PER CURIAM:

The State of Florida brings this action for declaratory relief under section 5 of the Voting Rights Act of 1965, 42 U.S.C. § 1973c. Section 5 prohibits the enforcement of any change in voting practices or procedures in certain states and other covered jurisdictions, including five of Florida's counties, unless and until the change is approved by the Attorney General of the United States or by a three-judge panel of the United States District Court for the District of Columbia. In this three-judge district court proceeding, Florida seeks a declaratory judgment that certain recent changes to its election laws "neither ha[ve] the purpose nor will have the effect of denying or abridging the right to vote on account of race or color" or membership in a language minority group. 42 U.S.C. § 1973c(a).

Florida amended its voting laws in 2011, making some 80 sets of changes from prior provisions. The State submitted the changes to the Attorney General for administrative approval, and the Attorney General precleared 76 of them. Florida then withdrew the remaining four sets of changes from its request for administrative preclearance, and instead filed a complaint seeking judicial preclearance of those changes. After the complaint was filed, Florida resubmitted one of the four changes (regarding the procedures for constitutional amendments proposed by initiative) to the Attorney General for administrative approval. Thereafter, the Attorney General precleared that change, and Florida voluntarily dismissed that count of its complaint. One week ago, Florida filed a motion to voluntarily withdraw another set of changes (regarding requirements for third-party voter registration organizations (TPROs)), and to inform the court that it had amended several of the TPRO changes for which it still seeks preclearance. Because this development requires a new round of briefing and review, we will address the remaining TPRO changes at a later date.

Two categories of voting changes will be addressed in this opinion. In brief, those changes would: (1) amend the available days and hours that Florida counties may use for early in-person voting, see Fla. Stat. § 101.657(d) (2011); and (2) amend the voting procedures for registered voters who move between Florida counties and seek to vote in their new county of residence ("inter-county movers"), see id. § 101.045. The parties have filed extensive submissions regarding the law applicable to our task of reviewing the voting changes for purposes of preclearance. As we discuss below, Florida's submission urges an unconventional reading of section 5, which we largely reject. The Attorney General, supported by more than two dozen individuals and organizations who were permitted to intervene as defendants in this action, proffers a more traditional reading, which we largely adopt.

The parties have also developed a voluminous evidentiary record, comprised of over 11,000 pages of legislative hearings, deposition transcripts, expert reports, and other exhibits.1 The parties then filed proposed findings of fact and conclusions of law on the basis of that written record. The parties agreed that the record was sufficient for the court to reach a decision, and that live trial testimony was unnecessary. The court did, however, hear five hours of oral argument on all aspects of the statutory preclearance question. Thereafter, the court received supplemental briefing and submissions from the parties on several discrete questions of law and fact.

Upon consideration of the entire record, our conclusions may be summarized as follows. First, we conclude that we cannot, at this time, preclear Florida's early voting changes because the State has failed to satisfy its burden of proving that those changes will not have a retrogressive effect on minority voters. Specifically, the State has not proven that the changes will be nonretrogressive if the covered counties offer only the minimum number of early voting hours that they are required to offer under the new statute, which would constitute only half the hours required under the prior law. Following an approach approved by the Supreme Court, however, we also conclude that if Florida and the covered counties were to submit a preclearance plan that offered early voting for the maximum number of hours authorized by the new statute, which would be exactly the same number as under the prior law, and did so on a standard 7 a.m. to 7 p.m. schedule, it is likely that Florida would be able to satisfy its burden of proving that the overall effect of its early voting changes would be nonretrogressive. Second, we conclude that Florida has satisfied its burden of proving that the changes to the procedures for inter-county movers neither were enacted with a discriminatory purpose nor will have a retrogressive effect on minority voters, and that those changes are therefore entitled to preclearance.

The opinion that follows summarizes our findings of fact and sets forth our conclusions of law on the question of statutory preclearance.2 The appendix to this opinion separately sets forth our findings of fact. See FED.R.CIV.P. 52.

I. Background and Procedural History

A. Statutory Background

This court has been convened as a three-judge district court with jurisdiction to hear and determine this declaratory judgment action under section 5 of the Voting Rights Act of 1965, 42 U.S.C. § 1973c. See 28 U.S.C. § 1346(a)(2); id. §§ 2201, 2284. The Act was enacted to protect the fundamental rights guaranteed by the Fifteenth Amendment, and to "banish the blight of racial discrimination in voting." South Carolina v. Katzenbach, 383 U.S. 301, 308, 86 S.Ct. 803, 15 L.Ed.2d 769 (1966). Section 5 requires covered states and political subdivisions to seek advance approval, or "preclearance," from the Attorney General or a three-judge district court before administering any new "voting qualification or prerequisite to voting, or standard, practice, or procedure with respect to voting." 42 U.S.C. § 1973c(a). To obtain preclearance, a covered jurisdiction must show that its proposed voting change "neither has the purpose nor will have the effect of denying or abridging the right to vote on account of race or color, or in contravention of the guarantees set forth in section 1973b(f)(2) of this title [proscribing voting restrictions based on membership in a language minority group]." Id.3

Five of Florida's sixty-seven counties are subject to the preclearance requirements of section 5. They, along with a number of states and other local jurisdictions, were subjected to section 5 coverage by the 1975 amendments to the coverage formula set forth in section 4(b) of the Act. Section 4(b)'s coverage formula subjects jurisdictions to section 5 preclearance obligations based on a combination of the maintenance of a prohibited "test or device" and low voter turnout in certain elections.4 The 1975 amendments expanded the definition of "test or device" to include "any practice or requirement by which any State or political subdivision provided," as of November 1, 1972, "any registration or voting notices, forms, instructions, assistance, or other materials or information relating to the electoral process, including ballots, only in the English language, where the Director of the Census determines that more than five per centum of the citizens of voting age residing in such State or political subdivision [we]re members of a single language minority." 42 U.S.C. § 1973b(f)(3). The statute defines "language minorities" to include "persons who are American Indian, Asian American, Alaskan Native or of Spanish heritage." Id. § 1973l(c)(3). As a result of these language minority amendments, five Florida counties — Hillsborough, Monroe, Collier, Hendry, and Hardee — became subject to coverage under section 5. See 28 C.F.R. Pt. 51, App.; see also 41 Fed.Reg. 34329 (Aug. 13, 1976); 40 Fed.Reg. 43746 (Sept. 23, 1975).

Although Florida itself is not a covered jurisdiction under section 5, it is well settled that "the Act's preclearance requirements apply to measures mandated by a noncovered State to the extent that these measures will effect a voting change in a covered county." Lopez v. Monterey Cnty., 525 U.S. 266, 269, 119 S.Ct. 693, 142 L.Ed.2d 728 (1999). Accordingly, to the extent that Florida seeks to administer any of its statewide voting changes in its five covered counties, those changes must be submitted for preclearance. Id. at 278, 119 S.Ct. 693; see 28 C.F.R. § 51.23(a). Moreover, for reasons explained in more detail below, see infra Part II.A, it is clear that both of the electoral changes for which Florida seeks approval come within the purview of the Act, because they both involve changes to voting qualifications or prerequisites to voting, or "standard[s], practice[s], or procedure[s] with respect to voting." 42 U.S.C. § 1973c(a).5 Indeed, although Florida has argued that certain formulations of the section 5 "effect" test do not apply to these changes and that the requirement that its statewide changes be submitted for preclearance is unconstitutional, the State does not dispute its statutory obligation under section 5 to submit the changes at issue here for either administrative or judicial preclearance. See [91] Fla.'s Proposed Findings of Fact & Conclusions of Law ("Fla. Br."); [147] Third Am. Compl. ("Fla. Compl.").

B. Procedural History

The voting changes that Florida has submitted for our preclearance review were included in an omnibus bill, Committee Substitute for Committee Substitute for House Bill 1355 ("HB 1355"), which made approximately 80 sets of changes to Florida's election procedures.6 That bill moved through the Florida House and Senate over a period of several months in the spring of 2011, and was signed into law on May 19, 2011 by the Governor. See Fla. Laws, ch. 2011-40; see also Fla. Stat. §§ 101.045, 101.657(d) (2011) (codifying the particular voting changes at issue in this case). As we have indicated, Florida originally submitted HB 1355 in its entirety for administrative preclearance by the Department of Justice, and the vast majority of the changes in that law — most of which were minor or ministerial in nature — were precleared by the Attorney General on August 8, 2011. See A118 (Revised Jointly Stip. Facts ¶ 8).

In the meantime, however, Florida withdrew four sets of its voting changes from the administrative preclearance process. See A219 (Letter from Daniel Nordby, Gen. Counsel, Fla. Dep't of State (July 29, 2011)). Those four changes included the early voting and inter-county mover changes at issue before us in this opinion, along with changes imposing additional requirements on third-party voter registration organizations (TPROs) and amendments to the provisions governing the time frame during which signatures on citizen initiative petitions are valid (the "citizen initiative changes"). Id. After withdrawing those changes from administrative review, Florida then filed a complaint in this court seeking judicial preclearance of those four changes. See [1] Compl. Early in the litigation, Florida amended that complaint twice: once to assert claims — which we will address in future proceedings — challenging the constitutionality of the Voting Rights Act, see [39] First Am. Compl. ¶¶ 107-116; and once to include a request for judicial preclearance of amended state regulations implementing the TPRO changes, see [54] Second Am. Compl. ¶¶ 85-86 & Ex. B.

Over the following months, the parties conducted extensive discovery, and, after the discovery period closed on February 29, 2012, continued to take de bene esse depositions and submit declarations by their respective experts. By late May, the parties had submitted proposed conclusions of law and findings of fact. They then requested that we forego a live trial and decide the case on the basis of the written record alone. The court held five hours of oral argument on June 21, followed by multiple rounds of supplemental briefing on issues raised during the argument.

Since the filing of Florida's Second Amended Complaint, the scope of our required review has narrowed considerably. First, in March 2012 the United States informed this court that, after conducting discovery and reviewing the record, it had concluded that Florida had met its burden of demonstrating that the citizen initiative changes were neither enacted with a discriminatory purpose nor will have a retrogressive effect on minority voters. See [79] March 8, 2012 Mem. Order at 2-3. Accordingly, Florida re-submitted those changes for administrative preclearance, and on March 21, 2012, the Attorney General informed the State that no objection would be interposed. See [84] United States' Notice to the Court. The parties then filed a stipulation of dismissal as to Count Two of Florida's Second Amended Complaint, which dealt with those citizen initiative changes. See [85] Stip. of Dismissal.

Florida's stipulation of dismissal left us with only three statutory preclearance counts to review, dealing respectively with the TPRO, inter-county mover, and early voting amendments. But recent developments have again narrowed the issues before us. In particular, a federal district court in Florida issued a preliminary injunction on May 31, 2012 against many of the TPRO changes in HB 1355 and its implementing rule, finding that the plaintiffs in that case were likely to prevail on their First Amendment and National Voter Registration Act challenges to those changes. See League of Women Voters of Fla. v. Browning, 863 F.Supp.2d 1155, 1166-67 (N.D.Fla.2012). On June 12, Florida informed us that it would "no longer seek in this action to preclear the [TPRO] changes that the [district court in Florida] preliminarily enjoined." [109] Statement Regarding the Effect of the Recent Order in the N.D. Fla. at 2. Florida stated in that filing that it would "voluntarily withdraw the enjoined [TPRO] Changes from this judicial preclearance action," id. at 5, indicating that it was waiting for a clarification from the Florida district court before doing so, see id. at 3 n. 2.

Several weeks passed without further word from the State. We convened a status conference on July 16 to determine whether Florida had made a decision regarding which TPRO provisions it intended to withdraw from this action, and which it would still seek to preclear. Florida advised that it had not yet made a final decision. Then, approximately one week ago, the State informed us that it had settled the TPRO litigation in Florida. See [104] Mem. of Points and Authorities in Support of Pl.'s Mot. for Leave to Amend the Compl. at 1. It filed a motion to amend its complaint in this case, seeking to remove from its request for preclearance those TPRO changes that the district court in Florida had enjoined, while keeping the remaining changes, some in a modified form. See id. at 4 n. 1, 5. On August 13, Florida also advised the court that it would promptly submit the remaining, non-enjoined TPRO changes to the Department of Justice for administrative preclearance. On August 15, the court granted Florida's unopposed motion to amend its complaint to reflect that it no longer seeks preclearance for the enjoined changes.

The effect of this development has been to remove from our consideration the most controversial provisions of the TPRO amendments — including a stringent 48-hour deadline for submission of completed voter registration applications, a requirement that registration agents sign a "sworn statement" listing the penalties for fraudulent voter registration, and a mandate that TPROs track and account for all registration forms provided to and received from their employees and volunteers, and then file monthly reports with the State reflecting those tallies. But Florida's recent filing also means that we must hold a new round of briefing on the TPRO changes that remain. We also must give the Department of Justice an opportunity to consider whether to preclear the remaining changes administratively. This process could take several more weeks, even as the covered counties seek to finalize their plans for the November election.

Accordingly, having previously bifurcated this case into a preclearance part and a constitutional part in order to expedite the decision on preclearance, see supra note 2, we now trifurcate it. This opinion will address only those matters fully before us — the early voting and inter-county mover changes — for which there is no reason for further delay, and for which the interests of judicial efficiency and finality counsel an expeditious resolution. Once the parties have had an opportunity to brief the TPRO changes for which Florida still seeks preclearance, and the Attorney General has made a determination regarding whether to preclear those changes administratively, we will address the remaining TPRO provisions if there are any. Thereafter, if our dispositions have not rendered a decision on Florida's constitutional challenge moot or otherwise unnecessary, we will hold oral argument on those issues.

C. Voting Changes at Issue

In their final form, the two sets of voting amendments at issue in this opinion make a number of changes to Florida's practices and procedures governing early voting and inter-county movers. We examine those changes in greater detail in the course of our analysis and discussion below. In brief, however, the amendments make the following changes from Florida's previous voting laws.

1. Early Voting

Before the passage of HB 1355, Florida's early voting law provided early in-person voting for a potential 14-day period, beginning on "the 15th day before an election and end[ing] on the 2nd day before the election." Fla. Stat. § 101.657(d) (2010). That prior law required each county to offer early voting for exactly 8 hours per day on weekdays and exactly 8 hours in the aggregate each weekend, yielding a total of 96 hours of early voting. Early voting sites were to "open no sooner than 7 a.m. and close no later than 7 p.m. on each applicable day," but within this constraint local supervisors of elections in each Florida county had the discretion to select the specific voting hours for each early voting day. Id. In Florida's five covered counties, local supervisors of elections chose a range of voting hours between 7 a.m. and 7 p.m. depending on whether it was a primary or general election and what voting site was involved. In addition, the prior law gave election supervisors discretion in choosing whether their 8 aggregate weekend early voting hours would fall on a Saturday, a Sunday, or both. See id. In each of Florida's five covered counties, local election supervisors exercised that discretion to offer 8 hours of early voting on each Saturday and none on either Sunday. This meant that in each of the five covered counties, only 12 of the available 14 days of early voting were actually used (Monday through Saturday of each of the two weeks before the election).

HB 1355 amends the number of days, the number of hours, the specific hours, and the weekend times that early voting may be offered in Florida. First, under the new law, the early voting period begins "on the 10th day before an election ... and end[s] on the 3rd day before the election," for a total of only 8 days instead of the previous 12. Fla. Stat. § 101.657(d) (2011).

Second, the law now gives local election supervisors the discretion to determine the number of daily hours of early voting in their counties, subject to the constraint that "no less than 6 hours and no more than 12 hours" be offered on each of the 8 early voting days. Id. As a result, Florida's covered counties might still offer the same total hours of early voting (96 hours) that were required under the pre-2011 law, but only if their local election supervisors decide to offer the maximum 12 hours of early voting on each of the 8 days. If, on the other hand, a local supervisor chooses to offer the minimum number of hours (i.e., 6 hours per day), then the early voting period would last only 48 hours in total — exactly half of the hours that were offered under the prior law.

Third, the new law also removes the requirement that voting take place between 7 a.m. and 7 p.m. each day. Instead, local supervisors of elections have the discretion to determine the specific hours of early voting in their counties, as long as they offer "no less than 6 hours and no more than 12 hours" each day. Id. In the event that local election supervisors offer the maximum 12 hours of early voting on a given day, then the early voting hours for that day would necessarily include hours that fall outside the standard 8-hour workday extending from 9 a.m. to 5 p.m. However, in the event election supervisors offer the minimum 6 hours of early voting on a given day, the 6-hour early voting day may be entirely within, entirely outside, or straddle the standard workday.

Fourth, the new early voting statute also mandates some additional weekend hours of early voting. In particular, the new early voting period runs from the Saturday two weekends before the election to the Saturday immediately before Election Day, see id., meaning that the early voting period under HB 1355 now requires three weekend days of early voting: two Saturdays and one Sunday. And as already explained, the new statute mandates anywhere from 6 to 12 early voting hours on each day. Id. Accordingly, if HB 1355 were implemented in the covered counties, it would result in at least 6 (and up to 12) hours of Sunday early voting that were never before offered in those counties. Moreover, election officials would also have the discretion to offer up to 36 total hours of weekend early voting (12 hours per day on each of 3 weekend days) — for a net gain of 20 weekend early voting hours over the prior law, which offered exactly 16 weekend hours (8 hours in the aggregate on each of two weekends).

2. Inter-County Movers

Under the pre-2011 statute governing inter-county movers, registered Florida voters who moved to a new county of residence without informing the relevant supervisor of elections could still update their addresses at the polls and then vote a regular ballot. Specifically, before the enactment of HB 1355, inter-county movers were simply required to complete an affidavit of change of address (or a new voter registration application), listing their new address of residence and affirming that they had not already voted in the precinct of their former residence. See Fla. Stat. § 101.045(2)(a), (c) (2010). After their eligibility to vote was verified, such voters were then permitted to cast a regular ballot, which would be canvassed and counted like all other ballots. Id. § 101.045(2)(d).

After the 2011 amendments, however, "an elector whose change of address is from outside the county may not change his or her legal residence at the polling place and vote a regular ballot." Fla. Stat. § 101.045(2)(b) (2011). Instead, such inter-county movers are now only "entitled to vote a provisional ballot." Id. The only exception is for active uniformed services voters and members of their families, who are still permitted to cast a regular ballot after affirming their inter-county change of address and having their eligibility verified. Id.

Once completed, provisional ballots cast by inter-county movers "shall be placed in a secrecy envelope and thereafter sealed in a provisional ballot envelope." Id. § 101.048(1). The provisional ballots will then be deposited in a ballot box and returned to election officials, whereupon the county canvassing board will examine the ballots to determine if the voters were eligible to vote at that precinct and had not already cast a ballot in the election. Id. § 101.048(2)(a). Florida law specifically provides that a provisional ballot "shall be counted unless the canvassing board determines by a preponderance of the evidence that the person was not entitled to vote." Id. (emphasis added); see also id. § 101.048(2)(b)(1).

Finally, as relevant here, the 2011 amendments to Florida's Election Code also make pre-election address changes easier for registered Florida voters. Under the pre-2011 law, inter-county address changes had to be completed "using a voter registration application signed by the elector." Fla. Stat. § 97.1031(2) (2010). Now, however, registered voters can notify the supervisor of elections of their change of address by "[s]ubmitting the change on a voter registration application or other signed, written notice," or by "[c]ontacting the supervisor of elections via telephone or electronic means." Fla. Stat. § 97.1031(1)(b) (2011).

* * *

Florida seeks judicial preclearance of each set of voting changes outlined above. To obtain preclearance under section 5, a covered jurisdiction must demonstrate that its proposed voting changes "neither ha[ve] the purpose nor will have the effect of denying or abridging the right to vote on account of race or color, or in contravention of the guarantees set forth in section 1973b(f)(2) of this title [proscribing voting restrictions based on membership in a language minority group]." 42 U.S.C. § 1973c(a). The burden of proving by a preponderance of the evidence that the voting changes at issue neither have an impermissible purpose nor will have an impermissible effect rests on the plaintiff — in this case, the State of Florida. Reno v. Bossier Parish Sch. Bd. ("Bossier Parish I"), 520 U.S. 471, 478, 117 S.Ct. 1491, 137 L.Ed.2d 730 (1997); City of Rome v. United States, 446 U.S. 156, 184 n. 18, 100 S.Ct. 1548, 64 L.Ed.2d 119 (1980); Beer v. United States, 425 U.S. 130, 140-41, 96 S.Ct. 1357, 47 L.Ed.2d 629 (1976); Georgia v. United States, 411 U.S. 526, 538, 93 S.Ct. 1702, 36 L.Ed.2d 472 (1973); City of Port Arthur v. United States, 517 F.Supp. 987, 1011 (D.D.C.1981), aff'd, 459 U.S. 159, 103 S.Ct. 530, 74 L.Ed.2d 334 (1982); City of Petersburg v. United States, 354 F.Supp. 1021, 1027 (D.D.C.1972), summarily aff'd, 410 U.S. 962, 93 S.Ct. 1441, 35 L.Ed.2d 698 (1973); see also Reno v. Bossier Parish Sch. Bd. ("Bossier Parish II"), 528 U.S. 320, 328, 120 S.Ct. 866, 145 L.Ed.2d 845 (2000), superseded by statute on other grounds, 42 U.S.C. § 1973c(c).

We consider each of Florida's proposed voting changes to determine whether the State has met that burden. Because any evidence relating to the retrogressive effects (or lack thereof) of the changes may inform our analysis of possible discriminatory purpose, see Vill. of Arlington Heights v. Metro. Housing Dev. Co., 429 U.S. 252, 266, 97 S.Ct. 555, 50 L.Ed.2d 450 (1977), we address the effect issue first. See City of Port Arthur, 517 F.Supp. at 1011.

II. Effects

We begin by setting forth the legal standards applicable to the effect prong of section 5. Thereafter, we will apply those standards to Florida's two proposed voting changes.

A. Legal Standards

The test for assessing whether a given voting change will have "the effect of denying or abridging the right to vote on account of race[,] color," or membership in a language minority group, 42 U.S.C. § 1973c(a), has been stated in a relatively straightforward manner. In Beer, the Supreme Court held that, to be entitled to preclearance under the effect prong of section 5, a proposed change must not "lead to a retrogression in the position of racial minorities with respect to their effective exercise of the electoral franchise." 425 U.S. at 141, 96 S.Ct. 1357; accord Bossier Parish I, 520 U.S. at 478, 117 S.Ct. 1491. In other words, a voting change "has a prohibited `effect' only if it is retrogressive," meaning that it "worsen[s] the position of minority voters" in comparison to the preexisting voting standard, practice, or procedure. Bossier Parish II, 528 U.S. at 324, 120 S.Ct. 866.

By definition, this "so-called `retrogression' analysis" requires us to "compar[e] the existing voting scheme" — often known as the "benchmark" practice — with the "scheme that would result from the proposed change." State of New York v. United States, 874 F.Supp. 394, 397 (D.D.C.1994). "If the position of minority voters is no worse under the new scheme than it was under the old scheme," then the proposed change does not have an impermissibly retrogressive effect. Id. (citing City of Lockhart v. United States, 460 U.S. 125, 132-35, 103 S.Ct. 998, 74 L.Ed.2d 863 (1983)). If, on the other hand, the change would lead to a retrogression in the position of minority voters in the covered jurisdiction, then preclearance must be denied. Beer, 425 U.S. at 141, 96 S.Ct. 1357.

Lurking in the background of this seemingly simple formulation, however, are a number of potential complications. The central problem for our purposes is that "[t]he Supreme Court has never comprehensively defined `retrogression,' nor has it engaged in any detailed discussion of what constitutes the `effective exercise of the electoral franchise' by minority voters." Georgia v. Ashcroft, 195 F.Supp.2d 25, 74 (2002), vacated on other grounds, 539 U.S. 461, 123 S.Ct. 2498, 156 L.Ed.2d 428 (2003). In particular, the Court has not specifically addressed how the retrogression test applies to "ballot access" laws (e.g., laws governing the procedures for voting and voter registration) such as the ones before us. Indeed, the case law interpreting the section 5 effect test deals primarily with so-called "second generation barriers ... to prevent[ing] minority voters from fully participating in the electoral process," H.R.Rep. No. 109-478, at 2 (2006), 2006 U.S.C.C.A.N. 618, 618: for example, changes involving annexations, redistricting, or the creation or expansion of at-large electoral systems.7 Despite — or perhaps because of — the Voting Rights Act's central concern with prohibiting practices and procedures that impede minority voters from casting a ballot, see, e.g., 42 U.S.C. § 1973b(a) (suspending all "test[s] or device[s]" in covered jurisdictions), there are no cases squarely addressing how the retrogression analysis should function in a ballot access case like this one.

Nonetheless, our examination of the statutory text and relevant lines of authority allows us to draw some conclusions about applying the section 5 effect test to this case. In brief, we conclude that a change that alters the procedures or circumstances governing voting and voter registration will result in retrogression if: (1) the individuals who will be affected by the change are disproportionately likely to be members of a protected minority group; and (2) the change imposes a burden material enough that it will likely cause some reasonable minority voters not to exercise the franchise. We emphasize that the two-part nature of this inquiry means that the retrogression test in ballot access cases is not solely one of "disparate impact," as Florida fears. See Fla. Br. 57. In other words, a change is not retrogressive simply because it deals with a method of voting or registration that minorities use more frequently, or even because it renders that method marginally more difficult or burdensome. Rather, to be retrogressive, a ballot access change must be sufficiently burdensome that it will likely cause some reasonable minority voters not to register to vote, not to go to the polls, or not to be able to cast an effective ballot once they get to the polls.

This inquiry is a "fact-intensive" one, and requires us to "carefully scrutinize the context in which the proposed voting changes will occur." Georgia v. Ashcroft, 195 F.Supp.2d at 76. In so doing, we do not focus solely on the burdens imposed by a voting change, but rather must also take account of any off-setting, or "ameliorative," adjustments. See City of Richmond, 422 U.S. at 370-71, 95 S.Ct. 2296; City of Petersburg, 354 F.Supp. at 1031. And in the context of this particular case, where only five of Florida's counties are subject to the requirements of section 5, we must look at the effects of the voting changes on minority voters in only those five covered counties. See Lopez, 525 U.S. at 284, 119 S.Ct. 693 ("Section 5, as we interpret it today, burdens state law only to the extent that that law affects voting in jurisdictions properly designated for coverage." (emphasis added)). Finally, as we have said, Florida bears the burden of proving that the voting changes at issue are non-retrogressive. See, e.g., City of Rome, 446 U.S. at 184 n. 18, 100 S.Ct. 1548; Beer, 425 U.S. at 140-41, 96 S.Ct. 1357.

In settling upon the retrogression standard that we have outlined above, we reject Florida's novel arguments concerning section 5 — arguments that question the traditional understanding of the section's effect test or that would read the test out of the statute altogether. We discuss those arguments in the following paragraphs.

1. Florida's most far-reaching contention is that the effect prong of section 5 does not apply to this case at all. This, the State says, is because the 2006 amendments to the Voting Rights Act made the Act inapplicable to changes in ballot access laws. As relevant here, the 2006 amendments added two sections to the Act. See Pub. L. No. 109-246, 120 Stat. 577 (2006). Section 5(b) clarified that a voting change "denies or abridges the right to vote within the meaning of section 5(a) if it "will have the effect of diminishing the ability of any citizens of the United States on account of race or color [or membership in a language minority group] ... to elect their preferred candidates of choice." 42 U.S.C. § 1973c(b). And section 5(d) further explained that the purpose of section 5(b) "is to protect the ability of [minority] citizens to elect their preferred candidates of choice." Id. § 1973c(d).

Florida maintains that these amendments made "ability to elect" the only relevant consideration under the section 5 effect test. And it further argues that, when Congress used the phrase "ability to elect," it meant the ability of minorities as a group to elect candidates of the group's choice. In Florida's view, only voting changes that dilute a minority group's vote — like redistricting or annexation — can normally impact a minority group's ability to elect, whereas ballot access measures — like those at issue here — normally cannot. Accordingly, Florida concludes, section 5's effect test can play no role in this case. See Fla. Br. 49, 51-53.

Contrary to Florida's view, however, the 2006 amendments did not inoculate covered jurisdictions from section 5 review when making changes in ballot access rules. Election law changes dealing with ballot access have long been regarded as the kinds of changes that are at the core of the section 5 preclearance requirement. See Nw. Austin Mun. Utility Dist. No. One v. Holder, 557 U.S. 193, 198, 129 S.Ct. 2504, 174 L.Ed.2d 140 (2009) ("We have interpreted the requirements of § 5 to apply not only to the ballot-access rights guaranteed by § 4, but to drawing district lines as well." (emphasis added)).8 Indeed, the Voting Rights Act itself defines "voting" to include "all action necessary to make a vote effective ..., including, but not limited to, registration, ... casting a ballot, and having such ballot counted properly." 42 U.S.C. § 1973l(c)(1).9 The voting changes at issue in this opinion — which impose additional restrictions on early voting and polling place procedures for inter-county movers — fall squarely within that expansive definition of voting rights. Moreover, although most of section 5 case law involves challenges to redistricting, annexation, and other potential forms of vote dilution, it is plain that laws that make it difficult for minority voters to register to vote or cast a ballot can just as readily — if not more readily — lead to a "retrogression in the position of racial minorities with respect to their effective exercise of the electoral franchise." Beer, 425 U.S. at 141, 96 S.Ct. 1357.

The 2006 amendments to the Voting Rights Act did nothing to alter that basic framework. As relevant here, those amendments simply clarify that retrogressive effects include any diminution in minority voters' ability to elect their candidates of choice. The language added to section 5(b), for example, provides:

42 U.S.C. § 1973c(b) (emphasis added). That language merely supplements rather than supplants the broad requirement imposed on covered jurisdictions — in the original and still extant section 5(a) — to demonstrate that their voting changes will not "have the effect of denying or abridging the right to vote on account of race or color" or membership in a language minority group. Id. § 1973c(a). Likewise, the language added to section 5(d) states: "The purpose of subsection (b) of this section is to protect the ability of [minority] citizens to elect their preferred candidates of choice." Id. § 1973c(d). But that provision refers back only to subsection (b); it does not affect or construe subsection 5(a), nor does it purport to make "ability to elect" the only relevant test for section 5 purposes.10

Furthermore, an examination of the legislative history of the 2006 reauthorization confirms that the purpose behind adding the "ability to elect" language in 42 U.S.C. § 1973c(b) and (d) was simply to legislatively overrule a particular interpretation of the effect prong offered by the Supreme Court in one case: Georgia v. Ashcroft, 539 U.S. 461, 123 S.Ct. 2498, 156 L.Ed.2d 428 (2003). In that case, which involved allegations of vote dilution in the redistricting context, the Court adopted a multi-factor "totality of the circumstances" test, emphasizing that retrogression for vote dilution purposes should be evaluated by reference to "all the relevant circumstances, such as the ability of minority voters to elect their candidate of choice, the extent of the minority group's opportunity to participate in the political process [through `influence districts' and other means], and the feasibility of creating a nonretrogressive plan." Id. at 479, 123 S.Ct. 2498. During the 2006 reauthorization debates, Congress expressed concern that the open-ended Ashcroft framework would lead to "substantial uncertainty" and would hamstring the Attorney General's ability to make expeditious preclearance determinations. See H.R. Rep. No. 109-478, at 68, 2006 U.S.C.C.A.N. 618, 671. Accordingly, Congress added the language of 42 U.S.C. § 1973c(b) and (d) as part of the so-called "Georgia v. Ashcroft fix," in order to return the analysis to the comparatively simpler Beer framework. But nothing in the 2006 legislative record indicates that Congress meant to make the "ability to elect" standard the only relevant criterion under section 5's effect test. And there is certainly nothing to suggest that Congress meant to permit covered jurisdictions to adopt retrogressive restrictions on ballot access without an examination of the effects of those changes under section 5.

The argument that the 2006 Congress exempted ballot access changes from the scope of section 5's effect test is all the more implausible because reading the statute in the manner that Florida suggests would not only contradict the plain language of section 5(a), it would undermine the purpose of the Voting Rights Act. After all, section 5 was passed primarily because certain states and other covered jurisdictions were using "tests and devices," such as poll taxes, literacy tests, and other voting qualifications, to restrict minority voters' access to the ballot. See Katzenbach, 383 U.S. at 310-15, 86 S.Ct. 803. Yet under Florida's proposed interpretation, such tests and devices would escape scrutiny under the effect prong of section 5 precisely because they deal with ballot access measures, rather than "second-generation" discriminatory measures such as vote dilution. We do not believe — nor do we find any evidence — that this was the purpose of the Congress that extended the Voting Rights Act in 2006.

In any event, even if Florida's premise were correct and the "ability to elect" standard were now the sole criterion by which retrogressive effects may be judged, the State's conclusion would still not follow. It is axiomatic that in order for voters to be able to elect the candidates of their choice, they must first be able to register to vote, make it to a polling place, and cast a ballot that will count. Accordingly, election law changes that retrogressively burden such activities may well result in a diminution in minority voters' "ability ... to elect their preferred candidates of choice." 42 U.S.C. § 1973c(b), (d).

2. Next, Florida seeks to narrow the scope of the section 5 inquiry by pointing to the language in section 5 requiring covered jurisdictions to demonstrate that their voting changes will not "have the effect of denying or abridging the right to vote on account of race or color, or in contravention of the guarantees set forth in section 1973b(f)(2) of this title." 42 U.S.C. § 1973c(a) (emphasis added). In Florida's view, the "on account of" language "requires a `but for' causal connection" between race and retrogression. Fla. Br. 55. In other words, Florida contends that under section 5, "the race, color, or language minority status of the affected persons must play a decisive role in producing the prohibited effect." Id. Florida cites no case that has ever interpreted section 5 in this narrow manner. Moreover, although the State proffers two slightly different constructions of its proposed "but for" test, each would render the section's effect prong toothless.

Under Florida's first construction, "a voting change violates the `effect' prong of the statute only if it both has a racially retrogressive effect and the covered jurisdiction made the voting change because it would have that racially retrogressive effect." Fla. Br. 56 (emphasis in original). This reading of the statute would render the effect prong entirely redundant of the purpose prong, thereby contradicting decades of controlling precedent. See, e.g., Lopez, 525 U.S. at 283, 119 S.Ct. 693 (noting that section 5 "guard[s] against both discriminatory animus and the potentially harmful effect of neutral laws" (emphasis added)).11 Such redundancy would also violate a basic canon of statutory construction; section 5, after all, has both a purpose and an effect prong, see 42 U.S.C. § 1973c(a), and each must be given independent force, see Mackey v. Lanier Collection Agency & Serv., Inc., 486 U.S. 825, 837 & n. 11, 108 S.Ct. 2182, 100 L.Ed.2d 836 (1988) (collecting cases stating that courts should hesitate "to adopt an interpretation of a congressional enactment which renders superfluous another portion of that same law").

Florida's second proposed construction is that "a change will violate the effect prong if it has a retrogressive effect on minority voters and the voters' race, color, or language minority status is the reason for the retrogression." Fla. Br. 56 (emphasis in original). Under that reading, "[a] voting change could not fail the `effect' prong merely based on an incidental correlation." Id. In other words, "if the retrogression is `because of' some other factor (e.g., socioeconomic status ...)," which just happens to be correlated with race, "it is not `because of' race, color, or language minority status and preclearance cannot be denied under the `effect' prong." Id. This argument, too, would read the effect test out of the statute. It is difficult to imagine an example of a voting law or practice that would meet such a stringent definition of race-based causation — unless, of course, the law were something so facially discriminatory as "African-Americans may not vote." The Voting Rights Act is plainly not limited to addressing such obvious forms of discrimination. Indeed, the purpose of the Act was to eliminate "the subtle, as well as the obvious, state regulations which have the effect of denying citizens their right to vote because of their race." Allen, 393 U.S. at 565-66, 89 S.Ct. 817. And the prototypical examples of voting laws that are disallowed by the Act — laws like literacy tests and poll taxes — were not based directly on race, but rather on the correlation of race with low literacy rates, poor educational opportunities, and poverty. Florida's proposed interpretation of the effect test would thus mean that section 5 could not prevent the adoption of modern-day equivalents of Jim Crow-era voting laws. To state that proposition is to refute it.

3. Third, Florida contends that the effects of its voting changes must be evaluated "vis-a-vis the 1972 benchmark" — that is, the laws that were "in force or effect" in Florida's covered counties "on November 1, 1972, the effective date of their coverage." Fla. Br. 58.12 This argument need not detain us long because it is foreclosed by numerous binding precedents that identify the relevant section 5 "baseline as the most recent practice that was both precleared and `in force or effect' — or, absent any change since the jurisdiction's coverage date, the practice that was `in force or effect' on that date." Riley v. Kennedy, 553 U.S. 406, 421, 128 S.Ct. 1970, 170 L.Ed.2d 837 (2008) (emphasis added).13 In other words, "[t]o determine whether there have been changes with respect to voting, we must compare the challenged practices with those in existence before they were adopted." Presley v. Etowah Cnty. Comm'n, 502 U.S. 491, 495, 112 S.Ct. 820, 117 L.Ed.2d 51 (1992). If the covered jurisdiction's voting practices have been unchanged since the date on which the jurisdiction became subject to section 5, then the practices in existence on the date of coverage serve as the benchmark. See id. ("Absent relevant intervening changes, the Act requires us to use practices in existence on [the date of coverage] as our standard of comparison." (emphasis added)). But once the jurisdiction amends its voting laws, the laws that existed on its coverage date are no longer "directly relevant, for differences once precleared normally need not be cleared again." Young v. Fordice, 520 U.S. 273, 281, 117 S.Ct. 1228, 137 L.Ed.2d 448 (1997). Instead, the new laws "become part of the baseline standard for purposes of determining" compliance with section 5. Id.

Florida seeks to cast the language in Riley (and other opinions) that defines the relevant benchmark as "the most recent practice that was both precleared and `in force or effect,'" 553 U.S. at 421, 128 S.Ct. 1970, as merely an unchallenged assumption. See Fla. Br. 58-59. But whether "unchallenged" or not, it has been consistently applied and has formed the basis of Supreme Court section 5 jurisprudence for decades. As an inferior court, we are not empowered to question such decisive language. See Overby v. Nat'l Ass'n of Letter Carriers, 595 F.3d 1290, 1295 (D.C.Cir. 2010) (noting that "carefully considered language of the Supreme Court, even if technically dictum, generally must be treated as authoritative[,] ... especially [where] the Supreme Court has reiterated the same teaching" (internal citations and quotation marks omitted)).14

4. Fourth, even if we adopt the retrogression test outlined in Beer and the traditional benchmark outlined in Riley, Florida asks us to make a number of narrowing constructions for purposes of applying the retrogression test to this case. We also reject these novel arguments because they are inconsistent with the central goal of the Voting Rights Act: to provide robust and meaningful protections for minority voting rights.

a. Florida's first such argument contends that the only ballot access changes that violate the effect test are those that make it impossible for minority citizens to vote. See Fla. Br. 61. We cannot agree, however, that the mere existence of other possible avenues for voting renders Florida's changes to its early voting and inter-county mover procedures immune from section 5 scrutiny. If that were the case, a state could close polling places in minority precincts and yet survive the effect test so long as voters still had the option to travel across town to more distant polls. No court has endorsed such a restrictive construction of the section 5 effect prong. See Bossier Parish Sch. Bd. v. Reno, 7 F.Supp.2d 29, 31 (D.D.C.1998) (noting that the section 5 effect test is "broad enough to identify `retrogression' ... if (to imagine an example not present in this case) polling places were located so that they are less convenient to black voters than before the change."), aff'd, 528 U.S. 320, 120 S.Ct. 866, 145 L.Ed.2d 845 (2000). Nor do we. Although the availability of other options may be relevant to the retrogression analysis, it is not dispositive.

b. We also decline Florida's invitation to import a requirement that a change constitute a "severe burden" or impose "excessively burdensome requirements" on the right to vote before it can be considered retrogressive. Fla. Br. 60 (citing Crawford v. Marion Cnty. Election Bd., 553 U.S. 181, 202, 128 S.Ct. 1610, 170 L.Ed.2d 574 (2008); Burdick v. Takushi, 504 U.S. 428, 433-34, 112 S.Ct. 2059, 119 L.Ed.2d 245 (1992); Storer v. Brown, 415 U.S. 724, 728-29, 94 S.Ct. 1274, 39 L.Ed.2d 714 (1974)). That language is drawn from election-law cases decided in the First and Fourteenth Amendment context, as part of a determination of whether a given voting law should be subject to strict scrutiny even if no suspect classification is involved. As such, it does not control here. Instead, the question for section 5 purposes is simply whether the burdens are sufficiently material to result in retrogression with respect to minority voters' exercise of their voting rights.

c. We also reject Florida's assertion that voting changes can have "de minimis," and therefore permissible, effects if they only burden a small number of voters. Cf. Fla. Br. 62-66. If even a small number of voters are sufficiently burdened by a voting change that they do not exercise the franchise when they otherwise would have done so, then that change can (under some circumstances) be considered retrogressive.15 Voting is a fundamental right, "preservative of other basic civil and political rights," Reynolds v. Sims, 377 U.S. 533, 562, 84 S.Ct. 1362, 12 L.Ed.2d 506 (1964), and no amount of voter disenfranchisement can be regarded as "de minimis."

d. Further, as we explain in more detail in the course of our discussion of specific voting changes, see infra Part II.B.1, we disagree with Florida's position that a change is not retrogressive if it affects "in the aggregate significantly more [w]hites than minorities." Fla. Br. 53 (emphasis added). Focusing on the effects of voting changes in absolute terms would mean that almost no ballot access change would be considered retrogressive; after all, the fact that fewer members of a particular group are present in the overall electorate is part of what it means to be a minority group. Applying the effect test in the manner Florida suggests would thus allow covered jurisdictions to enact changes with clearly adverse effects on minority voters so long as more white voters were also affected. That approach would fly in the face of the Voting Rights Act's primary goal of protecting minority voting populations. The retrogression assessment must therefore be conducted in relative terms, with reference to the proportions of each group affected by the change.

e. Finally, we reject Florida's contention that the effect prong of the statute must be read in accordance with the State's proposed interpretations because there are "constitutional concerns" with any contrary reading. See, e.g., Fla. Br. 57.16 We do not discern constitutional difficulties with the interpretation we have set forth above. Contrary to Florida's contention, "[t]he net effect" of this construction of the section 5 effect test is not "to require states and political subdivisions to take account of race in every decision they make in order to ensure that any change would have an ameliorative impact on minority groups." Id. at 57-58. Nothing in the effect test as we have construed it for application to ballot access cases requires covered jurisdictions to maximize voting opportunities for their minority citizens alone. See id. Rather, under section 5, covered jurisdictions must simply ensure that their ballot access changes do not have retrogressive effects on minority voting rights — that is, that they do not "worsen the position of minority voters" in comparison to the pre-existing voting standards, practices, or procedures. Bossier Parish II, 528 U.S. at 324, 120 S.Ct. 866; see City of Lockhart, 460 U.S. at 134 & n. 10, 103 S.Ct. 998.

Nor does this interpretation mean, as Florida suggests, see Fla. Br. 58 n.12, that voting changes that benefit white voters automatically result in retrogression for minority voters. Indeed, the United States and the intervenors explicitly disclaim that position. DOJ Br. 68 n.17. Rather, the question is whether these voting changes will have adverse effects on minority voters' "effective exercise of the electoral franchise" in Florida's covered counties. Beer, 425 U.S. at 141, 96 S.Ct. 1357.

* * *

We now proceed to apply the section 5 effect test to the two voting changes at issue in this opinion, beginning with the new procedures for early in-person voting. In examining each of the changes, we are guided by the two central inquiries that we have outlined above: (1) Are the individuals who will be affected by the change disproportionately likely to be members of a protected minority group? And (2) Does the change impose a burden material enough that it will likely cause some reasonable minority voters not to exercise the franchise?

B. Early Voting Changes.

As explained above, see supra Part I.C.1, the relevant changes to Florida's early voting procedures involve the number of days, the number of hours, the specific hours, and the weekend times that early voting may be offered in future elections. Under the benchmark law, the potential early voting period lasted a total of 14 days, beginning "on the 15th day before an election and end[ing] on the 2nd day before an election." Fla. Stat. § 101.657(d) (2010). Under that prior law, exactly 8 hours of early voting were mandated on each weekday, and exactly 8 hours in the aggregate each weekend, yielding a total of 96 early voting hours. Id. Early voting sites were required to "open no sooner than 7 a.m. and close no later than 7 p.m. on each applicable day." Id. Within this important constraint, local supervisors were free to select the specific voting hours for each voting day as they saw fit. See id. Historically, local supervisors of elections in the five covered counties have chosen a range of voting hours depending on the type of election and the location of the early voting site. With one exception, all of the permutations used by the covered counties captured at least a part of the early morning or evening hours thought to be most convenient for those early voters working the standard 9 a.m. to 5 p.m. workday.17 Finally, although the prior law permitted local supervisors of elections to choose whether their 8 aggregate weekend early voting hours would fall on a Saturday, a Sunday, or both, supervisors in each of Florida's five covered counties decided to offer early voting only on Saturdays. This meant that early voting was actually offered in each of the covered counties for a total of only 12 days under the benchmark statute.

In short, the benchmark practice against which we must compare the early voting changes was 96 hours over 12 days, none of which was a Sunday. See Riley, 553 U.S. at 421, 128 S.Ct. 1970 (holding that the benchmark is the last precleared law actually in practice in the covered jurisdiction). Each day included 8 hours of early voting, although the specific hours varied. See supra note 17.

Because we have not been presented with a specific voting plan from any of the five covered counties, there is much that we do not know about how the new law will be implemented in the five covered counties. We do know, however, that Florida's new law reduces the total number of days available for early voting. The new early voting period begins "on the 10th day before a [state or federal] election... and end[s] on the 3rd day before the election." Fla. Stat. § 101.657(d) (2011). Thus, under the new statute, early voting lasts from the Saturday two weekends before the election through the next Saturday (three days before Election Day) — for a total of only eight days. That is four fewer days than under the benchmark law.

Moreover, under the new law, election supervisors in each county now have the discretion to determine the number of daily hours of early voting in their counties, subject to a statutory requirement that "no less than 6 hours and no more than 12 hours" be offered on each of the 8 early voting days. Id. The end result is that, although Florida's counties may choose the maximum and still offer 96 total hours of early voting (12 hours per day over 8 days), they may also choose the minimum and offer only 48 total hours (6 hours per day over 8 days) — or anything in between. An early voting period lasting only 48 hours would reflect exactly half of the hours that were required under the benchmark practice.

The new law also grants election supervisors the discretion to select the specific hours of early voting for each voting day, unencumbered by the prior law's constraint that early voting sites could not open before 7 a.m. or close after 7 p.m. Compare Fla. Stat. § 101.657(d) (2011) (providing that "the supervisor has the discretion to determine the hours of operation of early voting sites"), with Fla. Stat. § 101.657(d) (2010) ("Early voting sites shall open no sooner than 7 a.m. and close no later than 7 p.m. on each applicable day."). If supervisors of elections decide to offer the maximum of 12 hours each day on a 7 a.m. to 7 p.m. schedule, then the daily voting hours would include both morning and evening hours that fall outside the standard 9 a.m. to 5 p.m. workday. If they offered 12 hours but a different schedule, that would not necessarily be so.18 And if they instead offered only the minimum of 6 hours, there would be many schedules that would not include non-working hours.

Finally, the new statute mandates some additional weekend voting that was not offered under the benchmark practice. In particular, the law requires early voting for at least 6 hours on the Sunday nine days before Election Day, whereas, as we have already explained, there was no Sunday early voting at all under the benchmark practice in the covered counties. And although the new law removes the theoretical possibility under the benchmark statute of early voting on the Sunday immediately before Election Day, compare Fla. Stat. § 101.657(d) (2011), with Fla. Stat. § 101.657(d) (2010), that does not affect our analysis because none of the covered counties offered any Sunday voting at all. See Riley, 553 U.S. at 421, 128 S.Ct. 1970 (holding that the relevant benchmark is the "most recent practice that was both precleared and in force or effect" (emphasis added) (internal citation and quotation marks omitted)); City of Lockhart, 460 U.S. at 132, 103 S.Ct. 998 ("The proper comparison is between the new system and the system actually in effect.").

In sum: With respect to days, the new law results in 4 fewer early voting days than the benchmark (8 days rather than 12). One of those days is now a Sunday, whereas there was no Sunday voting under the benchmark practice. With respect to hours, the new law may result in a total of as few as 48 hours of early voting — one half of the 96 required under the benchmark — because it permits counties to offer as few as 6 hours per day (over the 8 days). But the new law also permits counties to offer as many as 12 hours of voting per day, which, if adopted by the counties, would result in a maximum of 96 hours of early voting — the same number of hours as under the benchmark. With respect to the specific hours of voting, the new law leaves the supervisors unconstrained. Hence, it may result in fewer of the early morning and evening hours that are convenient for voters working the standard 9 a.m. to 5 p.m. workday. Or, it may result in more such hours.

The interrelationship between the days and hours of early voting under the old and new law introduces uncertainty into our evaluation of the effects of these early voting changes. For the reasons discussed below, we conclude that at least one permutation of the early voting changes would result in retrogression. In particular, we conclude that Florida has failed to meet its burden of showing that retrogression would not occur if the covered counties not only reduced the number of early voting days from 12 to 8 as required by the new law, but also reduced their total early voting hours from 96 to 48 (regardless of the specific hours chosen). As also discussed below, however, we conclude that if the counties were instead to utilize the maximum 96 hours permitted under the new law, and offer them on a standard 7 a.m. to 7 p.m. schedule that would provide the opportunity for voters to vote before and after the workday, Florida would likely be able to meet its burden of demonstrating that the overall effect of the changes would not be retrogressive. But neither Florida nor the counties have submitted the counties' intended hours for preclearance, nor do we have any real indication of what the counties will do. Accordingly, because retrogression may well follow from the new statute, we cannot grant preclearance at this time. Our analysis follows.

1. As an initial matter, we find that minority voters will be disproportionately affected by the changes in early voting procedures because they disproportionately use early in-person voting. As the intervenors' expert witness, Professor Paul Gronke,19 determined, the proportion of African-American usage of early in-person voting in Florida "has exceeded White usage of early in-person voting in four of five [recent] federal elections," and "substantially exceeded White usage in both the 2004 and 2008 presidential elections." A10092, 10104 (Am. Expert Report of Prof. Paul Gronke).20 The data is particularly striking for the 2008 general election, when more than half (54%) of African-American voters in Florida cast ballots using early in-person voting — twice the rate of white voters. A10106 (Am. Expert Report of Prof. Gronke, Ex. Seven); see A7818 (Expert Report of Dr. Stewart, Attach. J). And although rates of early voting declined across the board in 2010, the African-American usage rate still exceeded the white rate by a factor of about one-third in the 2010 general election. See A 10106 (Am. Expert Report of Prof. Gronke, Ex. Seven).21

These disproportionate usage rates hold true in Florida's five covered counties as well. In 2008, for example, 52% of all African-American voters in the covered counties cast an early in-person ballot, compared to only 28% of white voters. Id. at A10095-96, 10105; see also A7821 (Expert Report of Dr. Stewart, Attach. M). African-American rates of early in-person voting in the five covered counties also "remained statistically significantly higher [than white rates] in the 2010 primary and general elections." A10096 (Am. Expert Report of Prof. Gronke). Even Florida's own expert witness, Professor M.V. (Trey) Hood III, acknowledged that he "can see the pattern" of higher levels of African-American usage of early in-person voting in the covered counties after the 2008 primary election. A5855 (Hood Dep.); see also A9041-42 (Expert Report of Prof. Hood) (showing that the African-American rates of early voting in the covered counties exceeded the rates of white voters to a significant degree in the 2008 general election, the 2010 primary, and the 2010 general election (and, to a lesser extent, in the 2008 primary)).

Furthermore, all available evidence suggests that these trends "will continue into the 2012 general [election] and likely in[to] the future." See A9918 (De Bene Esse Dep. of Prof. Gronke). According to Professor Gronke:

A10096 (Am. Expert Report of Prof. Gronke); see A10079 (De Bene Esse Dep. of Prof. Gronke) ("I think that history will show that 2008 ha[d] a particularly high rate [of African-American early voting], but that that adoption rate by African-Americans had a lasting impact, and that the higher rate of usage will continue."); id. at A10034-36. Florida's expert likewise agreed that it is "more likely than not" that in the 2012 general election, the African-American usage rate of early in-person voting will be higher than the white usage rate. A5875-76 (Hood Dep.).22

Moreover, although the differences are not as stark as for the entire early voting period, the evidence also shows that African-American voters disproportionately used the first five days of the benchmark early voting period — i.e., the Monday through Friday of the week that falls two weeks before Election Day — all of which will now be eliminated under HB 1355. In the 2008 general presidential election, for example, approximately 17.25% of African-American voters in the covered counties cast an early in-person ballot during the first five days of early voting (the so-called "repealed days" of early voting), compared to only 9.3% of white voters. A10097 (Am. Expert Report of Prof. Gronke); see A7821 (Expert Report of Dr. Stewart, Attach. M). In other words, African-American voters used the repealed days of early voting at rates nearly double those of white voters in 2008. The difference between these percentages "far exceeds the... statistical significance criterion." A 10097 (Am. Expert Report of Prof. Gronke). And while the data with respect to use of the repealed days in the 2010 elections is more inconclusive,23 for reasons discussed below we find the 2008 presidential election to be the best indicator of future trends.

In finding that African-American voters in the covered counties will be disproportionately affected by the reduction in early voting days under the new law, we reject the contrary opinions of Florida's expert witness, Professor Hood.24 We do so because the analysis underlying his conclusions suffers from a number of methodological flaws.25

First, Professor Hood asserts that the effects of the early voting changes will be "disproportionately borne by [white] voters" because they "comprise the greatest share of total number of early votes cast" in the covered counties. A9061 (Expert Report of Prof. Hood). But as the intervenors' expert, Professor Gronke, convincingly explains, this "analytic method violates a basic tenet of comparative analysis" because it fails to "control for ... the size of the underlying subgroups." A9089 (Rebuttal Decl. of Prof. Gronke). After all, it is "no surprise that [white] voters are ... the majority of users of early in-person voting" because they also represent "the vast majority of all Florida voters," id. (emphasis added), including those in the covered counties, see A7683-84 (Expert Report and Decl. of Russell Weaver) (collecting census data in the covered counties).26 That is why, as Professor Gronke explains, adjusting for population demographics is critical in assessing the potential retrogressive effects of a given voting change:

A9888-90 (De Bene Esse Dep. of Prof. Gronke); see A9089 n.1 (Rebuttal Decl. of Prof. Gronke) (explaining how Prof. Hood's flawed analysis could lead to endorsing a poll tax that has an obviously disparate effect on minority voters). As even Professor Hood conceded, the accepted practice in the social sciences is to look at the rate of impact on different groups, not simply absolute numbers. A5926-28 (Hood Dep.). Yet his calculations do not follow that accepted method of statistical analysis. See id. at 5840-42.

Moreover, as we suggested above, adopting Professor Hood's approach would permit covered jurisdictions to enact changes with clearly adverse effects on minority voters as long as more white voters were also affected. See supra Part II.A.4.d.27 Since having fewer voters is part of what it means to be a "minority," that would render section 5 ineffectual in virtually all ballot access cases — an approach that cannot be squared with Congress' intent to use the section to protect minority voting populations. Accordingly, as we concluded above, the retrogression assessment must be conducted in relative terms, with reference to the proportions of each group affected by the change.

Second, we reject other calculations in Professor Hood's expert report because we agree with the intervenors' expert that "[i]n several instances Professor Hood inappropriately pools together groups of dissimilar data, which is not methodologically appropriate." A9092 (Rebuttal Decl. of Prof. Gronke). For example, Professor Hood "attempts to draw conclusions based on data `pooled' from different kinds of elections, without offering a reason to believe that early voting patterns are in fact common across the different types of elections." Id. And "[t]here is no evidence that Professor Hood conducted a pooling test, a statistical tool that helps determine whether it is valid to aggregate data" from those different types of elections. Id.; see A5762-64 (Hood Dep.) (conceding that he pooled the data without running any of the standard statistical tests to determine whether such pooling was appropriate). This problem is exacerbated, for reasons discussed in more detail below, by the fact that Professor Hood often "aggregat[es] data from all elections analyzed in his Report except for the 2008 general election," thereby further distorting the data. A9092 (Rebuttal Decl. of Prof. Gronke).

Professor Hood also frequently lumps African-Americans and Hispanics into a single category of "Minorities," which misleadingly flattens the data because, unlike African-Americans, Hispanic voters use early voting at about the same rate as whites. See A7818 (Expert Report of Dr. Stewart, Attach. J); A10105 (Am. Expert Report of Prof. Gronke, Ex. Six). This "masks significant differences between White and African-American rates of early in-person voting." A9092 (Rebuttal Decl. of Prof. Gronke). As Professor Gronke notes, "[i]t is inconsistent with political science research standards to pool together these two racial groups when they demonstrate different behaviors in voting." Id. at A9093. Moreover, it is at odds with the relevant inquiry under section 5 because the section forbids a covered jurisdiction from diminishing voting rights "on account of race or color, or" membership in a language minority group. 42 U.S.C. § 1973c(a) (emphasis added).

Finally, we reject Professor Hood's contention that the 2008 general election was an "outlier" that should be ignored. More specifically, Professor Hood asserts that the "anomalous" circumstances surrounding the 2008 election — namely, "the historic candidacy of Barack Obama, the first African-American presidential candidate representing one of the two major political parties in the United States," coupled with the "intensity of interest in President Obama's candidacy" among African-American voters — account for the "spike in early in-person voting turnout among African-Americans" in 2008. A9042 (Expert Report of Prof. Hood). According to Professor Hood, if "the 2008 general election is removed from the analysis, the numbers more accurately reflect the historic trends." Id. at A9043. And once the 2008 general election is discarded, the rates of white and minority usage of early in-person voting then appear — at least to Professor Hood's eye — "roughly comparable," meaning (in his view) that the early voting changes should not have disproportionate effects on African-American voters. Id. at A9041-42; see also id. at A9061 (describing the early voting turnout rates by race as "very similar"); Fla. Br. 65 ("Putting aside an anomalous 2008 general election,... [w]hite voters are slightly more likely to use early voting than their minority counterparts." (emphasis added)).

We find this reasoning unpersuasive for several reasons. First, and most important, we can hardly discard an election as an "outlier" or an "anomaly" simply because an African-American ran for President for the first time as a major-party candidate and many African-Americans voted for him. The Voting Rights Act was designed to ensure that, among other things, minorities can "elect their preferred candidates of choice." 42 U.S.C. § 1973c(d). Thus, we 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.

In addition, although the 2008 election was of course unique in certain respects, the record does not support the assertion that that election was an "outlier" that can be discarded out of hand. Rather, the record evidence suggests that the 2008 election is highly predictive of what is likely to happen in 2012. The expert witnesses in this case are all generally in agreement that, when assessing future usage rates of early voting, comparisons are best made between "like" elections, and that the most recent analogous election is the best predictor of what will happen in the future. Accordingly, the 2008 general presidential election represents the best guidepost for projecting how early in-person voting will be used in the upcoming 2012 general presidential election. See, e.g., A5657-58 (Stewart Dep.); A9091 (Rebuttal Decl. of Prof. Gronke); A10098 (Am. Expert Report of Prof. Gronke); see also A5867-68 (Hood Dep.) (conceding that "the 2008 presidential election would be our best gauge" of "what's going to occur in the 2012 general election"). For this reason alone, "it is not by any means an `outlier' in terms of predictive value." A9091 (Rebuttal Decl. of Prof. Gronke). And that is particularly true because many of the factors that supposedly rendered the 2008 election "anomalous" will again be factors in 2012:

Id.

Moreover, even setting aside the specifics of the 2008 and 2012 general elections, the record indicates that the 2008 general election was not a mere one-off phenomenon. The trend of increased African-American usage of early in-person voting predated 2008 to some degree: African-American early voting rates in Florida exceeded white early voting rates in 2004, as well. See A10092, 10104 (Am. Expert Report of Prof. Gronke). And after the 2008 election, the rates of African-American usage of early in-person voting in Florida have continued to exceed those of white voters to a statistically significant degree. See, e.g., A9090 (Rebuttal Decl. of Prof. Gronke). Hence, it is not true, in the words of Florida's expert, that once the 2008 election is removed from the picture the rates of white and African-American early voting are "roughly comparable." A9041 (Expert Report of Prof. Hood). And in any event, "[t]he term `roughly comparable' is not an accepted term of art or standard for statistical comparisons in the field of political science research." A9090 (Rebuttal Decl. of Prof. Gronke). Rather, "[t]he methodologically appropriate starting point is to determine whether the differences are statistically significant." Id.28 And, even discounting the 2008 election, the data shows a clear and statistically significant "trend of higher usage [of early voting] over time by African American voters in the Covered Counties and statewide." Id.29

In short, rather than being an "outlier," the evidence suggests that the 2008 general election is best seen as a "game-changer" that simply magnified already nascent trends in African-American preferences for early in-person voting. See A10069 (De Bene Esse Dep. of Prof. Gronke).30 Indeed, Florida's own expert witness acknowledges that the 2008 election was not an outlier when compared to the relative rates of African-American versus white early in-person voting in the four most recent federal elections in Florida. A5965-66 (Hood Dep.). As such, there is no sound basis for disregarding the 2008 election in our evaluation of the effects of the new statute's changes in early voting procedures.

2. Having determined that African-American voters in Florida's covered counties are disproportionately likely to use the early voting days that would be eliminated under the new law, we proceed to evaluate whether Florida's early voting changes will impose a materially increased burden on African-American voters' effective exercise of the electoral franchise relative to the benchmark practice. As we have already suggested, this evaluation is complicated by the fact that the new law makes certain potential trade-offs between the days and hours of early voting: it decreases the available days, but allows local election officials to maintain the same total early voting hours and to enhance the accessibility of those hours, if they so choose. And it requires a voting day on a Sunday. That said, we do not yet know how the covered counties will decide to exercise their discretion with respect to the early voting hours offered in their jurisdictions: under the new law, election officials could choose to offer as many as 96 or as few as 48 hours of early voting, meaning that the total hours might remain the same as under the benchmark practice, or might be cut in half (or might end up somewhere in between).31

Florida urges us to assume that "the Covered Counties may seek to provide the maximum early voting hours available under the Act (96 hours)." Fla. Br. 25. But because Florida bears the burden of proving that its new law will be nonretrogressive, the State must show — not simply assume — that given a menu of possible hours, its covered counties will choose nonretrogressive ones. Florida has submitted no evidence, however, that would permit us to predict with any confidence the number of early voting hours the covered counties will offer. No covered county's election supervisor has submitted a county plan for preclearance. Most of the supervisors' responses to questioning by the defendants on this subject may at best be described as equivocal.32 And at least one covered-county supervisor indicated that she would not use the full 96 hours of early voting, unless required to do so. See A3530-32 (Dep. of Lucretia Strickland, Hendry County SOE). Moreover, Florida acknowledges that it "has not issued any rule, directive, or guidance to Florida supervisors of elections ... regarding the manner in which [they] should exercise the discretion granted by HB 1355" to determine the number of early voting hours in their counties. A8283 (Fla. Resp. to Req. for Admiss. No. 18).

In light of these considerations, and in light of the fact that the burden is on Florida to prove that its new law will not have a retrogressive effect, we must proceed on the assumption that the counties will utilize the minimum rather than the maximum permitted hours.33 If the effects would be nonretrogressive even on that assumption, then we could conclude that Florida has met its burden. But if they would be retrogressive at the minimum number of hours, then the uncertainty as to whether more hours will be offered (and whether those hours will be more or less accessible) means that Florida has not met its burden.

We conclude that Florida has failed to sustain its burden of proving that, if the covered counties offer only 48 hours of early voting (i.e., only 6 hours per day) as the new law permits, that change would not impose a material burden on — and therefore a retrogressive effect with respect to — African-American voters' effective exercise of the electoral franchise. Under such circumstances, not only would the number of early voting days be reduced by one-third (from 12 to 8), but the total available early voting hours would be cut in half (from 96 to 48). Moreover, with only 6 hours available per day, it is likely that early voting would start after the workday starts and would end before the workday ends, making it even more inaccessible to many minority voters who have inflexible work schedules. See A9142-43 (Decl. of Cynthia Slater, 2d Vice-President of Florida NAACP). This dramatic reduction in a form of voting disproportionately used by African-Americans would be analogous to (although certainly not the same as) closing polling places in disproportionately African-American precincts. Although such action would not bar African-Americans from voting, it would impose a sufficiently material burden to cause some reasonable minority voters not to vote.

This conclusion is supported by testimony indicating that a two-week early voting period is important to get-out-the-vote (GOTV) efforts in minority communities. See, e.g., A9143 (Slater Decl.); A9227-28 (Decl. of Rev. Charles McKenzie, Florida state liaison for the Rainbow PUSH Coalition).34 Record evidence suggests that such efforts are important in enabling African-Americans "who want to vote but need help getting to the polls" to exercise the franchise. A9237 (Decl. of Ella Kate Coffee, African-American resident and GOTV volunteer in Hillsborough County); see A7693 (Expert Report and Decl. of Russell Weaver, Ex. 2) (noting that disproportionately more minority households in the covered counties have no motor vehicle available). With a substantially reduced early voting period, third-party groups would not be able to assist minority voters as effectively. See A9237 (Coffee Decl.). This, in turn, would likely make it more difficult for those minority voters who rely on such efforts to make it to the polls. Florida has not submitted any evidence to the contrary. Indeed, although Florida's expert witness initially hypothesized that early voters should be able to adjust to even such a dramatic contraction in the early voting period, A9059-60 (Expert Report of Prof. Hood), he later conceded that there is no empirical support for that claim, and that it was nothing more than an "assumption" or "supposition" on his part, A5861, 5917-22 (Hood Dep.); see A9095 (Rebuttal Decl. of Prof. Gronke) ("I know of no empirical support for the conclusion that voters will successfully adjust... under these conditions.").35

Moreover, even if all of the voters who would have used the repealed days of early voting did attempt to adjust to a shortened early voting schedule of only 48 hours over 8 days, that shift would create problems of its own for minority voting. In the 2008 general election, for example, 71,670 voters — 14,897 of whom were African-American — cast ballots in the covered counties during the early voting days that the new law has repealed. See A7820-21 (Expert Report of Dr. Stewart, Attachs. L & M); see also A9054-57 (Expert Report of Prof. Hood) (showing that approximately one-third to two-fifths of early voters voted in the first (repealed) week of early voting). According to testimony in the record, a shift of that magnitude to the remaining early voting days would lead to substantially increased lines, overcrowding, and confusion at the polls, which would in turn discourage some reasonable minority voters from waiting to cast their ballots. See, e.g., A3170-71, 3183-87 (Dep. of Harry Sawyer, Monroe County SOE); A3744-45, 3748 (Dep. of Earl Lennard, Hillsborough County SOE).36 Indeed, election officials in Florida have testified that an extensive early voting period is necessary because Florida's "electoral infrastructure is completely maxed out," A440 (House State Affairs Comm. (Apr. 1, 2011) (Statement of Ion Sancho, Leon County SOE)), such that the State "would not be able to process record numbers of voters" in a substantially shorter time frame, A976-78 (Senate Rules Comm. (Apr. 15, 2011) (Sancho Statement)).37 Florida legislators, too, have warned that a "shortened period of early voting will cause congestion and long lines in populous areas of the state, including predominantly African-American neighborhoods in Hillsborough County," and will thereby "discourage [minority] voters from voting." A9108 (Decl. of Sen. Arthenia Joyner).

Florida counters that there is relatively little definitive evidence regarding how early voting actually affects overall voter turnout. This observation is accurate, given the fact that early voting itself is a relatively recent phenomenon. But the problem for Florida is that, under the Voting Rights Act, the State bears the burden of proving that its action will not have a retrogressive effect; the defendants do not have to prove that it will have such an effect. See Bossier Parish I, 520 U.S. at 480, 117 S.Ct. 1491 ("Section 5 ... imposes upon a covered jurisdiction the difficult burden of proving the absence of discriminatory purpose and effect."). This is particularly so when it has already been shown — as it has been here — that a voting change does, in fact, disproportionately harm African-American voters. Hence, without some evidence indicating that minority voters would be able to adapt to such a substantial reduction in early voting hours, the State cannot justify a grant of preclearance. Florida has submitted no such evidence, whether empirical, anecdotal, or otherwise. Indeed, the most that Florida is willing to assert is that the "vast majority" — but apparently not all — of those African-American voters who would have voted during the repealed days "will still vote." Fla. Br. 25.

Meanwhile, the academic scholarship and commentary is currently in a state of flux as to how the availability of early in-person voting affects overall voter turnout. The consensus prior to the 2008 election appears to have been that early in-person voting was a convenience that had an "insignificant or marginal effect on increasing the likelihood [that] an individual will vote." A10131 (Robert Stein & Greg Vonnahme, Early, Absentee, and Mail-in Voting, in THE OXFORD HANDBOOK OF AM. ELECTIONS & POL. BEHAVIOR (Jan Leighly, ed., 2010), at 185); see A5496 (Stewart Dep.) ("[T]he research up until around 2008 ... is that, in general, early voting procedures make voting more convenient for people who look like they have a ... propensity to vote, but do[] not generally increase turnout overall."); see also A9093-95 (Rebuttal Decl. of Prof. Gronke).38

According to the intervenors' expert, however, "[t]he 2008 presidential and subsequent elections have challenged the conventional wisdom [regarding early voting], primarily because of changing voting patterns in the South." A10091 (Am. Expert Report of Prof. Gronke). And the United States' expert further states that "the issue that [post-2008] research raises is whether the use of early voting in 2008 may ... have been related to a surge in turnout in 2008. So there may be something new afoot." A5497 (Stewart Dep.). Moreover, even if early voting by itself does not affect overall voter turnout, there is evidence that it may do so when combined with other factors, such as GOTV drives. Indeed, "[t]here is a growing literature that takes a more nuanced approach to examining the relationship between convenience voting and turnout, conjecturing that although reforms may not increase overall participation, they may ... have an impact ... when combined with party mobilization efforts." A7926 (Paul Gronke et al., Convenience Voting, 11 ANN. REV. POL. SCI. 437, 443 (2008)); see A8009 (Paul Gronke, Early Voting Reforms & Am. Elections, WM. & MARY BILL OF RTS. J. 423, 432 (2008)) (stating that "[e]nough research has accumulated" for scholars to conclude that early voting "does encourage regular voters to participate in lower intensity contests that they might otherwise skip"); A9094 (Rebuttal Decl. of Prof. Gronke) (stating that, "[s]ince non-habitual voters are less likely to vote, early or on election day, convenience may have a significant and positive effect on their decision to vote" as well). In short, at the very least it can be said that "there is not currently a consensus ... that there is no effect of early in-person voting on turnout." A9984 (De Bene Esse Dep. of Prof. Gronke).

Moreover, even if the (still developing) academic scholarship could be read in Florida's favor, it would still be insufficient to carry Florida's burden of proof. The literature that Florida cites addresses only the question of how adding early voting days affects overall voter turnout. It does not address the specific question before us: how decreasing an established early voting period from 12 days to 8 days (and from 96 hours to 48) will affect African-American voter turnout. Hence, even if the addition of early voting days does not significantly increase turnout, "it is not methodologically sound to assume that there will ... be little or no impact on overall turnout when voters (who have habituated to early in-person voting) face a loss of previously available voting days." A9095 (Rebuttal Decl. of Prof. Gronke). Indeed, common sense suggests the opposite.39

Finally, although we acknowledge that Florida's new early voting statute makes some ameliorative adjustments in weekend early voting even at the minimum number of hours, we conclude that those changes would not outweigh the negative effects on GOTV drives and lines at the polls that we have discussed above. As discussed above, the new statute mandates a Sunday of early voting that was not previously offered in any of the covered counties. Thus, even if the covered counties choose to offer the minimum 6 hours per day of early voting, they would still be required to offer 6 hours of voting on the Sunday nine days before Election Day. And given the importance of Sunday early voting for "souls to the polls" drives in the African-American community, see infra Part II. B.3, that adjustment can be expected to have some positive effects on the ability of African-Americans to reach the polls.

Nonetheless, those 6 hours of Sunday early voting would represent the only ameliorative aspect of the new law if the covered counties were to implement it by choosing the bare minimum number of hours. Moreover, they would be accompanied by a net decrease of 2 hours on each Saturday as compared to the benchmark.40 Overall, we are not persuaded that merely shifting one of the remaining 8 days to a Sunday would make up for a 4-day and 48-hour decrease in early voting.

In sum, Florida is left with nothing to rebut either the testimony of the defendants' witnesses or the common-sense judgment that a dramatic reduction in the form of voting that is disproportionately used by African-Americans would make it materially more difficult for some minority voters to cast a ballot than under the benchmark law. We therefore conclude that Florida has not met its burden of proving that removing a third of the days and half of the hours from its benchmark early voting procedures would have a nonretrogressive effect on minority voting rights in the covered counties.

3. Although we have concluded that we cannot preclear Florida's early voting changes at this time because those changes authorize the covered counties to offer a statutory minimum number of hours that may result in retrogression, it is possible that the counties will instead, as Florida predicts, opt to provide substantially more hours than that minimum. As we discuss below, under at least one such scenario we are persuaded that Florida would likely satisfy its burden of showing a nonretrogressive effect: that is, if the covered counties were to provide the maximum authorized 96 hours on a standard 7 a.m. to 7 p.m. schedule. When a court finds that it cannot preclear one iteration of a submitted plan, but may be able to preclear a modified version, the Supreme Court has expressed approval for issuing a kind of "conditional order" indicating the circumstances under which approval may be obtained. See, e.g., City of Port Arthur, 459 U.S. at 167-68, 103 S.Ct. 530 (expressing approval of the district court's "conditional order" denying preclearance of the expansion of a city's borders unless the city agreed to eliminate a majority-vote requirement for certain elections); City of Richmond, 422 U.S. at 370, 95 S.Ct. 2296 (stating that the district court in City of Petersburg, 354 F.Supp. 1021, "was correct in conditioning approval of the annexation upon the adoption of the plan to elect councilmen by wards"). We do so here.

As we have noted above, although HB 1355 mandates a reduction in early voting days from 12 to 8 days, it authorizes the counties to offer a maximum of 12 hours on each of those early voting days — as compared to no more than 8 hours per day under the benchmark. (The counties are required to keep the polls open for 12 hours on Election Day itself. Fla. Stat. § 100.011(1) (2011).) Under those circumstances, voters would have exactly the same total number of hours for early voting as under the benchmark law: 96 hours. Those hours would simply be distributed over a fewer number of days.41

We do not discount the possibility that the reduction in days — considered alone — would make it more difficult for some voters to get to the polls. But not all reductions in early voting days have retrogressive effects, as the Justice Department has made clear by preclearing reductions in prior — albeit distinguishable — cases. See A9002-03 (preclearing Florida's 2005 shortening of the early voting period from 15 to 14 days); [128] United States' Notice in Resp. to Section 3(A) of the Court's June 22, 2012 Order (noting preclearances involving, e.g., Texas (changing the start date for the in-person early voting period from 20 days to 17 days prior to an election); Georgia (reducing the early voting period from 45 days to 21 days)).

More important, we cannot consider the reduction in days alone: We must consider not only the effect of the "negative" aspects of the changes in law, but also of its ameliorative aspects. See City of Richmond, 422 U.S. at 370-71, 95 S.Ct. 2296; DOJ Br. 68 ("The Department and courts consider whether a proposed change's negative impact on minority voters has been offset by ameliorative steps taken by the jurisdiction."). First among those, of course, is the tradeoff of hours for days. Although voters would not be able to vote during 4 days that were previously available, they would be able to vote during 4 hours each day that were not. Moreover, if election supervisors utilize a standard 7 a.m. to 7 p.m. voting day, this means that weekday voters would not have to vote during working hours, but could instead get to the polls before or after work — during the morning and evening hours that the record suggests would be most convenient for weekday voters and that would be more extended than those offered in any recent election. (Accordingly, when we conclude that there is a scenario that is likely to be nonretrogressive, that scenario includes a 7 a.m. to 7 p.m. schedule.) As the supervisor of elections in Hillsborough County stated, "some hours are more accessible to people than other hours," and both "[t]he early morning hours earlier in the day" and hours "in the afternoon ... following work" can be helpful for some voters who "have the opportunity" to vote at those times. A3765 (Lennard Dep.); see A9095 (Rebuttal Decl. of Prof. Gronke) ("Certainly more hours of voting per day can benefit voters, and there are studies... showing the positive benefits of expanded hours outside of the normal business day."); A10511 (testimony from an election official in a non-covered county that, "when the polls open [at] 7 a.m.," there will generally "be lines that have already formed" with voters who are "waiting for the precinct to open").42

Nor is the increase in weekday early voting hours the only ameliorative factor we must take into account. As we have discussed, the new early voting statute requires the covered counties to offer early voting for at least six hours on the Sunday nine days before Election Day. See Fla. Stat. § 101.657(d) (2011). Although the benchmark statute permitted counties to offer early voting on that Sunday, as well as on the Sunday immediately before Election Day, see Fla. Stat. § 101.657(d) (2010), none of the covered counties ever did so. Hence, the benchmark practice was no Sunday voting. See Riley, 553 U.S. at 421, 128 S.Ct. 1970; City of Lockhart, 460 U.S. at 132, 103 S.Ct. 998. Although the intervenors initially took a contrary position, they acknowledged in a post-argument filing that "[t]he change eliminating the discretion to conduct `last Sunday early voting' is not at issue in the retrogression analysis since the Covered Counties have not conducted early voting on that day." [134] Defendant-Intervenors' Supplemental Br. Regarding Preclearance Standards for Enabling Legislation at 2 n.1 (emphasis added).43 Accordingly, the new statute provides at least a net 6-hour increase in Sunday early voting compared to the relevant benchmark. And if the covered counties offer the maximum number of hours of early voting each day, as we assume for purposes of this subpart, the result would be an additional 12 hours of Sunday early voting that were never before offered in those counties.

The addition of Sunday voting is important — as the United States and intervenors themselves emphasized in initially arguing that the relevant benchmark should include the unrealized potential for last-Sunday voting under the prior statute. As the defendants stressed, many African-American churches organize "souls to the polls" drives to transport their congregants to early voting sites on that Sunday, see A9109 (Joyner Decl.); A9193-94 (Decl. of Rev. Thomas Scott), and that Sunday is therefore disproportionately used by African-American voters in jurisdictions that have early voting on that day, see A10098 (Am. Expert Report of Prof. Gronke); see also DOJ Br. 51; A9225 (Decl. of Rev. Charles McKenzie). But while the defendants have argued that it is the Sunday immediately before Election Day that is particularly important, there is evidence in the record to suggest that opening the polls on the previous Sunday would also facilitate voting by African-American voters. For example, even Senator Joyner, who opposed the reduction in early voting days, testified that if HB 1355 "were precleared and were put into effect in Hillsborough County," she would "[a]bsolutely... expect the predominantly African-American churches in Hillsborough County to organize Souls to the Polls drives" on that day. A9814 (Joyner Dep.). Likewise, Professor Gronke agreed that he has "no reason to believe that [the additional Sunday] would not be used" by African-American voters in such "souls to the polls" drives. A10012-13 (De Bene Esse Dep. of Prof. Gronke). As a result, we can expect that there will be at least some ameliorative effects from the new law's addition of the mandatory Sunday.

We further note that, for a county that offers the maximum 12 hours per day, the new law not only adds the opportunity to vote for 12 hours on the Sunday that falls two weekends before the election, but also increases the overall number of weekend hours on both that weekend and the last weekend before the election. Under the benchmark practice, the covered counties offered early voting for 8 hours on each Saturday, and not at all on Sunday, yielding 16 total weekend hours of early voting. But if a county offers the maximum hours under the new statute, it will provide early voting for 12 hours on each weekend day (i.e., on both Saturday and Sunday) approximately one week before the election, and for 12 hours on the last Saturday before the election. The total would then be 36 weekend hours of early voting (12 hours on each of 3 weekend days), for a net gain of 20 more weekend hours of early voting than under the benchmark.44

Like the added Sunday, those additional weekend hours of early voting are significant. Florida election officials have testified, based on their experience, that expanded weekend hours would provide increased accessibility for many minority voters. See, e.g., A3447-48, 3450 (Edwards Dep.); A3611-12, 3770, 3846-47 (Lennard Dep.); A10488 (Dep. of Ion Sancho, Leon County SOE). Florida legislators, including several of the legislators who opposed the early voting changes and who are now intervenors in this action, have concurred in that assessment. See A9801 (Joyner Dep.); A10254-55 (Cruz Dep.). And even the intervenors' expert, Professor Gronke, concedes that expanded early voting hours can be an added convenience for many voters. See A6161 (Gronke Dep.); A9095 (Rebuttal Decl. of Prof. Gronke); A10015 (De Bene Esse Dep. of Prof. Gronke); see also A10187 (EVIC blog post by Professor Gronke, stating that "weekend early voting" is "a potential inconvenience for officials to be sure, but one which citizens will find very helpful").

In sum, the record evidence persuades us that, if the covered counties offer the maximum available early voting hours each day on a standard 7 a.m. to 7 p.m. schedule, the negative effect of reducing the number of days from 12 to 8 would likely be offset by the ameliorative effects of adding non-working weekday hours, a Sunday, and additional weekend hours.45

Moreover, many of the concerns that we discussed in Part II.B.2 in connection with a contraction of the early voting period to only 48 total hours are not likely to materialize (certainly not to the same extent) if the full 96 hours of early voting are maintained on a standard 7 a.m. to 7 p.m. schedule. For example, although representatives and volunteers from minority voting rights groups have testified that "a two-week period has ... been essential to coordinating the logistics of GOTV efforts in the African American community," A9237 (Coffee Decl.); see A9227 (McKenzie Decl.), the record evidence suggests that GOTV groups could adjust to a redistribution of the total 96 hours over a different number of days, including weekend days and a "souls-to-the-polls" Sunday. Indeed, one of the central concerns expressed by such groups is that minority voters "frequently contend with issues such as lack of transportation or inflexible work and family schedules, which make it difficult to reach the polls on Election Day." A9143 (Slater Decl.). Expanding convenient non-working weekday and weekend voting hours should therefore help third-party efforts to provide transportation to the polls for such voters.

The same is true of concerns regarding overcrowding and confusion at polling places on the remaining early voting days (and Election Day) that could result from reducing the total number of early voting days. This problem should be alleviated, if not entirely eliminated, by the simultaneous expansion of early voting hours to 12 hours per day on a 7 a.m. to 7 p.m. schedule. Indeed, election officials have testified that they believe that, "by expanding the hours," they would be able to "accommodate [the] early voters" who may have been displaced by the elimination of the first five days of the old early voting period. A3859 (Lennard Dep.).46 For the first time, early voters in the five covered counties would have the benefit of an early voting period capturing the morning and evening hours on both ends of the standard 9 a.m. to 5 p.m. workday.

For the foregoing reasons, we conclude that if Florida and the covered counties were to submit a preclearance plan that offered early voting for 12 hours per day, from 7 a.m. to 7 p.m. over an 8-day early voting period, including one previously-unavailable Sunday, they would likely satisfy their burden of proving that the overall effect of the early voting changes would be nonretrogressive with respect to minority voters.47 We do not, however, reach any conclusions about other permutations of early voting hours that the covered counties might offer. Because at this point we do not know how many early voting hours the covered counties will offer or which specific hours they will provide, and because we find that retrogression may well follow from the statute as written, we cannot now conclude that Florida has met its burden of demonstrating that its statewide early voting changes will have a nonretrogressive effect in the covered counties.

C. Inter-County Mover Changes

We next address the changes in the procedures for so-called "inter-county movers" — registered voters who move from one Florida county to another, fail to notify the supervisor of elections of their address change in advance of the election, and then seek to vote in their new precinct. The relevant HB 1355 amendment provides that, with the exception of active uniformed services voters and members of their families, a registered voter "whose change of address is from outside the county [but inside the state] may not change his or her legal residence at the polling place and vote a regular ballot," Fla. Stat. § 101.045(2)(b) (2011), as he could under previous law, see Fla. Stat. § 101.045(2)(b) (2010). Instead, "such elector is entitled to vote a provisional ballot" after completing a "Provisional Ballot Voter's Certificate and Affirmation." Fla. Stat. § 101.045(2)(b) (2011). Under Florida law, such ballots "shall be counted unless the canvassing board determines by a preponderance of the evidence that the person was not entitled to vote." Id. § 101.048(2)(a) (emphasis added). As we explain below, Florida represents that this means that in the event that a canvassing board reaches the end of the canvassing period and has not reviewed some subset of the provisional ballots cast, then the unreviewed ballots must be counted as eligible ballots because the canvassing board would not have "determine[d] by a preponderance of the evidence that the person[s] w[ere] not entitled to vote." Id.

1. We again begin by finding that the inter-county mover changes will disproportionately affect minority voters, because minority voters have been disproportionately more likely to use the benchmark procedures that allowed voters to make inter-county address changes at the polls and then cast a regular ballot. See A8901 (Supplemental Decl. of Dr. Stewart) ("[M]inority voters are more likely to avail themselves of the law that is still in effect in the five covered counties, which allows registered voters who have moved between counties to simultaneously change their address and vote a regular ballot in their new home county."). In the most recent elections in Florida's covered counties for which there is data, the voters who moved inter-county and updated their addresses on or about Election Day were disproportionately members of a minority group. Id. at A8907-09 (Attach. 5). More specifically, African-American and Hispanic voters were statistically more likely than the average voter to be inter-county movers in the 2008 and 2010 elections, both on Election Day and during the early voting period, while white voters were statistically less likely to be inter-county movers. Id.48

For example, in the 2008 general election, African-American voters comprised only 13.3% of the total non-moving voters in Florida's covered counties, but made up 19.3% of all inter-county movers in those counties. Id. at A8908. Hispanic voters comprised only 10.1% of the non-movers, but accounted for 14.5% of the inter-county movers. Id. Meanwhile, white voters made up 71.3% of the non-movers, but only 58.8% of the inter-county movers. Id. These and other data points allowed the United States's expert witness, Dr. Charles Stewart, to conclude that "[i]n every single calculation [in his supplemental report], the percentage of black and Hispanic voters among the `movers' is greater than the percentage of black and Hispanic voters among the non-moving voters who used that mode [i.e., early voting or Election Day voting]." Id. at A8899. By contrast, "[i]n every calculation except one [the 2010 primary], the percentage of white voters among the movers is less than the percentage of white voters among the non-moving voters who used that mode." Id.; see also A10667 (De Bene Esse Dep. of Dr. Stewart) ("It's clear to me that blacks, whites, and Hispanics avail themselves of these procedures at different rates than the underlying voting population.").49

Indeed, even Florida's expert witness, Professor Hood, recognized that in the elections that he examined, the minority rates of inter-county movers in the covered counties were proportionately higher than the white rates. Professor Hood's data set is less complete than Dr. Stewart's, because he examined data from only three of Florida's five covered counties — excluding Hillsborough County, the most populous of the five — and two of those counties claimed not to have any inter-county movers in the election years that he analyzed. See A9073 (Expert Report of Prof. Hood); see also A5954 (Hood Dep.) (admitting that his data set is only "[a] quarter complete"). Nonetheless, Professor Hood conceded that in each of the elections he examined, the inter-county mover rate of "some minority group [whether African-American or Hispanic] is higher than the white rate and in most of [the elections] all of the minority rates are higher than the white rate." A5958-59 (Hood Dep.) (emphasis added).50

The bottom line is that, on average, minority voters in Florida's covered counties are approximately twice as likely as white voters in those same counties to move inter-county and seek to update their addresses at the polls. From 2008 through 2010, 0.44% of African-American voters and 0.50% of Hispanic voters in the covered counties were inter-county movers, compared to only 0.21% of white voters. See Findings of Fact ¶¶ 85-88. Accordingly, if the inter-county mover changes took effect in the covered counties, African-American and Hispanic voters would be disproportionately affected.

Florida correctly points out that, even if minorities are disproportionately affected, the total number of inter-county movers has historically been quite small, and the number who are minorities is even smaller. Indeed, the high-water mark for inter-county movers was the 2008 general election, when 2,240 voters (approximately 750 of whom were minorities) updated their addresses at the polls in the covered counties. See A8908 (Supplemental Decl. of Dr. Stewart). This is out of a total of over 530,400 voters who voted in that election in the covered counties. See id. But while the number of affected voters will have an impact on the burden analysis we conduct in the following section, we note again that a voting change with a retrogressive effect does not warrant preclearance merely because it affects a small number of voters. See supra Part II.A.4.

2. To be retrogressive, however, a ballot access change must not only affect proportionally more minority voters; it must also impose a burden material enough that it will likely cause some reasonable minority voters not to exercise the franchise. Accordingly, we must evaluate the burdens imposed by the inter-county mover changes to determine whether they are likely to dissuade reasonable minority voters in Florida's five covered counties from casting a ballot, or are likely to result in those minority voters' ballots not being properly tabulated. Although evidence has been submitted on both sides of those questions, we are persuaded by a preponderance of the evidence that the inter-county mover changes will not have impermissibly retrogressive effects on minority voters. This conclusion is based on the understanding, however, that consistent with the language of Florida statutory law and the interpretation the State has proffered to the court in this case, in the event that a canvassing board reaches the end of the canvassing period and has not reviewed some subset of the provisional ballots cast, the unreviewed ballots shall be counted as eligible ballots because the canvassing board would not have "determine[d] by a preponderance of the evidence that the person[s] w[ere] not entitled to vote." Fla. Stat. § 101.048(2)(a) (2011).

First, the record establishes that compliance with the new inter-county mover paperwork requirements will not take materially longer than compliance with the prior requirements. Under the benchmark statute, a voter who moved to another county without notifying the relevant election official and then sought to vote in his or her new county of residence was required to complete a change of address affirmation in "substantially the [same] form" set forth in Fla. Stat. § 101.045(2)(a) (2010), which was as follows:

Fla. Stat. § 101.045(2)(a); see, e.g., [129] Fla. Notice of Filing of Documents Re: Inter-County Address Changes at the Polls, Ex. A (Collier County change of address affirmation form).51 After completing the form, the voter was then entitled to cast a regular ballot.

Under the new law, inter-county movers must instead complete a "Provisional Ballot Voter's Certificate and Affirmation," as set forth in Fla. Stat. § 101.048(3) (2011). The new form must be "in substantially the [same] form" as the following:

Fla. Stat. § 101.048(3) (2011); see, e.g., [129] Fla. Notice of Filing of Documents Re: Inter-County Address Changes at the Polls, Ex. D (Collier County Provisional Ballot Certificate and Affirmation).52 After completing the affirmation, the inter-county mover is then entitled to cast a provisional ballot. Fla. Stat. § 101.048(1)-(2) (2011).

As a side-by-side comparison demonstrates, the form required under the new law requests essentially the same information as the prior form, including the voter's name, current and prior address of residence, and an affirmation that the voter is legally registered and qualified to vote.53 See A4068 (Dep. of Maria Matthews, Assistant Gen. Counsel, Fla. Dep't of State) ("[R]eally [inter-county movers are] not filling out ... any more information or any additional documentation than they would have if they had moved in from another county previously."). Although under the prior procedure the inter-county mover would then cast a regular ballot, while now the voter must cast a provisional ballot, the latter is in all physical respects identical to a regular ballot. Fla. Stat. § 97.021(29) (defining "provisional ballot" to simply mean "a conditional ballot"); Fla. Admin. R. 1 S-2.037(2)-(3) (providing guidelines for provisional ballot certificates and affirmations); see Oral Arg. Tr. 237:19-238:24 (representation by counsel for Florida that provisional ballots look the same as regular ballots, and that "[a]ll Florida voters vote by paper ballot"). The only difference — from a procedural point of view (but see further discussion infra) — is that instead of being fed into a machine for tabulation, a provisional ballot is placed into a secrecy envelope, which is then sealed in a provisional ballot envelope. Fla. Stat. § 101.048(1)-(2) (2011). The ballot is then deposited in a ballot box, and the secrecy envelope remains sealed until the county canvassing board reviews the certificate and affirmation to determine whether the ballot may be tabulated. See id.

On the basis of the foregoing, we are persuaded that the paperwork associated with these changes will not be materially more burdensome for inter-county movers. Election officials in the covered counties have estimated that the new provisional voter's certificate and affirmation should not take much longer to complete than the old affirmation of change of address. See, e.g., A3938 (Ussery Dep.) ("It's the same ballot they would have received.... I don't really see where it would be an obstacle [to voting].").54 At most, the total paperwork may take a few additional minutes because the voter may have to include some of the same information on the provisional ballot envelope. See Oral Arg. Tr. 237:11-237:16. The voter may also have to wait while an election official reviews and signs his or her affirmation and/or provisional envelope. See Fla. Stat. § 101.048(3) (2011). But we do not see how these minor burdens could deter a reasonable inter-county mover from voting. Moreover, some of the covered counties have sought to minimize even these burdens by affixing the Provisional Ballot Voter's Certificate and Affirmation to the outside of the provisional envelope, or even printing the certificate and affirmation directly onto the provisional envelope itself. See, e.g., A7395 (Hillsborough County Provisional Ballot Quick Reference Guide).

Second, we must consider whether, although the paperwork is not noticeably different, the overall balloting process for inter-county movers will take materially longer because of the additional interactions they will need to engage in with polling place officials. For example, although there does not appear to be any uniform process for inter-county mover voting under the statute, at least one election official in a non-covered county has indicated that provisional voters will be required to move into a different line, where they must interact with a particular election worker with the authority to process provisional ballots. See, e.g., A10583-84 (Sancho Dep.); A9183 (Sancho Decl.). Moreover, voters faced with the requirement of casting a provisional ballot may have questions for election officials, cf. A9105 (Joyner Decl.), and election officials are also required to communicate certain information to provisional voters, see Fla. Stat. § 101.048(5). All together, it does appear that some elements of the process may be slightly more time-consuming under the new law.

On the other hand, some aspects of the new inter-county mover process may move more quickly than the old procedures. For example, under the benchmark process, election officials were required to verify that inter-county movers were validly registered Florida voters before permitting them to cast a regular ballot. See Fla. Stat. § 101.045(d) (2010). Verification might involve consulting a voter registration database or telephoning election headquarters, with associated delay. Now, by contrast, verification is not necessarily required at the time of voting. Instead, the county canvassing board will determine the inter-county movers' eligibility when reviewing their provisional ballots after Election Day. See id. § 101.048(2)(a) (2011). Overall, then, we do not see much difference in the time required to cast an inter-county mover ballot under the new versus the benchmark law, let alone enough to deter reasonable voters from voting.

Third, the United States and intervenors urge us to consider the possibility that, although the new procedures may not take materially longer, voters will nonetheless be so deterred by the overall "negative experience" of being required to cast a provisional ballot that they will decide not to vote at all. Based on the record before us, however, we cannot conclude that a voter who has already committed to waiting in line to cast a ballot, and then learns that he or she is an "inter-county mover," will leave the precinct rather than vote a provisional ballot. It is certainly understandable that casting a provisional ballot may "not leave a good feel[ing] with the voter," A3778 (Lennard Dep.), and that intercounter movers may be upset or frustrated when informed that they have to vote using a provisional ballot, see A10583-84 (Sancho Dep.). But as the intervenors acknowledged at oral argument, "the retrogression standard is not in place to prevent people from being irritated." Oral Arg. Tr. 194:1-194:3.

In any event, much of the confusion and frustration appears to stem from a misunderstanding of the nature of provisional ballots. Although some voters "believe that provisional ballots will not be counted," A9106 (Joyner Decl.), under Florida law such ballots "shall be counted" as long as the voter is eligible, Fla. Stat. § 101.048(2)(a) (2011). Election officials in the covered counties have testified that they will explain these facts to their inter-county movers, and that under such circumstances, "if the voter is there to vote and they still have the opportunity to vote, most likely they are going to take advantage of it." A3972 (Ussery Dep.). Indeed, contrary to the uncorroborated assertion of a single supervisor of elections in a non-covered Florida county, see A10583-84 (Sancho Dep.), election officials in the covered counties have testified that they have never seen a voter leave a precinct without voting simply because he or she was required to vote a provisional ballot, A3972 (Ussery Dep.); A3778-79 (Lennard Dep.).55 In sum, there are no grounds for finding that the provisional ballot requirement for inter-county movers will result in reasonable minority voters no longer exercising the franchise.

Fourth, the United States and intervenors argue that the inter-county mover changes will "increase the lines at the polling places," which may in turn discourage or prevent other voters (i.e., those who are not inter-county movers) from voting. A7279 (Edwards E-mail (Apr. 13, 2011)); see also A3358 (Edwards Dep.). The theory appears to be that, even if the inter-county movers themselves will not be dissuaded from voting, the added time required to process them will result in longer lines and thereby discourage other voters from waiting in line to cast a ballot. We certainly do not disagree that there can be circumstances in which long lines at the polls may dissuade voters from voting. There is no evidence in the record of this case, however, that such dissuasion would result from the inter-county mover changes.

As we have just explained, the record indicates that the new procedures for inter-county movers will not take materially more time than the prior procedures. More important, whatever increased time an individual inter-county mover takes, there are so few of them that it is highly unlikely that any individual precinct would be affected to any material degree.56 As we have said, the high-water mark for inter-county movers was the 2008 general election, when 2,240 voters (approximately 750 of whom were minorities) updated their addresses at the polls in the covered counties. See A8908 (Supplemental Decl. of Dr. Stewart).57 (This is out of a total of over 530,400 voters who voted in that election in the covered counties. See id.) Moreover, of those 2,240, only about 1,000 updated their addresses on Election Day; the rest were spread over the early voting period. A8907-08. And when that total is itself spread among all of the precincts in Florida's five covered counties, it averages out to only a handful of inter-county movers at each precinct. See Oral Arg. Tr. 34:22-35:02. Indeed, it appears that even these numbers may over-estimate the number of such voters in future elections, for a variety of reasons.58

Fifth, we consider the possibility that the inter-county movers' provisional ballots will not be counted due to human error or lack of time to tabulate them. The United States and intervenors have introduced evidence that in the past, provisional ballots have been counted at lower rates than regular ballots. Of course, none of those provisional ballots involved inter-county movers because the provisional procedure did not apply to them under the prior law. Moreover, the record does not disclose why many of those provisional ballots were invalidated, and the United States and intervenors have proffered no evidence that invalidation was due to human error (or malevolence). Instead, the only evidence in the record is that, most often, provisional ballots were not counted "simply because the person [was] not eligible to vote," A5048 (Dep. of David Stafford, Escambia County SOE and Florida State Association of Supervisors of Elections ("FSASE") President); see also A8156 (Decl. of Gisela Salas, Fla. Director of Elections), a circumstance that would not apply to a true inter-county mover.

Hence, while there is a possibility of human error involved in casting and counting provisional ballots, we cannot deny preclearance based on a speculative risk. Rather, consistent with Florida law, "[a] ballot of a person casting a provisional ballot shall be counted unless the canvassing board determines by a preponderance of the evidence that the person was not entitled to vote." Fla. Stat. § 101.048(2)(a) (2011) (emphasis added); see id. § 101.048(2)(b)(1). Florida has already instructed its supervisors of elections accordingly. See A2764-65 (Fla. Dep't of State, Directive 2011-01 (May 19, 2011)) ("[T]he provisional ballot shall count unless the canvassing board determines more likely than not that the person was not entitled to vote. That would occur only if the voter was not registered or the voter voted in a precinct other than the one that corresponds to his or her new address ... or if evidence was available before the board that either the voter had already voted or that the voter was committing fraud."); A7232-33 (Fla. Dep't of State, Directive 2012-01 — Provisional Ballot Verification (Jan. 13, 2012)) (same).

We next address the possibility that the increase in the number of provisional ballots due to the inter-county mover changes will make it impossible to count them in a timely fashion. During the legislative debates, the FSASE expressed concern that "there's simply not enough time to canvass" the additional provisional ballots that will be cast under the new inter-county mover procedures. A967-69 (Senate Rules Comm. (Apr. 15, 2011) (Statement of David Stafford, FSASE President)); see also A9180-81 (Decl. of Ion Sancho, a supervisor of elections in a non-covered county). But supervisors of elections in the covered counties have testified that, consistent with their responsibilities under Florida law, they will "train and put on additional personnel to ensure that [the new provisional ballot procedures] will work." A3781-82 (Lennard Dep.). Indeed, the FSASE President — who opposed the inter-county mover changes on the ground that they would result in too many additional provisional ballots — has stated that, if more provisional ballots are cast than usual, his county's canvassing board will bring in additional personnel and work longer hours to ensure that the additional ballots are properly examined. A5044-45 (Stafford Dep.). Moreover, as we have noted above, the numbers of inter-county movers in the covered counties are so small as to make it even more unlikely that their provisional ballots will overwhelm their county canvassing boards.

In any event, because provisional ballots are presumptively valid under the new statute, which places the burden of proof on the canvassing board to disqualify them, see Fla. Stat. § 101.048(2)(a), Florida has confirmed that its election officials must count the provisional ballots of inter-county movers even if the officials run out of time to examine them in detail. See Fla. Br. 19; A7100-01 (Fla. Rule 30(b)(6) Dep.); A4467-68 (Dep. of Gary Holland, Assistant Gen. Counsel, Fla. Dep't of State). Accordingly, our grant of preclearance to the inter-county mover changes is based on our express understanding that Florida will follow its laws as written, see Fla. Stat. § 101.048(2), and will abide by the representations it has made to this court, see Fla. Br. 19; Oral Arg. Tr. 33:05-34:20 (argument by Florida's counsel that "these provisional ballots are going to be counted," even if the canvassing board does not have time to fully review them, and stating that "the Court can accept that representation and issue a decision accordingly"). We also expect that Florida will educate its supervisors of elections about their responsibilities, to avoid any confusion or misunderstandings that might otherwise result.59 And, of course, our decision in this section 5 case does not bar subsequent actions for violations of federal or state law.60

Finally, we note that a significant factor in our decision to grant preclearance to the inter-county mover changes is that, while in some respects the new law makes it marginally more difficult for such voters, in other respects the law makes it considerably easier for them. Cf. City of Richmond, 422 U.S. at 370-71, 95 S.Ct. 2296 (discussing not only the burdens imposed by a given voting change, but also any off-setting, or ameliorative, adjustments). Specifically, the new statute makes it easier for Florida voters to update their addresses before going to the polls. Under the benchmark law, inter-county address changes had to be completed "using a voter registration application signed by the elector." Fla. Stat. § 97.1031(2) (2010). Now, however, voters can update their addresses simply by "[c]ontacting the supervisor of elections via telephone or electronic means," or by submitting any kind of "signed, written notice." Fla. Stat. § 97.1031(1)(b) (2011). Because inter-county address changes are now so much easier to effectuate, there may well be fewer voters who wait until they arrive at the polling place to change their addresses. This will further reduce concerns regarding longer lines, extended wait times, and the added burdens that provisional ballots may impose on voters and election officials.

For the foregoing reasons, we are persuaded by a preponderance of the record evidence that the inter-county mover changes will not have a retrogressive effect on minority voters in Florida's five covered counties.

III. Purpose

Section 5 of the Voting Rights Act requires Florida to demonstrate that each of its proposed changes "neither has the purpose nor will have the effect of denying or abridging the right to vote on account of race or color, or in contravention of the guarantees set forth in section 1973b(f)(2) of this title [proscribing voting restrictions based on membership in a language minority group]." 42 U.S.C. § 1973c(a) (emphasis added). See Bossier Parish I, 520 U.S. at 478, 117 S.Ct. 1491; City of Rome, 446 U.S. at 172, 183 & n. 18, 100 S.Ct. 1548; see also City of Richmond, 422 U.S. at 378, 95 S.Ct. 2296 ("An official action ... taken for the purpose of discriminating ... on account of ... race has no legitimacy at all under our Constitution or under the [Voting Rights Act]."). In Part II, we addressed the effects of Florida's voting changes. In this part, we address their purpose.

As with our analysis of the effect prong, we first determine the appropriate legal standards for evaluating purpose. We then apply those standards to the specific voting changes at issue in this case.

A. Legal Standards

As Congress made clear in the 2006 reauthorization of the Voting Rights Act, a voting change passed with "any discriminatory purpose" does not qualify for preclearance under section 5. 42 U.S.C. § 1973c(c). The Supreme Court has directed that, in evaluating legislative purpose under that section, "courts should look to [the] decision in Arlington Heights for guidance." Bossier Parish I, 520 U.S. at 488, 117 S.Ct. 1491. There, the Court "set forth a framework for analyzing `whether invidious discriminatory purpose was a motivating factor' in a government body's decisionmaking." Id. (quoting Arlington Heights, 429 U.S. at 266, 97 S.Ct. 555). That "framework has ... been used, at least in part, to evaluate purpose [in] previous § 5 cases." Id. (citing City of Pleasant Grove v. United States, 479 U.S. 462, 469-70, 107 S.Ct. 794, 93 L.Ed.2d 866 (1987) (considering several of the Arlington Heights factors in the course of rejecting a proposed annexation of an African-American neighborhood); Busbee v. Smith, 549 F.Supp. 494, 516-17 (D.D.C. 1982) (referencing the Arlington Heights test); City of Port Arthur, 517 F.Supp. at 1019 (same)).

Before applying the Arlington Heights test to this case, we must first flesh out its framework.61 We must also address several of Florida's novel arguments for evaluating purpose under section 5, some of which would take us far outside the bounds of Arlington Heights.

1. Our examination of the text of section 5 and relevant case law construing the purpose prong of the Act convinces us that the basic approach we must apply in this case is as follows: First, Florida must present some prima facie evidence "to show that [its] voting changes are nondiscriminatory." Shelby Cnty. v. Holder, 811 F.Supp.2d 424, 431 (D.D.C.2011), aff'd, 679 F.3d 848 (D.C.Cir.2012). "As a practical matter, this means that the plaintiff must come forward with evidence of legitimate, nondiscriminatory motives for the proposed changes to [its] voting laws." New York, 874 F.Supp. at 400. If Florida can meet that initial burden of production, the production burden then "shifts to the Attorney General," id., to provide some evidence to "refute the covered jurisdiction's prima facie showing that a proposed voting change does not have a [discriminatory] purpose." Bossier Parish II, 528 U.S. at 332, 120 S.Ct. 866; see New York, 874 F.Supp. at 400. When each party has met its production burden, those burdens fall away and it remains for us to assess whether Florida (through its prima facie case and other evidence) has satisfied its ultimate burden to prove, by a preponderance of the evidence, that its election changes were not motivated by any discriminatory purpose.62

As noted above, in making that assessment we are to "look to" the Court's "decision in Arlington Heights for guidance." Bossier Parish I, 520 U.S. at 488, 117 S.Ct. 1491. "The `important starting point' for assessing discriminatory intent under Arlington Heights is `the impact of the official action [and] whether it bears more heavily on one race than another.'" Id. at 489, 117 S.Ct. 1491 (quoting Arlington Heights, 429 U.S. at 266, 97 S.Ct. 555) (additional internal quotation marks omitted). "Other considerations relevant to the purpose inquiry include, among other things, `the historical background of the [jurisdiction's] decision'; `[t]he specific sequence of events leading up to the challenged decision'; `[d]epartures from the normal procedural sequence'; and `[t]he legislative or administrative history, especially... [any] contemporary statements by members of the decisionmaking body.'" Id. (quoting Arlington Heights, 429 U.S. at 267-68, 97 S.Ct. 555) (brackets in Bossier Parish I.) Our inquiry may also include an examination of such other "circumstantial and direct evidence of intent as may be available," Arlington Heights, 429 U.S. at 266, 97 S.Ct. 555, including whether there are legitimate, race-neutral justifications for the change, see, e.g., City of Richmond, 422 U.S. at 374, 95 S.Ct. 2296.

2. In adopting this construction of the section 5 purpose prong, we reject several of Florida's contrary arguments.

First, Florida contends that, because it is a non-covered state and only five of its counties are covered jurisdictions, its sole burden under the purpose prong should be to "show[] ... that the covered count[ies] acted with no impermissible purpose." Fla. Br. 36. Florida then goes on to argue that, because "[t]here is no evidence that legislators from the Covered Counties were motivated by any discriminatory purpose" in passing HB 1355, the election law changes at issue here are entitled to preclearance. Id. at 41. Indeed, Florida argues that, even if the legislators from Florida's five covered counties had acted with a discriminatory purpose in voting for the changes, that "would be irrelevant as the Act was passed by such wide margins that the votes of the legislators from the Covered Counties had no effect on the Act becoming law." Id.

We cannot square this interpretation of the purpose test with the text of the statute or relevant precedent. Section 5 states that preclearance must be denied if a proposed voting change "has the purpose... of denying or abridging the right to vote on account of race[,] color," or membership in a language minority group. 42 U.S.C. § 1973c(a). And in Lopez v. Monterey County, 525 U.S. 266, 119 S.Ct. 693, 142 L.Ed.2d 728 (1999), the Supreme Court held that voting changes that are to be administered in a covered jurisdiction must be submitted for preclearance even if the changes are part of a state law of general applicability and the State itself is not covered by section 5. See also 28 C.F.R. § 51.23 (giving partially covered states the authority to submit voting changes on behalf of their covered jurisdictions). The Lopez Court reasoned that, while section 5 generally does not impose burdens on non-covered jurisdictions, it does "burden[] state law [in non-covered states] ... to the extent that that law affects voting in jurisdictions properly designated for coverage." 525 U.S. at 284, 119 S.Ct. 693.

In so holding, Lopez did not suggest that the only purpose that should be considered in evaluating whether such a state law is entitled to preclearance is the purpose that motivated the specific legislators from the covered jurisdictions.63 Nor does Florida cite any authority for that proposition. That is not surprising. The Supreme Court has done the opposite when applying section 5's purpose prong to partially covered jurisdictions. See Shaw v. Hunt, 517 U.S. 899, 912-13, 116 S.Ct. 1894, 135 L.Ed.2d 207 (1996) (concluding that the North Carolina "General Assembly" adequately "explained why it did not create a second minority district," id. at 912, 116 S.Ct. 1894).64 So has this court. See New York, 874 F.Supp. at 400 (preclearing New York's plan because "the legislature's primary motivation" was nondiscriminatory). Indeed, limiting the purpose inquiry to the legislators from covered jurisdictions would render the purpose prong virtually meaningless for states like Florida, where so few counties are covered that the votes of legislators from the covered jurisdictions are unlikely to be outcome-determinative.

Next, Florida cites Personnel Administrator of Massachusetts v. Feeney, 442 U.S. 256, 99 S.Ct. 2282, 60 L.Ed.2d 870 (1979), as "set[ting] forth the constitutional standard" for unlawful purposeful discrimination. Fla. Br. 38. We are not certain what standard Florida means to tease out of Feeney. At one point, Florida cites Feeney for the proposition that "the plaintiff must show that `the decisionmaker ... selected or reaffirmed a particular course of action at least in part `because of,' not merely `in spite of' its adverse effects upon an identifiable group.'" Id. (quoting Feeney, 442 U.S. at 279, 99 S.Ct. 2282). Except for the fact that in a section 5 case the burden to show the absence of an unlawful purpose is on the covered jurisdiction, we generally agree with that proposition: there is an impermissible purpose under section 5 when race was a "motivating factor" in a decision, Bossier Parish I, 520 U.S. at 488, 117 S.Ct. 1491 (quoting Arlington Heights, 429 U.S. at 266, 97 S.Ct. 555), not when the legislature merely took a particular action knowing that it might bear disproportionately on minority groups. Of course, the fact that a given action will have a disparate impact on minorities (and that the decisionmaker knew that) can provide powerful circumstantial evidence of discriminatory intent. As Arlington Heights explained, "[t]he impact of the official action" and "whether it bears more heavily on one race than another may provide an important starting point" in the analysis of legislative intent. 429 U.S. at 266, 97 S.Ct. 555 (internal citation and quotation marks omitted).

But Florida also cites Feeney for the propositions that "facially neutral laws of general applicability are presumptively constitutional," Fla. Br. 38, and that "even if a neutral law has a disproportionately adverse effect upon a racial minority, it is unconstitutional under the Equal Protection Clause only if that impact can be traced to a discriminatory purpose," Feeney, 442 U.S. at 272, 99 S.Ct. 2282. Florida then asserts that as a result of these two propositions, "Florida can meet its burden under Section 5 by showing that each voting change is a neutral law of general applicability and thus presumptively constitutional." Fla. Br. 40. That argument, however, conflates the constitutional standard with the statutory preclearance test. Feeney may set out the relevant constitutional standard, but as a matter of statutory preclearance under section 5, the traditional presumption is reversed, and the covered jurisdiction bears the affirmative burden of showing that its proposed voting changes do not have a discriminatory purpose or a retrogressive effect. See 42 U.S.C. § 1973c(a); see Bossier Parish II, 528 U.S. at 328, 120 S.Ct. 866; Bossier Parish I, 520 U.S. at 478, 117 S.Ct. 1491; City of Rome, 446 U.S. at 184 n. 18, 100 S.Ct. 1548; Beer, 425 U.S. at 140-41, 96 S.Ct. 1357; Georgia, 411 U.S. at 538, 93 S.Ct. 1702. As such, the fact that a voting law is "facially neutral" or "of general applicability" is not dispositive for section 5 purposes. Indeed, a central goal of the Voting Rights Act was to prevent covered jurisdictions from enacting laws that "may have been facially neutral" but that could be "easily manipulated to keep [minorities] from voting." Nw. Austin, 557 U.S. at 198, 129 S.Ct. 2504. And as Florida itself admits, see Fla. Br. 40-41, case law interpreting the Act requires jurisdictions to bring forth at least some "affirmative evidence that the proposed changes were not motivated by a discriminatory purpose," New York, 874 F.Supp. at 400; see City of Richmond, 422 U.S. at 374, 95 S.Ct. 2296.

Finally, Florida puts the preceding two arguments together and contends that we must evaluate only the motives of the legislators from the five covered counties, and must assume that facially neutral voting changes were enacted with a proper purpose, because to do otherwise "would increase further the serious federalism costs already implicated by [section] 5 and thus trigger serious constitutional concerns." Fla. Br. 41 (internal citation and quotation marks omitted); see id. at 36. But the doctrine of constitutional avoidance does not permit us to interpret section 5 in a way that would render its purpose prong meaningless for states with a small number of covered counties, or that would require reversing the burden of proof specified by the statutory text. See Commodity Futures Trading Comm'n v. Schor, 478 U.S. 833, 841, 106 S.Ct. 3245, 92 L.Ed.2d 675 (1986) ("[A]lthough this Court will often strain to construe legislation so as to save it against constitutional attack, it must not and will not carry this to the point of perverting the purpose of a statute or judicially rewriting it." (internal citations and quotation marks omitted)). And in any event, as we explain below, Florida in fact prevails under the purpose prong with respect to the voting changes that we address (the inter-county mover changes), even as that prong has traditionally been interpreted.

* * *

We now proceed to apply the section 5 purpose test, including the burden-shifting framework and the Arlington Heights factors outlined above, to the particular voting changes at issue in this opinion. Consistent with long-standing practice under section 5, we will evaluate each of the proposed changes separately. See DOJ Br. 58 ("The Attorney General analyzes each voting change in submitted legislation separately as to purpose and to effect."); cf. [79] March 8, 2012 Mem. Order at 2-3 (noting that the United States advised the court that it had concluded that Florida had met its burden of demonstrating that the citizen initiative changes in HB 1355 were neither enacted with a discriminatory purpose nor will have a retrogressive effect on minority voters); Georgia v. Ashcroft, 195 F.Supp.2d 25 (denying preclearance to a state senate redistricting plan, but preclearing the Congressional and state house plans adopted at roughly the same time). But because the parties often present more general arguments with respect to the purpose underlying HB 1355 as a whole, we cross-apply our analysis of those broader arguments where relevant.

B. Early Voting Changes

Our discussion of the early voting changes can be brief. As explained above, we have determined that Florida has failed to satisfy its burden of proving that those changes will not have a retrogressive effect. See supra Part II.B.2. And because we cannot preclear changes unless they satisfy both the effect and purpose prongs of section 5, we must deny preclearance regardless of the purpose of the early voting changes. This makes it unnecessary for us to reach the purpose prong at this time. See City of Port Arthur, 517 F.Supp. at 1019 ("[W]e need not reach the issue of whether the changes in voting standards, practices or procedures were undertaken with the purpose of denying or abridging the right to vote on account of race, color, or language-affiliation if we find that these same actions have a discriminatory effect."). Instead, we leave that question for another day, to be resolved if Florida or the covered counties submit nonretrogressive hours for preclearance. Of course, consistent with the important role that the effect of a voting change plays in our analysis of purpose, see Arlington Heights, 429 U.S. at 266, 97 S.Ct. 555, if the covered counties submit hours that will not result in retrogression, that will be a significant factor in favor of a finding of nondiscriminatory purpose.

C. Inter-County Mover Changes

We next consider the changes to the voting procedures governing inter-county movers. With respect to those changes, Florida has satisfied its prima facie burden to "come forward with evidence of legitimate, nondiscriminatory motives" for the amendments. New York, 874 F.Supp. at 400. The procedures at issue address a particular — and, by the numbers, unusual — scenario in which a voter arrives at the polls to cast his or her vote but is already registered to vote in a different county. Florida has submitted numerous contemporaneous statements by the proponents of the inter-county mover changes who stated that, in light of those circumstances, their goal was to ensure the efficiency and integrity of the electoral process by allowing election officials to verify the eligibility of such inter-county movers.

In particular, legislators stated that, under the prior law, it was often difficult if not impossible to determine whether an inter-county mover seeking to cast a ballot on Election Day had already voted in that election. See, e.g., A1155-56 (House Sess. HB 1355 1st Reading & Debate (Apr. 20, 2011) (Statement of Rep. Baxley)); A1351-52 (House Sess. HB 1355 2d Reading & Debate (Apr. 21, 2011) (Statement of Rep. McKeel)). According to the proponents of the inter-county mover changes, this left the system vulnerable to a specific form of potential fraud, whereby a voter might intentionally — or even inadvertently — cast an early in-person or absentee ballot in one county and then seek to vote again in a different county as an "inter-county mover." See id.; see also A1405 (House Sess. HB 1355 2d Reading & Debate (Apr. 21, 2011) (Statement of Rep. Gates) (stating that "I'm concerned that the status quo is so flawed that we don't even know in some instances where that abuse may be occurring")). To solve this problem, the legislators proposed requiring inter-county movers to cast a provisional ballot, which is "a real ballot" just like any other, A1156 (Rep. Baxley), but which "gives [election officials] time to go back and verify, and make sure everything's valid," A463 (Rep. Baxley). In sum, the sponsors of the inter-county mover amendments expressed their belief that the new system would help to prevent voter fraud because it would permit election officials to "check [inter-county movers] out, and eliminate ineligible voters," while still allowing eligible voters to cast provisional ballots that would be counted. A1417-18 (House Sess. HB 1355 2d Reading & Debate (Apr. 21, 2011) (Rep. Baxley)).

The sponsors of the inter-county mover amendments also indicated that the intercounty mover changes would represent an improvement over the benchmark law because they would encourage and facilitate the updating of addresses before voters went to the polls. That, in turn, would allow election officials to finalize their voter rolls before Election Day and thereby contribute to smoother overall operations at the polls on Election Day. See A713-16 (House State Affairs Comm. (Apr. 14, 2011) (Rep. Baxley)); A1204 (House Sess. HB 1355 1st Reading & Debate (Apr. 20, 2011) (Rep. Baxley)).

Taking all of these statements together, Florida has offered sufficient evidence to carry its initial burden of production to present facially neutral, nondiscriminatory justifications for the inter-county mover changes.

The United States and intervenors, however, have countered with significant evidence and arguments of their own. In particular, they have advanced a number of arguments tracking the key Arlington Heights factors — including arguments alleging disparate effects of the changes, discriminatory legislative statements, and departures from normal legislative procedures. They have also sought to undermine the stated justifications behind the inter-county mover amendments by contending that the changes were not necessary to prevent voter fraud, and hence that "verifiable reasons" are not "demonstrable in support" of the amendments. City of Richmond, 422 U.S. at 374, 95 S.Ct. 2296. Although we will consider each of those arguments in detail below, for now it is enough to say that the defendants have also met their burden of production to present evidence tending to "refute the covered jurisdiction's prima facie showing" of nondiscriminatory purpose. Bossier Parish II, 528 U.S. at 332, 120 S.Ct. 866; see New York, 874 F.Supp. at 400.

Because both sides have satisfied their initial production burdens, the burden-shifting framework falls away, and we are left with the task of evaluating all of the evidence in the record to determine whether Florida has carried its ultimate burden of persuasion under the purpose prong of section 5.

1. "The `important starting point' for assessing discriminatory intent under Arlington Heights is `the impact of the official action [and] whether it bears more heavily on one race than another.'" Bossier Parish I, 520 U.S. at 489, 117 S.Ct. 1491 (quoting Arlington Heights, 429 U.S. at 266, 97 S.Ct. 555) (second internal quotation omitted). That is so because "[s]ometimes a clear pattern, unexplainable on grounds other than race, emerges from the effect of the state action even when the governing legislation appears neutral on its face." Arlington Heights, 429 U.S. at 266, 97 S.Ct. 555 (citing, as examples, Gomillion v. Lightfoot, 364 U.S. 339, 81 S.Ct. 125, 5 L.Ed.2d 110 (1960) and Yick Wo v. Hopkins, 118 U.S. 356, 6 S.Ct. 1064, 30 L.Ed. 220 (1886)). In such a case, it is "relatively easy" to find discriminatory intent. Id. But "such cases are rare," the Court cautioned, and absent a "stark" pattern, disproportionate "impact alone is not determinative" of discriminatory intent. Id.

In this case, there is no evidence of starkly disparate effects on a minority group. To the contrary, as we have explained in detail in connection with our analysis of the section 5 effect prong, Florida has satisfied its burden of showing that the inter-county mover changes will not "bear more heavily" on any minority group. To be sure, the inter-county mover changes may affect disproportionately more minority voters than white voters because minorities are more likely to be inter-county movers. But the evidence indicates that the changes will not have materially adverse effects on the ability of minority voters to cast a ballot and effectively exercise the electoral franchise. See supra Part II.C.2. As such, the effects of these changes do not suggest that they were enacted with a discriminatory purpose.

Indeed, the lack of retrogressive effects constitutes evidence that the inter-county mover amendments were not passed for a discriminatory reason. After all, to find discriminatory intent with respect to these changes, we would have to conclude that Florida exhibited "malevolent incompetence" — that is, that the State wanted to harm minority voters by making it more difficult for them to cast a ballot but simply chose an ineffective means of doing so. Bossier Parish II, 528 U.S. at 332, 120 S.Ct. 866. Although such conduct is theoretically possible, the Supreme Court has advised that it is "unlikely." And because legislatures, like people, "usually intend the natural consequences of their actions," the absence of retrogressive effects on minority citizens suggests that the inter-county mover amendments were not motivated by a discriminatory purpose. Bossier Parish I, 520 U.S. at 487, 117 S.Ct. 1491. Indeed, we have been unable to find a case in which a court has found discriminatory purpose in the absence of retrogressive effects — except where the record contained strong statements by legislators that made their true purpose clear. See Busbee, 549 F.Supp. 494.

2. This brings us to the second Arlington Heights factor: "contemporary statements by members of the decisionmaking body," which the Court has said "may be highly relevant" in determining legislative intent. 429 U.S. at 268, 97 S.Ct. 555. As we have noted above in describing Florida's prima facie case, the State has submitted numerous contemporaneous statements by the proponents of the intercounty mover changes, indicating that their goal was to ensure the integrity of the electoral process and to encourage and facilitate the making of address changes before voters arrive at the polls. This evidence weighs in favor of a finding of nondiscriminatory intent. See id. at 270, 97 S.Ct. 555 (concluding that there was no discriminatory purpose, and relying in part on the fact that "the statements by the [decisionmaking body], as reflected in the official minutes, focused almost exclusively on" nondiscriminatory reasons for the decision).65

We must also consider, however, whether there are contemporaneous statements that do indicate a discriminatory purpose. On this point, the defendants point to a floor statement by State Senator Mike Bennett, who said that he did not want to make it easier for people to vote, but rather that it should be harder to vote — as it is "in Africa." A2242 (Senate Floor Debate (May 5, 2011)). The relevant text of Senator Bennett's statement is set out in the footnote below.66 Whether or not Senator Bennett actually intended his statement to have racial undertones, it certainly can be read that way. Hence, Senator Bennett's remarks could be considered to constitute evidence of discriminatory purpose by at least one Florida legislator.

Nonetheless, Senator Bennett's is the only statement to which the defendants point as evidencing a discriminatory purpose on the part of the Florida legislature. The purpose of a single legislator is normally too slim a reed upon which to rest a determination regarding the legislature as a whole. See Castaneda-Gonzalez v. Immigration & Naturalization Serv., 564 F.2d 417, 424 (D.C.Cir.1977) ("Statements by individual legislators should generally be given little weight when searching for the intent of the entire legislative body."). Also important is the fact that Senator Bennett was neither a sponsor nor a primary proponent of HB 1355, and did not play an important role in passage of the bill. Indeed, the floor statement was Senator Bennett's only public statement regarding the 2011 voting amendments. And although Senator Bennett was the Senate President Pro Tempore at the time, that office has limited authority, and there is no evidence it played any role in passage of the amendments.67 Cf. Busbee, 549 F.Supp. at 500-02, 508, 516-18 (finding that a Georgia redistricting plan was passed with a discriminatory purpose because the chairman of the Georgia redistricting committee was "a racist" who "made [o]vert racial statements" and "utilized the full power of his position and personality to insure passage of his desired Congressional plan" — a plan that was "motivated by a desire to minimize black voting strength").

Given all of these considerations, Senator Bennett's single statement is not enough to suggest that his purpose, whatever it was, represented the purpose of the Florida legislature as a whole. Accordingly, we conclude that the "contemporaneous statements" factor does not materially weigh in favor of a finding of discriminatory purpose.

3. Although it was not a factor expressly mentioned in Arlington Heights, the defendants contend that we may also infer discriminatory intent based on a finding that Florida's stated justifications for its inter-county mover amendments were pretextual. In particular, the defendants argue that concerns about voter fraud cannot have motivated these changes because there is no empirical evidence of inter-county mover fraud, nor have any local election officials expressed concerns that inter-county movers might be "double voting." See DOJ Br. 15-17, 60-61. And indeed, Florida has conceded that "[a]t the time of the Florida Legislature's consideration of HB 1355, the State of Florida knew of no instance" of such "double voting" by inter-county movers. A8281 (Fla. Resp. to Req. for Admiss. No. 12). Nor has it identified any instances since then. See Oral Arg. Tr. 24:17-25:2, 27:3-27:20. Likewise, election officials have confirmed that they have never seen double voting by inter-county movers at their precincts.68

We agree with the defendants that in some circumstances it is reasonable to infer discriminatory intent based on evidence of pretext. Cf. McDonnell Douglas Corp. v. Green, 411 U.S. 792, 93 S.Ct. 1817, 36 L.Ed.2d 668 (1973); Aka v. Wash. Hosp. Ctr., 156 F.3d 1284 (D.C.Cir.1998). But there are barriers to reaching such an inference here. The Supreme Court has instructed us that "[t]here is no question about the legitimacy or importance" of a state's "interest in counting only the votes of eligible voters," and that "[w]hile the most effective method of preventing election fraud may well be debatable, the propriety of doing so is perfectly clear." Crawford v. Marion Cnty. Election Bd., 553 U.S. 181, 196, 128 S.Ct. 1610, 170 L.Ed.2d 574 (2008) (plurality opinion of Stevens, J.). Moreover, it has instructed that this principle applies even where "[t]he record contains no evidence of any such fraud actually occurring" in the jurisdiction. Id. at 194, 128 S.Ct. 1610. Accordingly, the fact that a state has acted proactively to close a loophole in its election laws — i.e., that someone could "double vote" by going to precincts in multiple counties and claiming to have moved — does not by itself raise an inference of discriminatory intent. It might well be different, of course, had the State not only acted without evidence of fraud, but also acted in a way that materially increased the burden on minority voters. But as discussed above, we have found that the inter-county mover changes will not have such a materially adverse effect.

Furthermore, in evaluating whether Florida's stated justifications for its inter-county mover changes are pretextual, we regard it as significant that those changes also make ameliorative adjustments that will make it easier for inter-county movers to vote. As we have noted, the sponsors stated that one purpose of the changes was to encourage inter-county movers to update their addresses before going to the polls, in order to allow election officials to finalize their voter rolls before Election Day. Changes in the new law do help to accomplish this because, unlike the benchmark, the new law allows voters to update their addresses before Election Day by simply "[c]ontacting the supervisor of elections via telephone or electronic means." Fla. Stat. § 97.1031(1)(b) (2011). Because these ameliorative changes make it easier for inter-county movers to vote, they undercut the inference that Florida's true purpose was to disenfranchise them. And this, combined with the absence of retrogressive effects or direct statements of discriminatory intent, persuades us that an inference of unlawful pretext would be unwarranted.

4. Finally, citing Arlington Heights, the defendants contend that "[t]he specific sequence of events leading up to the ... decision" supports a finding of discriminatory intent. 429 U.S. at 267, 97 S.Ct. 555; see id. (noting that "[d]epartures from the normal procedural sequence ... might afford evidence that improper purposes are playing a role"). In particular, they argue that the process by which HB 1355 was passed and implemented was "unfair, severely limited, and unusual," and that the "rushed" process, "late-filed" strike-all amendments, and limitations on public debate all suggest discriminatory animus on the part of the Florida legislature. DOJ Br. 24-25; see id. at 6, 59, 76-78.

Based on the record before us, however, we are simply unable to determine whether the legislative process was unusual. Although there is some support for the defendants' position, the record of the debates on HB 1355 also reveals that the legislative process unfolded over a period of several months, and involved substantial testimony from a number of Florida election officials and members of the general public.69 In addition, several witnesses who are familiar with the standard procedures in the Florida legislature testified that strike-all or late-filed amendments are not necessarily unusual. See A4759, 4776, 4952-54 (Dep. of Pierce Schuessler, Legislative Affairs Director, Fla. Dep't of State); A9828-33 (Joyner Dep.).

The defendants also argue that "HB 1355's provision that the Voting Changes `shall take effect upon [HB 1355] becoming a law' is highly unusual," and that that departure from normal protocol requires a finding of discriminatory intent. DOJ Br. 52 (internal citation omitted). But the record contains several examples of Florida voting laws passed in recent years that were enacted to take immediate effect. See A10699-970 (collecting examples); see also A3869 (Lennard Dep.); A4918 (Schuessler Dep.). Relatedly, the defendants contend that the Florida Department of State's direction to election supervisors that they should implement the law in non-covered counties, even before the State receives preclearance for implementation in the covered counties, "deviates from prior precedent" and violates Florida law requiring uniform implementation of election provisions. DOJ Br. 52-53 (referring to A7269 (Fla. Dep't of State, Directive 2011-01 (May 19, 2011))). But the specifics of post-enactment implementation by Florida's executive branch tell us little — if anything — about whether the enacting legislature was motivated by discriminatory intent.

5. In sum, based on a consideration of the Arlington Heights factors and the arguments of the parties, we conclude that Florida has carried its burden of establishing that the inter-county mover changes were not enacted with "the purpose ... of denying or abridging the right to vote on account of race or color, or" membership in a language minority group. 42 U.S.C. § 1973c(a).

IV. Conclusions

To summarize, we have reached the following conclusions:

First, we conclude that we cannot preclear Florida's early voting changes at this time because the State has failed to satisfy its burden of proving that those changes will not have a retrogressive effect on minority voters if the covered counties offer only the minimum number of early voting hours required under the new statute, which would constitute only half the hours required under the prior law. We also conclude, however, that if Florida and the covered counties were to submit a preclearance plan that offered early voting for the maximum number of hours authorized by the new statute, which would be exactly the same number as under the prior law, and did so on a standard 7 a.m. to 7 p.m. schedule, Florida likely would satisfy its burden of proving that the overall effect of its changes in law would be nonretrogressive. Accordingly, we will deny Florida's request for a declaratory judgment granting preclearance of the early voting changes and dismiss Count Three of the Third Amended Complaint, without prejudice.

Second, we conclude that Florida has satisfied its burden of proving that the changes to the procedures for inter-county movers neither were enacted with a discriminatory purpose nor will have retrogressive effects on minority voters. We will therefore enter a declaratory judgment on Count Two of Florida's Third Amended Complaint preclearing those changes.

A separate order implementing these decisions will be issued this day.

FINDINGS OF FACT

I. Background and Procedural History

A. Factual Background

1. Five of Florida's sixty-seven counties are covered jurisdictions subject to the requirements of section 5 of the Voting Rights Act of 1965, 42 U.S.C. § 1973c. Those five counties are Collier County, Hardee County, Hendry County, Hillsborough County, and Monroe County. See A117 (Revised Jointly Stip. Facts ¶¶ 1-2). Each county became subject to coverage by operation of the language minority amendments to the Voting Rights Act in 1975. See 28 C.F.R. Pt. 51, App.; see also 41 Fed.Reg. 34329 (Aug. 13, 1976); 40 Fed.Reg. 43746 (Sept. 23, 1975). The State of Florida is not itself covered by section 5.

2. According to the 2010 census, the covered counties have a combined population of approximately 1,691,000, of which 13% are African-American and 26% are Hispanic. A7677 (Expert Report and Decl. of Russell Weaver) (compiling data from the 2010 U.S. Census Community Survey dataset). The covered counties' citizen voting age population is approximately 12% African-American, 16% Hispanic, and 68% white. (Other minorities, including Asian-Americans and Native Americans, make up the remainder.) Id. at A7683.

3. Hillsborough County is the largest of the covered counties, constituting 73% of the covered counties' combined population. Id. at A7677. The other covered counties, in descending order of population, are: Collier County, which accounts for 19% of the covered counties' total population; Monroe County, which makes up 4% of the covered counties' total population; and Hardee and Hendry Counties, which each constitute 2% of the covered counties' total population. Id.

4. On average, African-American and Hispanic populations in the covered counties have a lower per capita income and a lower median household income level than the white population. See id. at A7694-97. African-American and Hispanic households in the covered counties are also less likely than white households to own or have access to a motor vehicle. Id. at A7693.

5. County supervisors of elections administer the elections in each of Florida's sixty-seven counties. These supervisors are elected constitutional officers, see id. § 98.015(1), and are charged with, among other things, conducting elections, verifying, entering, and updating voter registration information, transmitting updated voter histories to the Department of State after an election, training poll workers, and reporting any instances of voter fraud, see id. §§ 98.015, 98.0981, 101.001-102.171; see also A7076 (Fla. Rule 30(b)(6) Dep.).

6. The Florida Secretary of State is Florida's chief election officer and the head of the Department of State, which houses the Florida Division of Elections. Id. §§ 20.10, 97.012. The Secretary of State is charged with, among other things, interpreting and implementing the requirements of the Florida Election Code, maintaining voter registration files, providing voter education assistance to the public, and providing written direction and opinions to the supervisors of elections regarding the performance of their official duties. Id. § 97.012.

7. The Florida Election Code regulates such matters as voter registration and the times, places, and methods of voting. The Election Code has been amended several times in recent years, including in 2001, 2005, and most recently in 2011. See Fla. Laws ch. 2001-40; Fla. Laws chs. 2005-277, 2005-278; Fla. Laws chs. 2011-40.

B. Procedural History

8. This case involves several of the 2011 amendments to the Florida Election Code. Those amendments were included in an omnibus bill, Committee Substitute for Committee Substitute for House Bill No. 1355 ("HB 1355"), which made approximately 80 sets of changes to the provisions of the Election Code. The two particular sets of changes at issue in the accompanying opinion involve: (1) amendments to the days and hours of early in-person voting; and (2) amendments to the voting procedures for inter-county movers (i.e., voters who move between Florida counties without updating their registered address and then seek to vote in their new county of residence).

9. The final 2011 voting changes were the product of two similar bills — HB 1355 and SB 2086 — that moved through the Florida House and Senate over a period of several months.

10. HB 1355 was first introduced on March 7, 2011, by Representative Dennis Baxley. A119 (Revised Jointly Stip. Facts ¶ 18). The bill was approved by the House Government Operations Subcommittee, A119 (Revised Jointly Stip. Facts ¶ 24); see A464-65, and then the House State Affairs Committee, A692-788 (House State Affairs Comm. (Apr. 14, 2011), after undergoing substantial revisions throughout the committee process. On April 20 and 21, 2012, the House debated and amended HB 1355, A120 (Revised Jointly Stip. Facts ¶¶ 32-35); see A1149-1333 (House Sess. HB 1355 1st Reading & Debate (Apr. 20, 2011)), before approving the bill by a vote of 79 to 37, A120 (Revised Jointly Stip. Facts ¶ 36).

11. SB 2086 was introduced by Senator Miguel Diaz de la Portilla on March 29, 2011. A119 (Revised Jointly Stip. Facts ¶ 21). The bill was approved by the Senate Rules Committee on Ethics and Elections, see A605-29 (Senate Rules Comm. (Apr. 4, 2011)), then the Senate Rules Committee, see A120 (Revised Jointly Stip. Facts ¶ 30), and finally the Senate Budget Committee, A1640-42 (Senate Budget Comm. (Apr. 26, 2011)), after undergoing substantial revisions in each committee. On May 4, 2011, the Senate adopted an amendment substituting HB 1355 for SB 2086, A121 (Revised Jointly Stip. Facts ¶ 45), as well as an amendment that would shorten Florida's early voting period from 14 possible days to 8 days, but expand the maximum available hours on each day from 8 to 12. See A1861-74 (Senate Sess. — SB 2086/HB 1355 2nd Reading & Debate (May 4, 2011)). The next day, the Senate debated HB 1355 and approved it by a vote of 25 to 13. A121 (Revised Jointly Stip. Facts ¶ 49); see A2213 (Senate Floor Debate (May 5, 2011)).

12. The House received the Senate's version of HB 1355, debated the bill, and approved it by a vote of 77 to 38 on May 5. A 121-22 (Revised Jointly Stip. Facts ¶¶ 50-52).

13. On May 19, 2011, the bill was signed into law by Florida Governor Rick Scott, and was subsequently codified at Florida Laws, ch. 2011-40. A122 (Revised Jointly Stip. Facts ¶ 54). The particular changes at issue before us now appear in Fla. Stat. §§ 101.045, 101.657(d) (2011).

14. Florida submitted the voting changes in HB 1355 to the Attorney General for administrative preclearance on June 8, 2011. The submission was received by the Department of Justice on June 9, 2011. A118 (Revised Jointly Stip. Facts ¶ 5); see [147] Third Am. Compl. ("Fla. Compl.") ¶ 30.

15. On July 29, 2011, the State withdrew four sets of voting changes from that administrative preclearance process. See A219 (Letter from Daniel E. Nordby, Gen. Counsel, Fla. Dep't of State (July 29, 2011)). Those four sets of changes included the inter-county mover and early voting changes at issue before us, along with changes imposing additional requirements upon third-party voter registration organizations (TPROs) and amendments to the provisions governing the time frame during which signatures on citizen initiative petitions are valid (the "citizen initiative changes"). Id.; see Fla. Compl. ¶¶ 28, 30.

16. The other 76 sets of changes in HB 1355 remained before the Department of Justice for preclearance, and on August 8, 2011, the Attorney General precleared them. See A118 (Revised Jointly Stip. Facts ¶ 8); Fla. Compl. ¶ 30.

17. On August 1, 2011, Florida commenced this action seeking judicial preclearance of its inter-county mover, TPRO, early voting, and citizen initiative changes. See [1] Compl. Florida also filed a motion to convene a three-judge district court; that motion was granted by an order dated August 9, 2011.

18. Early in the litigation, Florida amended its complaint twice: once on October 11, 2011, adding claims challenging the constitutionality of the Voting Rights Act, see [39] First Am. Compl. ¶¶ 107-116; and once on October 24, 2011, to include a request for judicial preclearance of the amended state regulations implementing the TPRO changes, see [54] Second Am. Compl. ¶¶ 85-86 & Ex. B.

19. Throughout late August and early September, 2011, more than two dozen individuals and organizations opposed to Florida's voting changes — including the Florida State Conference of the NAACP, the National Council of La Raza, the League of Women Voters of Florida, Project Vote, Voting for America, several Florida legislators and election officials, and a number of registered Florida voters who are members of racial and language minority groups — filed motions to intervene as defendants in this action. The court granted those motions on October 19, 2011, giving the defendant-intervenors leave to intervene permissively pursuant to Rule 24(b)(1) of the Federal Rules of Civil Procedure. See [42] Oct. 19, 2011 Order.

20. On October 18, 2011, Florida filed a motion to expedite, seeking a final ruling on all of its claims in advance of its presidential preference primary on January 31, 2012. See [41] Pl.'s Mot. to Expedite. This court denied that motion in part, for the reasons set forth in a Memorandum Opinion issued on October 28, 2011, 820 F.Supp.2d 85 (D.D.C.2011). See [55] Oct. 28, 2011 Order; [56] Oct. 28, 2011 Mem. Op. Nonetheless, the court did adopt a somewhat expedited schedule for both discovery and briefing. See [61] Scheduling & Procedures Order (Nov. 3, 2011). Moreover, in an effort to further streamline the resolution of this case, the court also bifurcated Florida's statutory preclearance claims from its constitutional challenges, and ordered separate briefing on those two subjects. See id. at 5; see also [94] Apr. 20, 2012 Order.

21. During the discovery phase of this case, the intervenors moved to compel deposition testimony from four Florida legislators and two legislative staff members, but a federal district court in Florida denied the motion on the grounds of legislative privilege. See [70] Defendant-Intervenors' Second Notice of Ancillary Proceedings at 2 & Ex. A; see also Florida v. United States, 4:12mc3 (N.D.Fla. Feb. 3, 2012).

22. In early March 2012, the United States informed the court that after conducting discovery and reviewing the record, it had concluded that Florida had met its burden of demonstrating that the citizen initiative changes were neither enacted with a discriminatory purpose nor will have retrogressive effects. See [79] March 8, 2012 Mem. Order at 2-3. Accordingly, Florida re-submitted those changes for administrative preclearance on March 9, 2012, and on March 21, 2012, the Attorney General informed Florida that no objection would be interposed. See [84] United States' Notice to the Court. The parties then filed a stipulation of dismissal as to Count II of Florida's Second Amended Complaint, which dealt with those citizen initiative changes. See [85] Stip. of Dismissal.

23. The discovery and briefing schedule regarding the three remaining sets of voting changes was revised several times at the request of various parties. See, e.g., [87] Scheduling & Procedures Order (March 27, 2012); [94] Apr. 20, 2012 Order. Ultimately, the discovery period closed on February 29, 2012, although the parties were permitted to take de bene esse depositions of a number of witnesses throughout April and May 2012, and to submit supplemental or rebuttal declarations by some of their expert witnesses. The parties then filed proposed findings of fact and conclusions of law on the basis of the written record. Briefing on the statutory preclearance question was completed by May 24, 2012.

24. During an on-the-record telephonic conference on June 1, 2012, the parties indicated that they saw no need for live testimony regarding Florida's statutory preclearance claims, "and represented that they [we]re satisfied with the record as it [had] been submitted." [106] June 5, 2012 Order at 1 n.1. Accordingly, the court did not hold a live trial in this case. The full three-judge panel did, however, hear oral argument for 5 hours on the statutory preclearance issue on June 21, 2012. Thereafter, the court ordered and received supplemental briefing from the parties with respect to several specific issues raised during that hearing. The court also accepted a modified amicus curiae brief filed by U.S. Senator Bill Nelson, the senior senator from Florida.

25. Meanwhile, the TPRO changes were the subject of collateral litigation in the Northern District of Florida, where they were challenged on the grounds, inter alia, that they violate the First Amendment and the National Voter Registration Act (NVRA). On May 31, 2012, the district court issued an order preliminarily enjoining many of the TPRO changes, finding that the plaintiffs were likely to prevail on their First Amendment and NVRA objections to those changes. See League of Women Voters of Fla. v. Browning, 863 F.Supp.2d 1155 (N.D.Fla.2012).

26. Florida notified this court on June 12, 2012, that in light of that preliminary injunction, it would "no longer seek in this action to preclear the [TPRO] changes that the [district court in Florida] preliminarily enjoined." [109] Statement Regarding the Effect of the Recent Order in the N.D. Fla. at 2. Florida stated in that June 12 filing that it would "voluntarily withdraw the enjoined [TPRO] Changes from this judicial preclearance action," id. at 5, indicating that it was waiting for a minor clarification from the Florida district court before doing so, see id. at 3 n. 2. Likewise, during the oral argument on June 21, 2012, Florida stated that it would "withdraw the [provisions] that have been preliminarily enjoined ... [a]nd proceed on the remainder." Oral Arg. Tr. 6:08-6:12.

27. On July 16, 2012, after weeks had passed without further word from Florida, this court convened a telephonic status conference to determine whether the State had made a decision regarding which TPRO provisions it would withdraw from the action, and which it was still seeking to preclear. The State advised that it had not yet made a final decision.

28. On August 10, 2012, the State informed the court that it had settled the TPRO litigation in Florida. See [104] Mem. of Points and Authorities in Support of Pl.'s Mot. for Leave to Amend the Compl. at 1. It filed a motion to amend its complaint in this case, seeking to remove from its request for preclearance those TPRO changes that the district court in Florida had enjoined, while keeping the remaining changes, some in modified form. See id. at 4 n. 1, 5. On August 13, Florida advised the court that it would promptly submit the remaining, non-enjoined TPRO changes to the Attorney General for administrative preclearance. On August 15, the court granted Florida's unopposed motion to amend its complaint.

C. Voting Changes at Issue

1. Early Voting

29. Florida enacted the law that was in effect immediately prior to HB 1355 in 2005. That law provided that the State's potential early voting period was 14 days, beginning on the 15th day before an election and ending on the second day before that election. See Fla. Laws ch. 2005-277. That law also limited the available early voting hours to exactly 8 hours per day on weekdays and 8 hours in the aggregate over each weekend, yielding a total of 96 hours of early voting. See Fla. Stat. § 101.657(d) (2006). Early voting sites were required to "open no sooner than 7 a.m. and close no later than 7 p.m. on each applicable day." Id. Local supervisors were free to select the specific voting hours for each voting day as they saw fit. See id. The statute was precleared by the Attorney General on September 6, 2005, see A9002-03, and it is the most recent law in force or effect in Florida and the covered counties prior to the 2011 amendments in HB 1355.

30. Under that prior law, local supervisors of elections in each of the five covered counties have chosen a range of voting hours varying by election and early voting site. The following five permutations were used by the covered counties in recent elections: (1) 8 a.m. to 1 p.m. and 2 p.m. to 5 p.m.; (2) 8:30 a.m. to 4:30 p.m.; (3) 9 a.m. to 5 p.m.; (4) 10 a.m. to 6 p.m.; and (5) 10:30 a.m. to 6:30 p.m. See A8533-34, 8539, 8542-43, 8549, 8560, 8564-65, 8570, 8581, 8584-85, 8591, 8604, 8607-09, 8614-15 (Fla.'s Resps. to First Set of Interrogs. of Def. United States, Ex. D); [136] United States' Resp. to the Benchmark Questions Raised by the July 3, 2012 Minute Order, Attach. 1 at 5, 9-10, 12, 24-25, 27, 31.

31. In each of Florida's five covered counties, local supervisors of elections also exercised their discretion to offer weekend early voting only on Saturdays. See A3194 (Sawyer Dep.); A3415 (Dep. of Jennifer Edwards, Collier County SOE); A3495 (Dep. of Lucretia Strickland, Hendry County SOE); A3841 (Lennard Dep.); A3912 (Dep. of Jeffery Ussery, Hardee County SOE). In other words, election officials satisfied the mandate that they offer "8 hours [of early voting] in the aggregate each weekend" by offering 8 Saturday hours of early voting on each of the two weekends that were required under the pre-2011 early voting law, Fla. Stat. § 101.657(d) (2010). This meant that in each covered county, only 12 of the available 14 days of early voting were actually used, and no covered county ever offered early voting on Sundays.70

32. HB 1355 amended the number of days, the number of hours, the specific hours, and the weekend times that early voting may be offered in Florida. First, under the new law, the early voting period begins "on the 10th day before an election... and end[s] on the 3rd day before the election," for a total of only 8 days instead of the previous 12. Fla. Stat. § 101.657(d) (2011).

33. Second, the law now gives local election supervisors the discretion to determine the number of daily hours of early voting in their counties, subject to the constraint that "no less than 6 hours and no more than 12 hours" be offered on each of the 8 early voting days. Id. As a result, Florida's covered counties might still offer the same total hours of early voting (96 hours) that were required under the pre-2011 law, but only if their local election supervisors decide to offer the maximum 12 hours of early voting on each of the 8 days. If, on the other hand, a local supervisor chooses to offer the minimum number of hours (i.e., 6 hours per day), then the early voting period would last only 48 hours in total — exactly half of the hours that were offered under the prior law.

34. Third, the new law also removes the requirement that voting take place between 7 a.m. and 7 p.m. each day. Instead, local supervisors of elections have the discretion to determine the specific hours of early voting in their counties, as long as they offer "no less than 6 hours and no more than 12 hours" each day. Id. In the event that local election supervisors offer the maximum 12 hours of early voting on a given day, then the early voting hours for that day would necessarily include hours that fall outside the standard 8-hour workday extending from 9 a.m. to 5 p.m. However, in the event election supervisors offer the minimum 6 hours of early voting on a given day, the 6-hour early voting day may be entirely within, entirely outside, or straddle the standard workday.

35. Fourth, the new early voting statute also mandates some additional weekend hours of early voting. In particular, the new early voting period runs from the Saturday two weekends before the election to the Saturday immediately before Election Day, see id., meaning that the early voting period under HB 1355 now requires three weekend days of early voting: two Saturdays and one Sunday. And as already explained, the new statute mandates anywhere from 6 to 12 early voting hours on each day. Id. Accordingly, if HB 1355 were implemented in the covered counties, it would result in at least 6 (and up to 12) hours of Sunday early voting that were never before offered in those counties. Moreover, election officials would also have the discretion to offer up to 36 total hours of weekend early voting (12 hours per day on each of 3 weekend days) — for a net gain of 20 weekend early voting hours over the prior law, which offered exactly 16 weekend hours (8 hours in the aggregate on each of two weekends).

2. Inter-County Movers

36. In 2005, Florida amended its Election Code to allow voters to make inter-county address changes at the polls and then cast a regular ballot. Under that version of the law — which was precleared by the Department of Justice on September 6, 2005, see A9002-03, and which is the most recent pre-2011 statute in force and effect in Florida — inter-county movers were required to complete an affidavit of change of address or a voter registration application, listing their new address of residence and affirming that they had not already voted in the precinct of their former residence. See Fla. Stat. § 101.045(2)(a) (2006). After their eligibility to vote was verified, such voters could then cast a regular ballot, which would be canvassed and counted like all other ballots. Id. § 101.045(2)(c)-(d).

37. Under HB 1355, however, "an elector whose change of address is from outside the county may not change his or her legal residence at the polling place and vote a regular ballot." Fla. Stat. § 101.045(2)(b) (2011). Instead, "such elector is entitled to vote a provisional ballot." Id. The only exception is for "active uniformed services voter[s]" and members of their families, who are still permitted to cast a regular ballot after affirming their inter-county change of address and having their eligibility verified. Id.

38. Once completed, provisional ballots cast by inter-county movers "shall be placed in a secrecy envelope and thereafter sealed in a provisional ballot envelope." Id. § 101.048(1). The provisional ballots will then be deposited in a ballot box and returned to election officials, whereupon the county canvassing board will examine the ballots to determine if the voters were eligible to vote at that precinct and had not already cast a ballot in the election. Id. § 101.048(2)(a).

39. Under Florida law (both pre- and post-2011), "[a] ballot of a person casting a provisional ballot shall be counted unless the canvassing board determines by a preponderance of the evidence that the person was not entitled to vote." Id. (emphasis added); see also id. § 101.048(2)(b)(1)-(2). Moreover, voters who are required to cast a provisional ballot are permitted — though not required — "to present written evidence supporting [their] eligibility to vote to the supervisor of elections by not later than 5 p.m. on the second day following the election." Id. § 101.048(1). Provisional voters must be given written instructions concerning that right, and "[e]ach supervisor of elections shall establish a free access system that allows each person who casts a provisional ballot to determine whether his or her provisional ballot was counted in the final canvass of votes and, if not, the reasons why." Id. § 101.048(5)-(6).

40. HB 1355 also makes pre-election address changes easier for Florida voters. Under the pre-2011 law, inter-county address changes had to be completed "using a voter registration application signed by the elector." Fla. Stat. § 97.1031(2) (2010). Now, however, voters can notify the supervisor of elections of their change of address by "[s]ubmitting the change on a voter registration application or other signed, written notice," or by "[c]ontacting the supervisor of elections via telephone or electronic means." Fla. Stat. § 97.1031(1)(b) (2011).

II. Effects

41. In the appendices to their proposed findings of fact and conclusions of law, the parties have submitted reports, declarations, and deposition testimony from a number of witnesses, including expert witnesses for each party, several Florida state legislators, representatives and volunteers from third-party voter registration organizations, local election supervisors from each of the five covered counties (as well as several other non-covered Florida counties), employees of the Florida Department of State, and several other individuals with knowledge of how the new law might affect voters in Florida's covered counties. That testimony, combined with other record evidence, allows us to make the following findings regarding the likely effects of the 2011 voting changes.

A. Early Voting Changes

1. Voters Affected by Early Voting Changes

42. We find that minority voters disproportionately use early in-person voting, and therefore will be disproportionately affected by the changes in early voting procedures.

43. As the intervenors' expert witness, Professor Paul Gronke,71 determined, the proportion of African-American usage of early in-person voting in Florida "has exceeded White usage of early in-person voting in four of five [recent] federal elections," and "substantially exceeded White usage in both the 2004 and 2008 presidential elections." A 10092, 10104 (Am. Expert Report of Prof. Paul Gronke).72

44. In the 2008 general election, more than half (54%) of African-American voters in Florida cast ballots using early in-person voting — twice the rate of white voters. A10106 (Am. Expert Report of Prof. Gronke, Ex. Seven); see A7818 (Expert Report of Dr. Stewart, Attach. J). And although rates of early voting declined across the board in 2010, the African-American usage rate still exceeded the white rate by a factor of about one-third in the 2010 general election. See A10106 (Am. Expert Report of Prof. Gronke, Ex. Seven); see also A9091 (Rebuttal Decl. of Prof. Gronke) ("[T]here was a massive increase, of nearly 75 percent, in the use of early in-person voting by African Americans" between 2006 and 2010, and "[t]his increase in early in-person voting by African Americans dramatically exceeded the increase among White voters.").

45. These disproportionate usage rates hold true in Florida's five covered counties as well. In 2008, for example, 52% of all African-American voters in the covered counties cast an early in-person ballot, compared to only 28% of white voters. Id. at A10095-96, 10105; see also id. at A7821 (Expert Report of Dr. Stewart, Attach. M). African-American rates of early in-person voting in the five covered counties also "remained statistically significantly higher [than white rates] in the 2010 primary and general elections." A10096 (Am. Expert Report of Prof. Gronke). Even Florida's own expert witness, Professor M.V. (Trey) Hood III, acknowledged that he "can see the pattern" of higher levels of African-American usage of early in-person voting in the covered counties after the 2008 primary election. A5855 (Hood Dep.); see also A9041-42 (Expert Report of Prof. Hood) (showing that the African-American rates of early voting in the covered counties exceeded the rates of white voters to a significant degree in the 2008 general election, the 2010 primary, and the 2010 general election (and, to a lesser extent, in the 2008 primary)).

46. Furthermore, all available evidence suggests that these trends "will continue into the 2012 general [election] and likely in[to] the future." See A9918 (De Bene Esse Dep. of Prof. Gronke); A10096 (Am. Expert Report of Prof. Gronke); A10079 (De Bene Esse Dep. of Prof. Gronke) ("I think that history will show that 2008 ha[d] a particularly high rate [of African-American early voting], but that that adoption rate by African-Americans had a lasting impact, and that the higher rate of usage will continue."); id. at A10034-36. Florida's expert likewise agreed that it is "more likely than not" that in the 2012 general election, the African-American usage rate of early in-person voting will be higher than the white usage rate. A5875-76 (Hood Dep.).

47. The evidence also shows that African-American voters disproportionately used the first five days of the preexisting early voting period — i.e., the Monday through Friday of the week that falls two weeks before Election Day — all of which will now be eliminated under HB 1355. In the 2008 general Presidential election, for example, approximately 17.25% of African-American voters in the covered counties cast an early in-person ballot during the first five days of early voting (the so-called "repealed days" of early voting), compared to only 9.3% of white voters. A10097 (Am. Expert Report of Prof. Gronke); see A7821 (Expert Report of Dr. Stewart, Attach. M). In other words, African-American voters used the repealed days of early voting at rates nearly double those of white voters in 2008. The difference between these percentages "far exceeds the... statistical significance criterion." A10097 (Am. Expert Report of Prof. Gronke).

48. The data with respect to use of the repealed days in the 2010 elections is more inconclusive. In the 2010 primary election, African-Americans "continued to vote at a higher rate" than whites during the first week of early voting. A10097 (Am. Expert Report of Prof. Gronke) (describing a usage rate of 7.91% for African-American voters, compared to 6.84% for white voters). In the 2010 general election, however, African-Americans voted at a slightly lower rate than whites during that first week. Id. (indicating that 6.88% of African-Americans voted during the repealed days in the 2010 general election, compared to 8.34% of whites). It is unclear whether either of these differences is statistically significant: Professor Gronke testified that he believed that both were, see A6211-13 (Gronke Dep.), but the record does not contain any statistical significance calculations, and Professor Gronke's expert report does not mention any conclusions to that effect, see A10097.

49. In finding that African-American voters in the covered counties will be disproportionately affected by the reduction in early voting days under the new law, we reject the contrary opinions of Florida's expert witness, Professor Hood.73 We do so because we find that the analysis underlying his conclusions suffers from a number of methodological flaws.

50. First, Professor Hood asserts that the effects of the early voting changes will be "disproportionately borne by [white] voters" because they "comprise the greatest share of total number of early votes cast" in the covered counties. A9061 (Expert Report of Prof. Hood). But as the intervenors' expert, Professor Gronke, convincingly explains, this "analytic method violates a basic tenet of comparative analysis" because it fails to "control for... the size of the underlying subgroups." A9089 (Rebuttal Decl. of Prof. Gronke). After all, it is "no surprise that [white] voters are ... the majority of users of early in-person voting" because they also represent "the vast majority of all Florida voters," id. (emphasis added), including those in the covered counties, see A7683-84 (Expert Report and Decl. of Russell Weaver) (collecting census data in the covered counties). See also A7821 (Expert Report of Dr. Stewart, Attach. M) (showing that 525,324 white voters cast a ballot in the 2008 general election, compared to only 86,314 African-American voters and 66,391 Hispanic voters); A7684 (Expert Report and Decl. of Russell Weaver) (collecting census data showing that in the 2010 general election, approximately 70% of the registered voters in the covered counties were white, while African-Americans and Hispanics each made up only approximately 12% of the voting population in those counties). That is why, as Professor Gronke explains, adjusting for population demographics is critical in assessing the potential retrogressive effects of a given voting change:

A9888-90 (De Bene Esse Dep. of Prof. Gronke); see A9089 n.1 (Rebuttal Decl. of Prof. Gronke) (explaining how Prof. Hood's flawed analysis could lead to endorsing a poll tax that has an obviously disparate impact on minority voters). As even Professor Hood concedes, the accepted practice in the social sciences is to look at the rate of impact on different groups, not simply absolute numbers. A5926-28 (Hood Dep.). Yet his calculations do not follow that accepted method of statistical analysis. See id. at 5840-42.

51. Second, we reject other calculations in Professor Hood's expert report because we agree with the intervenors' expert that "[i]n several instances Professor Hood inappropriately pools together groups of dissimilar data, which is not methodologically appropriate." A9092 (Rebuttal Decl. of Prof. Gronke). For example, Professor Hood "attempts to draw conclusions based on data `pooled' from different kinds of elections, without offering a reason to believe that early voting patterns are in fact common across the different types of elections." Id. And "[t]here is no evidence that Professor Hood conducted a pooling test, a statistical tool that helps determine whether it is valid to aggregate data" from those different types of elections. Id.; see A5762-64 (Hood Dep.) (conceding that he pooled the data without running any of the standard statistical tests to determine whether such pooling was appropriate). This problem is exacerbated by the fact that Professor Hood often "aggregat[es] data from all elections analyzed in his Report except for the 2008 general election," thereby further distorting the data. A9092 (Rebuttal Decl. of Prof. Gronke).

52. Professor Hood also frequently lumps African-Americans and Hispanics into a single category of "Minorities," which misleadingly flattens the data because, unlike African-Americans, Hispanic voters use early voting at about the same rate as whites. See A7818 (Expert Report of Dr. Stewart, Attach. J); A10105 (Am. Expert Report of Prof. Gronke, Ex. Six). This "masks significant differences between White and African-American rates of early in-person voting." A9092 (Rebuttal Decl. of Prof. Gronke). As Professor Gronke notes, "[i]t is inconsistent with political science research standards to pool together these two racial groups when they demonstrate different behaviors in voting." Id. at 9093.

53. Finally, for several reasons, we reject Professor Hood's contention that the 2008 general election was an "outlier" that should be ignored. A9042 (Expert Report of Prof. Hood); see id. at A9043, A9061. First, although the 2008 election was of course unique in certain respects, the record does not support the assertion that that election was an "outlier" that can be discarded out of hand. Rather, the record evidence suggests that the 2008 election is highly predictive of what is likely to happen in 2012. The expert witnesses in this case are all generally in agreement that, when assessing future usage rates of early voting, comparisons are best made between "like" elections, and that the most recent analogous election is the best predictor of what will happen in the future. Accordingly, the 2008 general Presidential election represents the best guidepost for projecting how early in-person voting will be used in the upcoming 2012 general Presidential election. See, e.g., A5657-58 (Stewart Dep.); A9091 (Rebuttal Decl. of Prof. Gronke); A10098 (Am. Expert Report of Prof. Gronke); see also A5867-68 (Hood Dep.) (conceding that "the 2008 presidential election would be our best gauge" of "what's going to occur in the 2012 general election"). For this reason alone, "it is not by any means an `outlier' in terms of predictive value." A9091 (Rebuttal Decl. of Prof. Gronke). And that is particularly true because many of the factors that supposedly rendered the 2008 election "anomalous" will again be factors in 2012:

Id.

54. Moreover, even setting aside the specifics of the 2008 and 2012 general elections, the record indicates that the 2008 general election was not a mere one-off phenomenon. The trend of increased African-American usage of early in-person voting pre-dated 2008 to some degree: African-American early voting rates in Florida exceeded white early voting rates in 2004, as well. See A10092, 10104 (Am. Expert Report of Prof. Gronke). And after the 2008 election, the rates of African-American usage of early in-person voting in Florida have continued to exceed those of white voters to a statistically significant degree. See, e.g., A9090 (Rebuttal Decl. of Prof. Gronke). Hence, it is not true, in the words of Florida's expert, that once the 2008 election is removed from the picture the rates of white and African-American early voting are "roughly comparable," A9041 (Expert Report of Prof. Hood). And in any event, "[t]he term `roughly comparable' is not an accepted term of art or standard for statistical comparisons in the field of political science research." A9090 (Rebuttal Decl. of Prof. Gronke). Rather, "[t]he methodologically appropriate starting point is to determine whether the differences are statistically significant." Id.74 And, even discounting the 2008 election, the data shows a clear and statistically significant "trend of higher usage [of early voting] over time by African American voters in the Covered Counties and statewide." Id.; see A9090 (Rebuttal Decl. of Prof. Gronke).

55. In short, rather than being an "outlier," the evidence suggests that the 2008 general election is best seen as a "game-changer" that simply magnified already nascent trends in African-American preferences for early in-person voting. See A10069 (De Bene Esse Dep. of Prof. Gronke).75 Indeed, Florida's own expert witness acknowledges that the 2008 election was not an outlier when compared to the relative rates of African-American versus white early in-person voting in the four most recent federal elections in Florida. A5965-66 (Hood Dep.). As such, there is no sound basis for disregarding the 2008 election in our evaluation of the effects of the new statute's changes in early voting procedures.

2. Burdens Imposed by Early Voting Changes (at 48 Hours)

56. As a preliminary matter, we cannot predict with any confidence the number of early voting hours the covered counties will offer. No covered county's election supervisor has submitted a county plan for preclearance. Most of the supervisors' responses to questioning by the defendants on this subject may at best be described as equivocal. See, e.g., A3400-3401, 3447-48 (Edwards Dep.); A3999 (Ussery Dep.). And at least one covered-county supervisor indicated that she would not use the full 96 hours of early voting, unless required to do so. See A3530-32 (Dep. of Lucretia Strickland, Hendry County SOE). Moreover, Florida acknowledges that it "has not issued any rule, directive, or guidance to Florida supervisors of elections ... regarding the manner in which [they] should exercise the discretion granted by HB 1355" to determine the number of early voting hours in their counties. A8283 (Fla. Resp. to Req. for Admiss. No. 18).

57. If the covered counties offer only 48 hours of early voting (i.e., only 6 hours per day) as the new law permits, that change would impose a material burden on African-American voters' effective exercise of the electoral franchise. Under such circumstances, not only would the number of early voting days be reduced by one-third (from 12 to 8), but the total available early voting hours would be cut in half (from 96 to 48). Moreover, with only 6 hours available per day, it is likely that early voting would start after the workday starts and would end before the workday ends, making it even more inaccessible to many minority voters who have inflexible work schedules. See A9142-43 (Decl. of Cynthia Slater, 2d Vice-President of Florida NAACP).

58. A two-week early voting period is important to get-out-the-vote (GOTV) efforts in minority communities. See, e.g., A9143 (Slater Decl.); A9227-28 (Decl. of Rev. Charles McKenzie, Florida state liaison for the Rainbow PUSH Coalition); A9237-38 (Decl. of Ella Kate Coffee, African-American resident and GOTV volunteer in Hillsborough county). Record evidence suggests that such efforts are important in enabling African-Americans "who want to vote but need help getting to the polls" to exercise the franchise. A9237 (Coffee Decl.); see A7693 (Expert Report and Decl. of Russell Weaver, Ex. 2). With a substantially reduced early voting period, third-party groups would not be able to assist minority voters as effectively. See A9237. This, in turn, would likely make it more difficult for those minority voters who rely on such efforts to make it to the polls. Florida has not submitted any evidence to the contrary. Indeed, although Florida's expert witness initially hypothesized that early voters should be able to adjust to even such a dramatic contraction in the early voting period, A9059-60 (Expert Report of Prof. Hood), he later conceded that there is no empirical support for that claim, and that it was nothing more than an "assumption" or "supposition" on his part, A5861, 5917-22 (Hood Dep.); see A9095 (Rebuttal Decl. of Prof. Gronke) ("I know of no empirical support for the conclusion that voters will successfully adjust ... under these conditions.").

59. Moreover, even if all of the voters who would have used the repealed days of early voting did attempt to adjust to a shortened early voting schedule of only 48 hours over 8 days, that shift would create problems of its own for minority voting. In the 2008 general election, for example, 71,670 voters — 14,897 of whom were African-American — cast ballots in the covered counties during the early voting days that the new law has repealed. See A7820-21 (Expert Report of Dr. Stewart, Attachs. L & M); see also A9054-57 (Expert Report of Prof. Hood) (showing that approximately one-third to two-fifths of early voters voted in the first (repealed) week of early voting). According to testimony in the record, a shift of that magnitude to the remaining early voting days would lead to substantially increased lines, overcrowding, and confusion at the polls, which would in turn discourage some reasonable minority voters from waiting to cast their ballots. See, e.g., A3170-71, 3183-87 (Dep. of Harry Sawyer, Monroe County SOE); A3744-45, 3748 (Dep. of Earl Lennard, Hillsborough County SOE). Indeed, election officials in Florida have testified that an extensive early voting period is necessary because Florida's "electoral infrastructure is completely maxed out," A440 (House State Affairs Comm. (Apr. 1, 2011) (Statement of Ion Sancho, Leon County SOE)), such that the State "would not be able to process record numbers of voters" in a substantially shorter time frame, A976-78 (Senate Rules Comm. (Apr. 15, 2011) (Sancho Statement)). Florida legislators, too, have warned that a "shortened period of early voting will cause congestion and long lines in populous areas of the State, including predominantly African-American neighborhoods in Hillsborough County," and will thereby "discourage [minority] voters from voting." A9108 (Decl. of Sen. Arthenia Joyner).

60. The academic scholarship and commentary is currently in a state of flux as to how the availability of early in-person voting affects overall voter turnout. The consensus prior to the 2008 election appears to have been that early in-person voting was a convenience that had an "insignificant or marginal effect on increasing the likelihood [that] an individual will vote." A10131 (Robert Stein & Greg Vonnahme, Early, Absentee, and Mail-in Voting, in THE OXFORD HANDBOOK OF AM. ELECTIONS & POL. BEHAVIOR (Jan Leighly, ed., 2010), at 185); see A5496 (Stewart Dep.) ("[T]he research up until around 2008 ... is that, in general, early voting procedures make voting more convenient for people who look like they have a ... propensity to vote, but do[] not generally increase turnout overall."); see also A9093-95 (Rebuttal Decl. of Prof. Gronke). Although Professor Gronke contended in his rebuttal declaration that "his own comprehensive review of the literature in 2008" showed that "convenience voting has a small but statistically significant impact on turnout, with most estimates of the increase in the 2%-4% range," A9094 (Rebuttal Decl. of Prof. Gronke), during his de bene esse deposition he backed off that claim, admitting that he "cannot put [his] finger on one particular statement" in any of the 2008 or pre-2008 literature "that indicates that early in-person voting specifically has a relationship to overall turnout." A9986 (De Bene Esse Dep. of Prof. Gronke); see id. at A9981-83.

61. According to the intervenors' expert, however, "[t]he 2008 presidential and subsequent elections have challenged the conventional wisdom [regarding early voting], primarily because of changing voting patterns in the South." A10091 (Am. Expert Report of Prof. Gronke). And the United States' expert further states that "the issue that [post-2008] research raises is whether the use of early voting in 2008 may ... have been related to a surge in turnout in 2008. So there may be something new afoot." A5497 (Stewart Dep.). Moreover, even if early voting by itself does not affect overall voter turnout, there is evidence that it may do so when combined with other factors, such as GOTV drives. Indeed, "[t]here is a growing literature that takes a more nuanced approach to examining the relationship between convenience voting and turnout, conjecturing that although reforms may not increase overall participation, they may ... have an impact ... when combined with party mobilization efforts." A7926 (Paul Gronke et al., Convenience Voting, 11 ANN. REV. POL. SCI. 437, 443 (2008)); see A8009 (Paul Gronke, Early Voting Reforms & Am. Elections, WM. & MARY BILL OF RTS. J. 423, 432 (2008)) (stating that "[e]nough research has accumulated" for scholars to conclude that early voting "does encourage regular voters to participate in lower intensity contests that they might otherwise skip"); A9094 (Rebuttal Decl. of Prof. Gronke) (stating that, "[s]ince non-habitual voters are less likely to vote, early or on election day, convenience may have a significant and positive effect on their decision to vote" as well). The court finds that "there is not currently a consensus ... that there is no effect of early in-person voting on turnout." A9984 (De Bene Esse Dep. of Prof. Gronke).

62. Moreover, the literature that Florida cites addresses only the question of how adding early voting days affects overall voter turnout. It does not address the specific question before us: how decreasing an established early voting period from 12 days to 8 days (and from 96 hours to only 48) will affect African-American voter turnout. Hence, even if the addition of early voting days does not significantly increase turnout, "it is not methodologically sound to assume that there will ... be little or no impact on overall turnout when voters (who have habituated to early in-person voting) face a loss of previously available voting days." A9095 (Rebuttal Decl. of Prof. Gronke). Indeed, common sense suggests the opposite.

63. Finally, HB 1355 mandates a Sunday of early voting that was not previously offered in any of the covered counties. And given the importance of Sunday early voting for "souls to the polls" drives in the African-American community, see infra ¶ 68, that adjustment can be expected to have some positive effects on the ability of African-Americans to reach the polls. Nonetheless, those 6 hours of Sunday early voting would represent the only ameliorative aspect of the new law if the covered counties were to implement it by choosing the bare minimum number of hours. Indeed, the county would offer 2 fewer hours on each Saturday (6 rather than the previous 8). We find that shifting one of the remaining 8 days to a Sunday would not make up for a 4-day and 48-hour decrease in early voting.

64. In sum, Florida is left with nothing to rebut either the testimony of the defendants' witnesses or the common-sense judgment that a dramatic reduction in the form of voting that is disproportionately used by African-Americans would make it materially more difficult for some minority voters to cast a ballot than under the prior law.

3. Burdens Imposed by Early Voting Changes (at 96 Hours)

65. It is possible, however, that the counties will opt to provide substantially more hours than the minimum of 48 hours permitted by the early voting changes. If the covered counties did offer the maximum of 96 possible hours of early voting, voters would have exactly the same total number of hours for early voting as under the preexisting law: 96 hours. Those hours would simply be distributed over a fewer number of days.76

66. Although voters would not be able to vote during 4 days that were previously available even if the counties offer 96 hours of early voting, they would be able to vote during 4 hours each day that were not. Moreover, if election supervisors utilize a standard 7 a.m. to 7 p.m. voting day, this means that weekday voters would not have to vote during working hours, but could instead get to the polls before or after work — during the morning and evening hours that would be most convenient for weekday voters and that would be more extended than those offered in any recent election. As the supervisor of elections in Hillsborough County stated, "some hours are more accessible to people than other hours," and both "[t]he early morning hours earlier in the day" and hours "in the afternoon ... following work" can be helpful for some voters who "have the opportunity" to vote at those times. A3765 (Lennard Dep.); see A9095 (Rebuttal Decl. of Prof. Gronke) ("Certainly more hours of voting per day can benefit voters, and there are studies ... showing the positive benefits of expanded hours outside of the normal business day."); A10511 (testimony from an election official in a non-covered county that, "when the polls open [at] 7 a.m.," there will generally "be lines that have already formed" with voters who are "waiting for the precinct to open").77

67. The new early voting statute also requires the covered counties to offer early voting for at least six hours on the Sunday nine days before Election Day. See Fla. Stat. § 101.657(d) (2011). Although the old early voting law permitted counties to offer early voting on that Sunday, as well as on the Sunday immediately before Election Day, see Fla. Stat. § 101.657(d) (2010), none of the covered counties ever did so. And if the covered counties offer the maximum number of hours of early voting each day, the result would be an additional 12 hours of Sunday early voting that were never before offered in those counties.

68. The addition of Sunday voting is important. As the defendants stressed, many African-American churches organize "souls to the polls" drives to transport their congregants to early voting sites on the Sunday immediately before Election Day, see A9109 (Joyner Decl.); A9193-94 (Decl. of Rev. Thomas Scott), and that Sunday is therefore disproportionately used by African-American voters in jurisdictions that have early voting on that day, see A10098 (Am. Expert Report of Prof. Gronke); see also DOJ Br. 51; A9225 (Decl. of Rev. Charles McKenzie). The court finds that opening the polls on the previous Sunday would also facilitate voting by African-American voters. Even Senator Joyner, who opposed the reduction in early voting days, testified that if HB 1355 "were precleared and were put into effect in Hillsborough County," she would "[a]bsolutely ... expect the predominantly African-American churches in Hillsborough County to organize Souls to the Polls drives" on that day. A9814 (Joyner Dep.). Likewise, Professor Gronke agreed that he has "no reason to believe that [the additional Sunday] would not be used" by African-American voters in such "souls to the polls" drives. A10012-13 (De Bene Esse Dep. of Prof. Gronke).

69. For a county that offers the maximum 12 hours per day, the new law not only adds the opportunity to vote for 12 hours on the Sunday that falls two weekends before the election, but also increases the overall number of weekend hours on both that weekend and the last weekend before the election. Under the preexisting practice, the covered counties offered early voting for 8 hours on each Saturday, and not at all on Sunday, yielding 16 total weekend hours of early voting. But if a county offers the maximum hours under the new statute, it will provide early voting for 12 hours on each weekend day (i.e., on both Saturday and Sunday) approximately one week before the election, and for 12 hours on the last Saturday before the election. The total would then be 36 weekend hours of early voting (12 hours on each of 3 weekend days), for a net gain of 20 more weekend hours of early voting than under the preexisting law.

70. Like the added Sunday, those additional weekend hours of early voting are significant. Florida election officials have testified, based on their experience, that expanded weekend hours would provide increased accessibility for many minority voters. See, e.g., A3447-48, 3450 (Edwards Dep.); A3611-12, 3770, 3846-47 (Lennard Dep.); A10488 (Dep. of Ion Sancho, Leon County SOE). Florida legislators, including several of the legislators who opposed the early voting changes and who are now intervenors in this action, have concurred in that assessment. See A9801 (Joyner Dep.); A10254-55 (Cruz Dep.). And even the intervenors' expert, Professor Gronke, concedes that expanded early voting hours can be an added convenience for many voters. See A6161 (Gronke Dep.); A9095 (Rebuttal Decl. of Prof. Gronke); A10015 (De Bene Esse Dep. of Prof. Gronke); see also A10187 (EVIC blog post by Professor Gronke, stating that "weekend early voting" is "a potential inconvenience for officials to be sure, but one which citizens will find very helpful").

71. In sum, we find that, if the covered counties offer the maximum available early voting hours each day on a standard 7 a.m. to 7 p.m. schedule, the negative effect of reducing the number of days from 12 to 8 would likely be offset by the ameliorative effects of adding non-working hours, a Sunday, and additional weekend hours.

72. Moreover, many of the concerns that we discussed in connection with a contraction of the early voting period to only 48 total hours are not likely to materialize (certainly not to the same extent) if the full 96 hours of early voting are maintained on a standard 7 a.m. to 7 p.m. schedule. For example, although representatives and volunteers from minority voting rights groups have testified that "a two-week period has ... been essential to coordinating the logistics of GOTV efforts in the African American community," A9237 (Coffee Decl.); see A9227 (McKenzie Decl.), the record evidence suggests that GOTV groups could adjust to a redistribution of the total 96 hours over a different number of days, including weekend days and a "souls-to-the-polls" Sunday. Indeed, one of the central concerns expressed by such groups is that minority voters "frequently contend with issues such as lack of transportation or inflexible work and family schedules, which make it difficult to reach the polls on Election Day." A9143 (Slater Decl.). Expanding convenient non-working weekday voting hours should therefore help third-party efforts to provide transportation to the polls for such voters.

73. The same is true of concerns regarding overcrowding and confusion at polling places on the remaining early voting days (and Election Day) that could result from reducing the total number of early voting days. This problem should be alleviated, if not entirely eliminated, by the simultaneous expansion of early voting hours to 12 hours per day on a 7 a.m. to 7 p.m. schedule. Indeed, election officials have testified that they believe that, "by expanding the hours," they would be able to "accommodate [the] early voters" who may have been displaced by the elimination of the first five days of the old early voting period. A3859 (Lennard Dep.). For the first time, early voters in the five covered counties would have the benefit of an early voting period capturing the morning and evening hours on both ends of the standard 9 a.m. to 5 p.m. workday. One supervisor from a non-covered county did say that he did not think voters would adjust to the expanded early voting hours because peak early voting times are not typically in the early morning or after dark. See A9179 (Sancho Decl.). We find that this observation is of little import, however, because early morning and evening voting hours were never offered under the preexisting early voting plan. And even that supervisor acknowledged that "[u]sually, we have a line when the polls open [on Election Day at] 7 a.m." A10511 (Sancho Dep.).

74. For the foregoing reasons, we find that if the covered counties were to offer early voting for 12 hours per day, from 7 a.m. to 7 p.m. over an 8-day early voting period, including one previously-unavailable Sunday, they would likely satisfy their burden of proving that the overall effect of the early voting changes would not materially burden minority voters' effective exercise of the franchise.

B. Inter-County Mover Changes

1. Voters Affected by Inter-County Mover Changes

75. We find that, because minority voters were disproportionately more likely to use prior procedures that allowed voters to make inter-county address changes at the polls and then cast a regular ballot, the inter-county mover changes will disproportionately affect minority voters. See A8901 (Supplemental Decl. of Dr. Stewart) ("[M]inority voters are more likely to avail themselves of the law that is still in effect in the five covered counties, which allows registered voters who have moved between counties to simultaneously change their address and vote a regular ballot in their new home county.").

76. The relevant inter-county mover data was compiled and analyzed by the expert witness for the United States, Dr. Charles Stewart III. It is derived from the results of a database query regarding: (a) early voters who moved into a covered county and changed their address during the early voting period; and (b) voters who moved into a covered county and changed their address on or shortly after Election Day. See A8896 (Supplemental Decl. of Dr. Stewart); see also A9035 (E-mail from Wren Fowler, VR Systems (Feb. 16, 2012)).

77. In reaching conclusions based on this data, Dr. Stewart was required to make certain assumptions about whether the voters in the data set had actually changed their addresses at the polls, rather than contacting election officials and updating their addresses before going to vote. See A10648-52 (De Bene Esse Dep. of Dr. Stewart). As such, it is possible that the data may over-estimate the numbers of inter-county movers to some degree. The data set may also be over-inclusive because the results of the database query include some voters whose addresses were changed in the voter registration database shortly after Election Day. See A8895-96 (Supplemental Decl. of Dr. Stewart); A9035.

78. Dr. Stewart reasonably assumed that many of those voters changed their addresses on or slightly before Election Day, and that election officials simply did not update the database until after the election. See id. at A10653-54. Ultimately, therefore, we are persuaded that although the data set may not be perfect, it provides an "apt proxy for the group of voters that are affected by the intercounty mover changes." Id. at A10655-56. Even if the total estimated number of inter-county movers is slightly overstated, there is no reason to believe that the racial proportions of those voters would be affected in any way.

79. The data demonstrates that in the most recent elections in Florida's covered counties for which there is data, the voters who moved inter-county and updated their addresses on or about Election Day were disproportionately likely to be members of a minority group. Id. at A8907-09 (Attach. 5). More specifically, African-American and Hispanic voters were statistically more likely than the average voter to be inter-county movers in the 2008 and 2010 elections, both on Election Day and during the early voting period, while white voters were statistically less likely to be inter-county movers. Id.

80. For example, in the 2008 general election, African-American voters comprised only 13.3% of the non-moving voters in Florida's covered counties, but made up 19.3% of all inter-county movers in those counties. Id. at A8908. Hispanic voters comprised only 10.1% of the non-moving voters, but accounted for 14.5% of the inter-county movers. Id. Meanwhile, white voters made up 71.3% of the non-moving voters, but only 58.8% of the inter-county movers. Id.

81. The following table, which appears as part of Attachment 5 in Dr. Stewart's supplemental declaration, shows the percentages of both moving and non-moving voters who were members of each of three racial groups (African-American, Hispanic, and white) in the five elections that took place between 2008 and 2010 in Florida's five covered counties. Id. at 9808-09; see also id. at 8898-89 & n. 18 (explaining the table).

Election                             Black   Hispanic   White    Total

2008 Pres.     Non-moving voters      2.1%    3.9%       92.1%    90,123
Preference     Inter-county movers    5.3%   14.9%       75.5%        94
Primary*       Difference             3.2%   11.0%      -16.6%

2008 Primary   Non-moving voters     11.7%    4.7%       81.1%    82,585
               Inter-county movers   21.9%    9.4%       64.6%        96
               Difference            10.2%    4.7%      -16.5%

2008 General   Non-moving voters     13.3%   10.1%       71.3%   530,409
               Inter-county movers   19.3%   14.5%       58.8%     2,240
               Difference             6.0%    4.4%      -12.5%

2010 Primary   Non-moving voters      8.7%    4.2%       84.7%   135,451
               Inter-county movers    9.5%    8.8%       78.8%       137
               Difference             0.8%    4.6%       -5.9%

2010 General   Non-moving voters     10.9%    7.1%       78.3%   355,398
               Inter-county movers   19.1%   16.0%       61.8%       613

               Difference             8.2%    8.9%      -16.5%

* Data from Hillsborough County is not available for this particular election (the 2008
Presidential preference primary)

82. In examining this table, Dr. Stewart found that "[i]n every single calculation, the percentage of black and Hispanic voters among the `movers' is greater than the percentage of black and Hispanic voters among the non-moving voters." Id. at A8899. By contrast, "[i]n every calculation [in the table reproduced above], the percentage of white voters among the movers is less than the percentage of white voters among the non-moving voters." Id.78

83. These disparities are statistically significant, with the exception of the 2010 primary election, which had very low overall voter turnout. See A8900-01 (Supplemental Decl. of Dr. Stewart); see also id. at A8910 (Attach. 6, showing results of chi-squared tests of statistical significance)

84. Thus, we credit Dr. Stewart's conclusions that "[i]t's clear to me that blacks, whites, and Hispanics avail themselves of [the inter-county mover] procedures at different rates than the underlying voting population," A10667 (De Bene Esse Dep. of Dr. Stewart), and that "minority voters [African-Americans and Hispanics] are more likely [than white voters] to avail themselves" of the inter-county mover procedures, A8901 (Supplemental Decl. of Dr. Stewart).

85. Because Dr. Stewart's tables demonstrate this phenomenon in a rather oblique way, we have also used his data to calculate the percentages of voters within each racial group (African-Americans, Hispanics, and whites) in the five covered counties who were inter-county movers in each of the elections that Dr. Stewart analyzed. The results of those calculations are displayed in the following table:

Election                                              Black    Hispanic   White

2008 Pres.            # of voters of given race        1,893   3,513      83,003
Preference Primary*   # of inter-county movers of      5       14         71
                      given race
                      % of voters of given race who    0.26%   0.40%      0.09%
                      were inter-county movers

2008 Primary          # of voters of given race        9,662   3,881      66,976
                      # of inter-county movers of      21      9          62
                      given race
                      % of voters of given race who   0.22%    0.23%      0.09%
                      were inter-county movers

2008 General          # of voters of given race       70,544   53,571     378,182

                      # of inter-county movers of     432      325          1,317
                      given race
                      % of voters of given race who   0.61%    0.61%        0.35%
                      were inter-county movers

2010 Primary          # of voters of given race       11,784   5,689        114,727
                      # of inter-county movers of     13       12           108
                      given race
                      % of voters of given race who   0.11%    0.21%        0.09%
                      were inter-county movers

2010 General          # of voters of given race       38,738   25,233       278,277
                      # of inter-county movers of     117      98           379
                      given race
                      % of voters of given race who   0.30%    0.39%        0.14%
                      were inter-county movers

* Data from Hillsborough County is not available for this particular election (the 2008
Presidential preference primary)

86. This table was generated by performing the following calculations with Dr. Stewart's data. First, we determined the total number of voters of a given race who voted in each election from 2008 to 2010 ("# of voters of given race") by multiplying the percentage of non-moving voters of a given race who voted in each election (i.e., the last column in Dr. Stewart's table) by the total number of non-moving voters in that election; then multiplying the percentage of inter-county movers of the same race by the total number of inter-county movers; and then adding together the two resulting numbers. Next, we determined the total number of inter-county movers in each racial group ("# of inter-county movers of given race") by multiplying the percentage of inter-county movers of that race by the total number of inter-county movers. Finally, we generated the percentage of voters of each given race who were inter-county movers in that particular election ("% of voters of given race who were inter-county movers") by dividing the first figure by the second.

87. We combined this data further in order to produce an aggregate percentage of voters of each race who were inter-county movers in the covered counties from 2008 to 2010. First, we added together the numbers of voters in a given race in each of the five elections in the previous table. Then, we added the number of inter-county movers of each race in each of these elections. Last, we divided the first number by the second to generate the percentage of inter-county movers of that race. That data is displayed in the table below:

Grand Totals                               Black     Hispanic   White

# of voters                                132,622   91,890     921,165

# of inter-county movers                   588       458        1,937

% of voters who were inter-county movers   0.44%     0.50%      0.21%

88. Based on this table, we find that, on average, minority voters in Florida's covered counties are approximately twice as likely as white voters in those same counties to move inter-county and seek to update their addresses at the polls. Specifically, from 2008 through 2010, 0.44% of African-American voters and 0.50% of Hispanic voters in the covered counties were inter-county movers, compared to only 0.21% of white voters.

89. The more limited data examined by Florida's own expert witness, Professor M.V. (Trey) Hood III, confirms the conclusion that, as a percentage of their overall demographic population, African-American and Hispanic voters are more likely than white voters to seek to change their addresses at the polls. Professor Hood's data set is less complete than Dr. Stewart's, because he examined data from only three of Florida's five covered counties — excluding Hillsborough County, the most populous of the five — and two of those counties (Hardee County and Hendry County) claimed not to even have any inter-county movers in the election years that he analyzed. See A9073 (Expert Report of Prof. Hood); see also A5954 (Hood Dep.) (admitting that his data is only "[a] quarter complete"). Nonetheless, Professor Hood's conclusions do not differ from those of Dr. Stewart.

90. The following chart reproduces Table 4.4 in Professor Hood's expert report, which displays "Out-of-County Change-of-Address Affirmations by Race/Ethnicity as a Percentage of Total In-Person Turnout by Racial/Ethnic Group," A9073 (Expert Report of Prof. Hood):

Election/County             White    Black    Hispanic   Other

Hardee County
2010 Primary                0%       0%       0%         0%
2010 General                0%       0%       0%         0%

Hendry County
2008 General                0%       0%       0%         0%
2010 General                0%       0%       0%         0%

Collier County
2006 Primary                0.019%   0%       0%         0%
2006 General                0.052%   0.260%   0.161%     0.204%
2008 Presidential Primary   0.049%   0.600%   0.088%     0.155%
2008 Primary                0.023%   0%       0.100%     0.238%
2008 General                0.125%   0.394%   0.326%     0.192%
2010 General                0.103%   0.374%   0.290%     0.288%

91. As the table above shows, and as Professor Hood acknowledges, in each of the elections that Professor Hood examined, the inter-county mover rate of "some minority group [whether African-American or Hispanic] is higher than the white rate[,] and in most of [the elections] all of the minority rates are higher than the white rate." A5958-59 (Hood Dep.) (emphasis added).

92. Notwithstanding that racial minorities in the covered counties will be disproportionately affected by changes to Florida's inter-county mover laws, the total number of inter-county movers has been quite small historically, and the number who are minorities has been even smaller. The high-water mark for inter-county movers was the 2008 general election, when 2,240 voters (approximately 750 of whom were minorities) updated their addresses at the polls in the covered counties. See A8908 (Supplemental Decl. of Dr. Stewart). This is out of a total of over 530,400 voters who voted in that election in the covered counties. See id.79

2. Burdens Imposed by Inter-County Mover Changes (at 48 Hours)

93. We find that, despite their disproportionate effect on minority voters, the burdens imposed by the inter-county mover changes are not likely to dissuade reasonable minority voters in Florida's five covered counties from casting a ballot, or to result in those minority voters' ballots not being properly tabulated.

94. Under the prior law, a voter who moved to another county without notifying the relevant election official and then sought to vote in his or her new county of residence was required to complete a change of address affirmation in "substantially the [same] form" set forth in Fla. Stat. § 101.045(2)(a) (2010), which was as follows:

Fla. Stat. § 101.045(2)(a); see, e.g., [129] Fla. Notice of Filing of Documents Re: Inter-County Address Changes at the Polls, Ex. A (Collier County change of address affirmation form).80 After completing the form, the voter was then entitled to cast a regular ballot.

95. Under the new law, inter-county movers must instead complete a "Provisional Ballot Voter's Certificate and Affirmation," as set forth in Fla. Stat. § 101.048(3) (2011). The new form must be "in substantially the [same] form" as the following:

Fla. Stat. § 101.048(3) (2011); see, e.g., [129] Fla. Notice of Filing of Documents Re: Inter-County Address Changes at the Polls, Ex. D (Collier County Provisional Ballot Certificate and Affirmation).81 After completing the affirmation, the inter-county mover is then entitled to cast a provisional ballot. Fla. Stat. § 101.048(1)-(2) (2011).

96. The form required under the new law requests essentially the same information as the prior form, including the voter's name, current and prior address of residence, and an affirmation that the voter is legally registered and qualified to vote. See A4068 (Dep. of Maria Matthews, Assistant Gen. Counsel, Fla. Dep't of State) ("[R]eally [inter-county movers are] not filling out ... anymore information or any additional documentation than they would have if they had moved in from another county previously."). Under both the old law and the new one, Florida voters were required to disclose either their Florida driver's license number or the last four digits of their social security number when they registered to vote. See [129] Florida's Notice of Filing of Documents Related to Inter-County Address Changes at the Polls, Ex. B (attaching a copy of the Florida Voter Registration Application, which was unchanged in HB 1355).82

97. A provisional ballot is in all physical respects identical to a regular ballot. Fla. Stat. § 97.021(29) (defining "provisional ballot" to simply mean "a conditional ballot"); Fla. Admin. R. 1 S-2.037(2)-(3) (providing guidelines for provisional ballot certificates and affirmations); see Oral Arg. Tr. 237:19-238:24 (representation by counsel for Florida that provisional ballots look the same as regular ballots, and that "[a]ll Florida voters vote by paper ballot"). The only difference — from a procedural point of view — is that instead of being fed into a machine for tabulation, a provisional ballot is placed into a secrecy envelope, which is then sealed in a provisional ballot envelope. Fla. Stat. § 101.048(1)-(2) (2011). The ballot is then deposited in a ballot box, and the secrecy envelope remains sealed until the county canvassing board reviews the certificate and affirmation to determine whether the ballot may be tabulated. See id.

98. The new provisional voter's certificate and affirmation will not take substantially longer to complete than the old affirmation of change of address. See A3938 (Ussery Dep.) ("It's the same ballot they would have received.... I don't really see where it would be an obstacle [to voting]."); A3261-62 (Sawyer Dep.); A3452-53 (Edwards Dep.); A3830 (Lennard Dep.). The total paperwork might take a few additional minutes because the voter may have to include some of the same information on the provisional ballot envelope. See Oral Arg. Tr. 237:11-237:16. The voter might also have to wait while an election official reviews and signs his or her affirmation and/or provisional envelope. See Fla. Stat. § 101.048(3) (2011). Some of the covered counties have sought to minimize these burdens by affixing the Provisional Ballot Voter's Certificate and Affirmation to the outside of the provisional envelope, or printing the certificate and affirmation directly onto the provisional envelope itself. See, e.g., A7395 (Hillsborough County Provisional Ballot Quick Reference Guide).

99. There will be little difference in the time required to cast an inter-county mover ballot under the new versus the prior law. Some elements of the process may be slightly more time-consuming. Provisional voters may be required to move into a different line, where they will be required to interact with a particular election worker with the authority to process provisional ballots, see, e.g., A10583-84 (Sancho Dep.); A9183 (Sancho Decl.). Voters faced with the requirement of casting a provisional ballot may also have questions for election officials, cf. A9105 (Joyner Decl.), who will in turn be required to communicate certain information to provisional voters, see Flat. Stat. § 101.048(5). However, some aspects of the new inter-county mover process may move more quickly than the old procedures. Election officials will no longer be required to verify that inter-county movers were validly registered Florida voters before permitting them to cast a regular ballot, see Fla. Stat. § 101.045(d) (2010), but will instead determine the inter-county movers' eligibility when reviewing their provisional ballots after Election Day, see Fla. Stat. § 101.048(2)(a) (2011).

100. We find that reasonable inter-county movers are unlikely to be so deterred by the experience of casting a provisional ballot that they will decide not to vote at all. Election officials in the covered counties have testified that they will explain to inter-county movers that their ballots "shall be counted" as long as the voter is eligible. Fla. Stat. § 101.048(2)(a) (2011). They have also testified that "if the voter is there to vote and they still have the opportunity to vote, most likely they are going to take advantage of it," A3972 (Ussery Dep.). These officials have never seen a voter leave a precinct without voting simply because he or she was required to vote a provisional ballot. A3972 (Ussery Dep.); A3778-79 (Lennard Dep.); but see A10583-84 (Sancho Dep.). The one contrary anecdote in the record involved a substantially different situation, in which a voter was told that there was "something wrong" and that he had to "come back in 48 hours and ... clear it up." A9794-95 (Joyner Dep.). We therefore accord that anecdote little weight.

101. The inter-county mover changes will not materially increase lines at polling places. The new procedures for inter-county movers will not take much more time than prior procedures. See ¶¶ 98-99, supra. Moreover, there are so few inter-county movers in the covered counties that it is unlikely that any individual precinct would be affected to any material degree. See A8908-09 (Supplemental Decl. of Dr. Stewart). Only 1,000 of approximately 2,240 movers in the 2008 general election updated their addresses on election day; the rest were spread over the early voting period. A8907-08. This total averages out to only a handful of inter-county movers at each precinct. See Oral Arg. Tr. 34:22-35:02. Even these numbers may overestimate the number of inter-county movers in future elections, as the data set on which they are based may include voters whose addresses were changed before or after election day, see A8895-96 (Supplemental Decl. of Dr. Stewart); A9035, military personnel and members of their families exempt from the new inter-county mover procedures, see Fla. Stat. § 101.045(2)(b) (2011), and voters who will no longer seek their addresses at the polls under the new law, given that it makes it considerably easier for voters to update their addresses before going to the polls to vote, see Fla. Stat. § 97.1031(1)(b) (2011) (allowing registered voters to update their addresses electronically or by telephone).

102. While provisional ballots may have been counted at lower rates than regular ballots in the past, see A7381 ("New Election Law Talking Points" for Hillsborough County); A3393-94 (Edwards Dep.), none of these ballots involved inter-county movers, and the United States has proffered no evidence that invalidation was due to human error or malevolence. The record evidence suggests that, most often, such ballots were not counted "simply because the person [was] not eligible to vote," A5048 (Stafford Dep.); see also A8156 (Decl. of Gisela Salas, Fla. Director of Elections), a circumstance that would not apply to inter-county movers. Inter-county mover ballots "shall be counted" if the voter is eligible. Fla. Stat. § 101.048(2)(a) (2011); id. § 101.048(2)(b)(1). Florida has instructed its supervisors of elections accordingly. See A2764-65 (Fla. Dep't of State, Directive 2011-01 (May 19, 2011)); A7232-33 (Fla. Dep't of State, Directive 2012-01 — Provisional Ballot Verification (Jan. 13, 2012)).

103. During legislative debates, the Florida Association of Supervisors of Elections ("FSASE") expressed concerns that "there's simply not enough time to canvass" the additional provisional ballots that will be cast under the new inter-county mover procedures. A967-69 (Senate Rules Comm. (Apr. 15, 2011) (Statement of David Stafford, FSASE President)); see also A9180-81 (Sancho Dep.). But supervisors of elections in the covered counties have testified that they will "train and put on additional personnel to ensure that [the new provisional ballot procedures] will work." A3781-82 (Lennard Dep.); see also A5044-45 (Stafford Dep.). As noted above, the numbers of inter-county movers in the covered counties are so small as to make it unlikely that their provisional ballots will overwhelm the county canvassing boards.

104. Election officials in Florida must count the provisional ballots of inter-county movers even if the officials run out of time to examine them in detail. See A7100-01 (Fla. Rule 30(b)(6) Dep.); A4467-68 (Dep. of Gary Holland, Assistant Gen. Counsel, Fla. Dep't of State). The accompanying opinion is based on the understanding that Florida will follow its laws as written, see Fla. State. § 101.048(2) (2011), and will abide by the representations it has made to this court to that effect, see Fla. Br. 19; Oral Arg. Tr. 33:05-34:20. Most of the elections supervisors in the covered counties appear to already understand the requirements of the new law, see A3207-08 (Lennard Dep.); A3265, 3269-71 (Sawyer Dep.).

105. Finally, in some respects the new inter-county mover changes will make it considerably easier for inter-county movers to vote. Voters will be able to update their addresses "via telephone or electronic means," or by submitting any "signed, written notice," see Fla. Stat. § 97.1031(1)(b) (2011), rather than only via "a voter registration application signed by the elector," Fla Stat. § 97.1031(2) (2010). As a result, there may be fewer voters who wait until they arrive at the polling place to change their addresses.

106. For the foregoing reasons, we find that the inter-county mover changes will not materially burden minority voters in Florida's five covered counties in their effective exercise of the franchise.

III. Purpose of the 2011 Inter-County Mover Amendments83

107. The legislative proponents of these changes stated that they were necessary to prevent a possible form of "double voting" voter fraud, whereby voters might fraudulently or negligently cast ballots in more than one county (e.g., their old and new counties of residence). See, e.g., A463 (House State Affairs Comm. (Apr. 1, 2011)); A1 155-56 (House Sess. HB 1355 1st Reading & Debate (Apr. 20, 2011)); A1351-52, 1403-06, 1417-18 (House Sess. HB 1355 2d Reading & Debate (Apr. 21, 2011)). The sponsors of those amendments also expressed a desire to promote the efficiency and integrity of the electoral process by encouraging voters to update their addresses before going to the polls, so that the voter rolls could be finalized before Election Day. See A713-16, 722 (House State Affairs Comm. (Apr. 14, 2011)); A948 (Senate Rules Comm. (Apr. 15, 2011)); A1 196-97, 1204 (House Sess. HB 1355 1st Reading & Debate (Apr. 20, 2011)); see also A386-87, 454-55, 463 (House State Affairs Comm. (Apr. 1, 2011)); A688, 722-24 (House State Affairs Comm. (Apr. 14, 2011)); A947-48, 956-57, 984-85 (Senate Rules Comm. (Apr. 15, 2011)); A1 149-51, 1167-68, 1196-97 (House Sess. HB 1355 1st Reading & Debate (Apr. 20, 2011)); A1351-52, 1403-06 (House Sess. HB 1355 2d Reading & Debate (Apr. 21, 2011)); A1586-87, A1595-1602, 1612-13 (Senate Budget Comm. (Apr. 26, 2011)); A1822-24, 1870 (Senate Sess. — SB 2086/HB 1355 2d Reading & Debate (May 4, 2011)); A2215, 2255-57 (Senate Floor Debate (May 5, 2011)).

108. The inter-county mover changes will not "bear more heavily" on any minority group. Although the inter-county mover changes may affect disproportionately more minority voters than white voters because minorities are more likely to be inter-county movers, the evidence indicates that the changes will not have materially adverse effects on the ability of minority voters to cast a ballot and effectively exercise the electoral franchise.

109. We must also consider whether there are contemporaneous statements that indicate a discriminatory purpose. On this point, the defendants point to a floor statement by State Senator Mike Bennett, who said that he did not want to make it easier for people to vote, but rather that it should be harder to vote — as it is "in Africa." A2242 (Senate Floor Debate (May 5, 2011)). The relevant text of Senator Bennett's statement is as follows:

A2242 (Senate Floor Debate (May 5, 2011) (Statement of Sen. Bennett)).

110. Whether or not Senator Bennett actually intended his statement to have racial undertones, it certainly can be read that way. Nonetheless, Senator Bennett's is the only statement to which the defendants point as evidencing a discriminatory purpose on the part of the Florida legislature. Senator Bennett was neither a sponsor nor a primary proponent of HB 1355, and did not play an important role in passage of the bill. Indeed, the floor statement was Senator Bennett's only public statement regarding the 2011 voting amendments. And although Senator Bennett was the Senate President Pro Tempore at the time, that office has limited authority, and there is no evidence it played any role in passage of the amendments. Under the rules of the Florida Senate, "the President Pro Tempore shall assume the duties of the chair" "[i]f for any reason the President is absent and fails to name a Senator" to carry out those duties. Rules and Manual of the Senate of the State of Florida, Rule 1.7(2) (adopted Nov. 16, 2010). And "[i]n the event of extended absence of the President or the President's disability or incapacity, the President Pro Tempore shall assume the duty of referring bills." Id., Rule 4.6(2).

111. Florida has conceded that "[a]t the time of the Florida Legislature's consideration of HB 1355, the State of Florida knew of no instance" of such "double voting" by inter-county movers. A8281 (Fla. Resp. to Req. for Admiss. No. 12). Nor has it identified any instances since then. See Oral Arg. Tr. 24:17-25:2, 27:3-27:20. Likewise, election officials have confirmed that they have never seen double voting by inter-county movers at their precincts. See, e, g., A3648-39, 3771-72 (Dep. of Earl Lennard, Hillsborough County SOE); A3106-07, 3132, 3198-3200 (Dep. of Harry Sawyer, Monroe County SOE); A3937 (Dep. of Jeffery Ussery, Hardee County SOE).

112. The inter-county mover changes also make ameliorative adjustments that will make it easier for inter-county movers to vote. Unlike the benchmark law, the new law allows voters to update their addresses before Election Day by simply "[c]ontacting the supervisor of elections via telephone or electronic means." Fla. Stat. § 97.1031(1)(b) (2011).

113. The bills that ultimately became HB 1355 were amended several times during the legislative process, often by means of "strike-all" amendments. See A120 (Revised Jointly Stip. Facts ¶ 27); A683-90; A942-50 (Senate Rules Comm. (Apr. 15, 2011)); A1586-92 (Senate Budget Comm. (Apr. 26, 2011)). But several witnesses who are familiar with the standard procedures in the Florida legislature testified that such strike-all amendments are not necessarily unusual in Florida. See A4759, 4776, 4952-54 (Dep. of Pierce Schuessler, Legislative Affairs Director, Fla. Dep't of State); A6766-67; A9828-33 (Dep. of Sen. Arthenia Joyner). In addition, the record of the debates on HB 1355 reveals that the legislative process unfolded over a period of several months, and involved substantial testimony from a number of Florida election officials and members of the general public. See, e.g., A417-45 (House State Affairs Comm. (Apr. 1, 2011)); A740-71 (House State Affairs Comm. (Apr. 14, 2011)); A964-99 (Senate Rules Comm. (Apr. 15, 2011)).

114. The record contains several examples of Florida voting laws passed in recent years that were enacted to take immediate effect. See A10699-970 (collecting examples); see also A3869 (Lennard Dep.); A4918 (Schuessler Dep.).