U.S. at 334-35, 120 S.Ct. 866.

At the same time, however, no case explicitly holds that prior laws or practices cannot be considered in the Section 2 "totality of circumstances" analysis. The Supreme Court has in fact found a Section 2 violation under the "totality of circumstances" in part based on changes to the electoral system. League of United Latin Am. Citizens v. Perry, 548 U.S. 399, 439, 126 S.Ct. 2594, 165 L.Ed.2d 609 (2006) (noting that "[t]he changes to District 23 undermined the progress of a racial group that has been subject to significant voting-related discrimination and that was becoming increasingly politically active and cohesive") (emphasis added). Neither does the legislative history of Section 2 indicate that prior laws or practices cannot factor into the analysis. The Senate Report to the 1982 amendments states only that "Plaintiffs could not establish a Section 2 violation merely by showing that a challenged reapportionment or annexation, for example, involved a retrogressive effect on the political strength of a minority group." S.REP. NO. 97-417 at 68 n.224 (1982) (emphasis added). Oxford English Dictionary defines "merely" as "entirely ... only (what is referred to) and nothing more." 9 Oxford English Dictionary 629 (2d ed.1989). In other words, the Report simply establishes that challengers cannot show a Section 2 violation only on the basis of retrogressive effects. Indeed, Senate factor nine in fact invites comparison to prior laws. The Senate Report explains that, in assessing the "tenuous[ness]" of policy justification for the challenged law or procedure, the fact that "the procedure markedly departs from past practices or from practices elsewhere in the jurisdiction ... bears on the fairness of its impact." S.REP. NO. 97-417 at 239 n.117 (emphasis added).

Moreover, the Supreme Court has made clear that "some parts of the § 2 analysis may overlap with the § 5 inquiry." Georgia v. Ashcroft, 539 U.S. 461, 478, 123 S.Ct. 2498, 156 L.Ed.2d 428 (2003). Both Section 2 and Section 5 speak to "abridgement." 42 U.S.C. § 1973(a) (Section 2) ("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 ... as provided in subsection (b) of this section.") (emphasis added); id. § 1973c(b) (Section 5) ("Any voting qualification or prerequisite to voting, or standard, practice, or procedure with respect to voting that has the purpose of or will have the effect of diminishing the ability of any citizens of the United States on account of race or color ... to elect their preferred candidates of choice denies or abridges the right to vote within the meaning of subsection (a) of this section.") (emphasis added). Abridgement "necessarily entails a comparison." Bossier II, 528 U.S. at 333-34, 120 S.Ct. 866.

Rather, what is distinct between a Section 5 analysis and a Section 2 analysis is the role that prior law plays in the comparison. The retrogression analysis under Section 5 involves comparing voting opportunities enjoyed by minorities under the status quo as compared to voting opportunities minorities would have under the electoral system if the proposed change is implemented. The focus is solely on voting opportunities enjoyed by minorities, and whether those opportunities would be reduced under the proposed law. In contrast, under the Section 2 analysis, the focus is whether minorities enjoy less opportunity to vote as compared to other voters. The fact that a practice or law eliminates voting opportunities that used to exist under prior law that African Americans disproportionately used is therefore relevant to an assessment of whether, under the current system, African Americans have an equal opportunity to participate in the political process as compared to other voters.

As discussed above, Plaintiffs have not asserted that SB 238 and Directive 2014-17 burden their right to vote merely because they take away EIP voting times that used to exist under prior law. Rather, they have presented evidence that the eliminated EIP voting times are those that African Americans disproportionately use, and that racial inequalities in socioeconomic status and other factors make it much more difficult for African Americans to vote at the remaining times or through the other methods now available under the status quo as compared to other groups.11 Thus, the district court did not improperly engage in a retrogressive analysis.

Finally, we find unpersuasive Defendants' claim that our decision would have far-reaching implications beyond Ohio. Defendants fail to cite any Supreme Court or Sixth Circuit authority to support their argument that we must necessarily consider the practices of other states with regard to EIP voting.

In fact, the text of Section 2 and Supreme Court decisions indicate that the opposite is true. The text of Section 2 directs courts to examine whether the political processes "in the State or political subdivision are not equally open to participation by members of a [protected] class... in that its members have less opportunity than other members of the electorate to participate." 42 U.S.C. § 1973(b) (emphasis added). The focus is on the internal processes of a single State or political subdivision and the opportunities enjoyed by that particular electorate. The text of Section 2 does not direct courts to compare opportunities across States. The Supreme Court has likewise characterized the determination of whether a practice violates Section 2 as "an intensely local appraisal of the design and impact of the contested electoral mechanisms." Gingles, 478 U.S. at 79, 106 S.Ct. 2752 (emphasis added) (internal quotation marks omitted). The Court explained that "the [Senate] Committee determined that `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. (quoting S.REP. NO. 97-417 at 30) (internal citations omitted). "This determination is peculiarly dependent upon the facts of each case." Id. (emphasis added) (quoting Rogers v. Lodge, 458 U.S. 613, 621, 102 S.Ct. 3272, 73 L.Ed.2d 1012 (1982)). See also Holder, 512 U.S. at 881-82, 114 S.Ct. 2581 ("It makes little sense to say ... that the sole commissioner system should be subject to a [Section 2] dilution challenge if it is rare — but immune if it is common.") (Kennedy, J.); Gonzalez, 677 F.3d at 406 ("Because a § 2 analysis requires the district court to engage in a `searching practical evaluation of the `past and present reality,' Gingles, 478 U.S. at 45, 106 S.Ct. 2752, ... a district court's examination in such a case is `intensely fact-based and localized,' Salt River, 109 F.3d at 591.'").

Defendants tell us that the Court's statement in Gingles is not relevant because the Court was considering a vote dilution claim, rather than a vote denial claim like that at issue here. Appellants Br. at 52. But they provide no explanation as to why that distinction matters or why assessments of early voting systems do not likewise require an intensely localized assessment of their impact on African American voters.

Ohio faced unique problems in administering the 2004 elections. The General Assembly introduced early voting in 2005 largely to remedy those problems. In the nearly ten years since, EIP voting has come to play a special role in Ohio in ensuring that African Americans have an equal opportunity to participate in the political process that is not necessarily true elsewhere. There is no reason to think our decision here compels any conclusion about the early-voting practices in other states, which do not necessarily share Ohio's particular circumstances.

In sum, we conclude that Plaintiffs are likely to succeed on their Section 2 of the Voting Rights Act claim.

E. The Remaining Preliminary Injunction Factors

"When a party seeks a preliminary injunction on the basis of a potential constitutional violation, `the likelihood of success on the merits often will be the determinative factor.'" Obama for America, 697 F.3d at 436 (quoting Jones v. Caruso, 569 F.3d 258, 265 (6th Cir.2009)). Although we have concluded that Plaintiffs are likely to succeed on their Equal Protection Clause claim and their Section 2 of the Voting Rights Act claim, we still consider the remaining three preliminary injunction factors.

First, we conclude that Plaintiffs have established that they will suffer irreparable harm absent an injunction. "When constitutional rights are threatened or impaired, irreparable injury is presumed. A restriction on the fundamental right to vote therefore constitutes irreparable injury." Id. (internal citations omitted).

Regarding the final two factors, we conclude that the issuance of the injunction would not cause substantial harm to others and that the public interest weighs in its favor. The significant burden on the ability to vote of the voters Plaintiffs represent outweighs any burden on Defendants, who have failed to demonstrate that BOEs would not be able to administer the extra days and evening hours of EIP voting required by the preliminary injunction, a schedule BOEs had previously administered in past elections. As we explained in Obama for America:

Obama for America, 697 F.3d at 436-37. We have previously rejected Defendants' argument that upholding the preliminary injunction risks confusing voters. Ohio State Conference of N.A.A.C.P. v. Husted, No. 14-3877, ___ F.3d ___, ___, 2014 WL 4494938, at *5 (6th Cir. Sept. 12, 2014). Regarding Defendants' laches argument, we "review[] a district court's resolution of a laches question for an abuse of discretion." Chirco v. Crosswinds Communities, Inc., 474 F.3d 227, 231 (6th Cir. 2007) (internal quotation marks omitted). Defendants have not explained how Plaintiffs could have more quickly produced the voluminous record evidence in this case in support of their motion for a preliminary injunction, nor have Defendants produced any evidence that Plaintiffs purposefully delayed. We therefore find that the district court did not abuse its discretion in determining that Defendants "ha[d] not shown ... a lack of diligence" by Plaintiffs to move promptly for a preliminary injunction. R. 72 (D. Ct. Op. and Order at 6) (Page ID # 5853).

F. The District Court's Remedy

Defendants suggest in one paragraph in their brief on appeal that the district court's preliminary injunction itself might violate the Equal Protection Clause by allowing BOEs to set EIP voting hours in addition to those set forth in the preliminary injunction order and Directive 2014-17 because BOEs might set unequal EIP voting hours. Appellants Br. at 28-29. This argument appears to implicate issues of state law that have not been fully developed on appeal. The Ohio Revised Code authorizes BOEs to set their own EIP voting hours. Ohio Rev.Code §§ 3501.10(b), 3501.11. The district court's order expressly ties its remedy to what is already permitted under Ohio law: "Secretary Husted is enjoined from preventing individual county Boards of Election from adopting, by a majority vote of their members and in accordance with the procedures established by Ohio election law," additional EIP voting hours. R. 72 (D. Ct. Op. and Order at 71) (Page ID # 5918) (emphasis added). Neither party has directly addressed Secretary Husted's authority to impose uniform EIP voting hours despite the provisions of the Ohio Revised Code that appear to vest discretion in BOEs to set their own hours. Id. at 4 (Page ID # 5851). At this stage in the litigation, with only days until early voting is set to begin, and having found that all four factors weigh in favor of granting Plaintiffs a preliminary injunction, we do not address this argument at this time.

III. CONCLUSION

For the foregoing reasons, we AFFIRM the district court's judgment granting a preliminary injunction.