DECISION AND ORDER

LYNN ADELMAN, District Judge.

In May 2011, the Wisconsin Legislature passed 2011 Wisconsin Act 23 ("Act 23"), which requires Wisconsin residents to present a document including photo identification ("photo ID") in order to vote. 2011 Wis. Sess. Laws 104 (codified as amended in scattered sections of Wis. Stat. Ch. 5 and 6).1 The plaintiffs in the two cases captioned above claim the law violates the Fourteenth Amendment and/or Section 2 of the Voting Rights Act, 42 U.S.C. § 1973.

In the Frank case, individuals who are eligible to vote in Wisconsin contend that Act 23 violates both the Fourteenth Amendment and Section 2 of the Voting Rights Act. In the LULAC case, four organizations argue that Act 23 violates Section 2 of the Voting Rights Act. With the agreement of the parties, I handled the cases together without formally consolidating them and, in November 2013, conducted a two week trial to the court. In this decision, which constitutes my findings and conclusions under Federal Rule of Civil Procedure 52, I address the major issues presented. In an effort to make the opinion as readable as possible, I have placed several relatively technical discussions of expert testimony in appendices rather than in the text.

Before proceeding, I note that I am only addressing two of the plaintiffs' claims — the Frank plaintiffs' claim that Act 23 places an unjustified burden on the right to vote and the claim of both the Frank and LULAC plaintiffs that Act 23 violates Section 2 of the Voting Rights Act. I do not address the Frank plaintiffs' remaining claims, which are all constitutional claims. My reason for not addressing the remaining claims is based on the "longstanding principle of judicial restraint" under which courts are to "avoid reaching constitutional questions in advance of the necessity of deciding them." Camreta v. Greene, ___ U.S. ___, 131 S.Ct. 2020, 2031, 179 L.Ed.2d 1118 (2011) (internal quotation marks omitted). As explained below, all of the plaintiffs are entitled to permanent injunctive relief against enforcement of the photo ID requirement on the ground that the requirement violates Section 2 of the Voting Rights Act. This makes consideration of any of the Frank plaintiffs' constitutional claims unnecessary. Still, I believe it is wise to consider the constitutional claim of whether Act 23 places an unjustified burden on the right to vote. As my analysis below will demonstrate, the Section 2 statutory claim and the unjustified-burden constitutional claim overlap substantially, in that many factual findings are relevant to both claims. Indeed, the Section 2 analysis is largely identical to the unjustified-burden analysis, except that the Section 2 analysis involves the additional question of whether Act 23 has a disproportionate impact on Blacks and Latinos and produces a "discriminatory result."2 Thus, it would likely not be a wise use of judicial resources to address the Section 2 claim but leave the unjustified-burden claim unresolved. Addressing only the former claim could result in an appeal and then a remand to this court for consideration of the constitutional claim, and then a second appeal involving only the constitutional claim. Of course, by not addressing all constitutional claims, I am leaving the door open to successive appeals. But unlike the unjustified-burden constitutional claim, the remaining constitutional claims do not overlap substantially with the Section 2 claim and could more easily be addressed in separate proceedings.

My analysis proceeds as follows. First, I give an overview of the relevant provisions of Act 23. Second, I address the Frank plaintiffs' claim that Act 23 violates the Fourteenth Amendment because it imposes substantial burdens on the many eligible voters who do not currently possess photo IDs, and because such burdens are not justified by the state interests that Act 23 purports to serve. Third, I address the plaintiffs' claim that Act 23 violates Section 2 of the Voting Rights Act because it has a disproportionate impact on the voting rights of Blacks and Latinos. Finally, I briefly address some remaining procedural matters, namely, the Frank plaintiffs' motion for class certification and the defendants' motion to dismiss the claims of certain Frank plaintiffs.

I. Overview of Act 23

Under Act 23, in order to vote, a person must present one of nine forms of photo ID to prove his or her identity.3 An acceptable photo ID includes one of the following that is unexpired or that expired after the most recent general election:4 (1) a Wisconsin driver's license, (2) a Wisconsin state ID card, (3) an ID card issued by a United States uniformed service, or (4) a United States passport. Wis. Stat. § 5.02(6m)(a). A person may also present: (5) a naturalization certificate issued within the last two years, (6) an unexpired receipt issued when a person applies for a Wisconsin driver's license, which is valid for 60 days as a temporary license, (7) an unexpired receipt issued when a person applies for a state ID card, which is valid for 60 days as a temporary ID card, (8) an unexpired ID card issued by a federally recognized Indian tribe in Wisconsin or (9) an unexpired ID card issued by an accredited Wisconsin university or college that contains the date of issuance, the person's signature and an expiration date no later than two years from the date of issuance. Wis. Stat. § 5.02(6m)(b) — (f). If a person presents a student ID, the person must also produce a document showing that he or she is currently enrolled. Wis. Stat. § 5.02(6m)(f).

Act 23 does not allow an individual to use a Veteran's ID Card, the photo ID that the United States Department of Veterans' Affairs issues when veterans leave the military. Trial Transcript ("Tr.") 871. An individual also cannot use an ID from one of Wisconsin's 16 two-year technical colleges. The Wisconsin Government Accountability Board ("GAB"), a non-partisan board consisting of six retired judges which administers Wisconsin elections, found that technical college IDs which met the requirements set out for student IDs were acceptable, but a legislative committee required the GAB to promulgate an administrative rule on the matter. The GAB did so, but both the legislative committee and the Governor must approve the rule and neither has done so. Tr. 879-80, 883.

When voting in-person, an individual must state his or her name and address and produce one of the accepted forms of photo ID. The clerk or poll worker will then check the poll list to determine if there is a registered voter with matching information and inspect the ID to see if the name on it conforms to the name on the poll list and the photograph reasonably resembles the individual. Wis. Stat. § 6.79(2)(a). If these requirements are met, the individual will be allowed to sign the poll book and receive a ballot. If an individual does not have a qualifying ID, he or she may cast a provisional ballot. However, such ballot will be counted only if the individual appears at the municipal clerk's office with an acceptable ID by 4:00 p.m. on the Friday after the election. Wis. Stat. §§ 6.79(3)(b), 6.97(3)(b). Individuals requesting absentee ballots must also present photo IDs. Wis. Stat. §§ 6.86(1)(ar), 6.87(1). A requester must mail in a photocopy of an acceptable photo ID with his or her request. Wis. Stat. § 6.87(1).

The statute provides limited exceptions. The photo ID requirement does not apply to: (1) absentee voters who have previously supplied acceptable photo IDs and whose names and addresses have not changed, Wis. Stat. § 6.87(4)(b)3, (2) absentee voters who are in the military or overseas, Wis. Stat. § 6.87(1), (3) voters who have confidential listings as a result of domestic abuse, sexual assault or stalking, Wis. Stat. § 6.79(6), (4) voters who have surrendered their driver's licenses due to a citation or notice of intent to revoke or suspend the license who present a copy of the citation or notice, Wis. Stat. § 6.79(7), and (5) absentee voters who are elderly, infirm or disabled and indefinitely confined to their homes or certain care facilities, Wis. Stat. §§ 6.86(2), 6.875. Additionally, an individual with a religious objection to being photographed can apply for a Wisconsin state ID card that does not include a photo. Wis. Stat. § 343.50(4g).

Individuals who lack a qualifying photo ID can apply for a Wisconsin state ID card at the Wisconsin Department of Motor Vehicles ("DMV"). The cost for such a card is normally $18.00, but Act 23 requires the DMV to waive the fee if the applicant is a citizen who will be at least 18 on the date of the next election, and the applicant asks that the card be issued without charge for voting purposes. Wis. Stat. § 343.50(5)(a)3. To obtain a state ID card, a person must obtain certain primary identification documents and appear at a DMV service center to submit an application and be photographed.

II. Fourteenth Amendment Claim: Unjustified Burden on the Right to Vote

The Frank plaintiffs are eligible Wisconsin voters who claim that Act 23's photo ID requirement violates the Fourteenth Amendment because it imposes an unjustified burden on their right to vote. The Constitution does not expressly provide a right to vote, but it does so implicitly. Harper v. Va. State Bd. of Elections, 383 U.S. 663, 665-66, 86 S.Ct. 1079, 16 L.Ed.2d 169 (1966); Reynolds v. Sims, 377 U.S. 533, 554-55, 84 S.Ct. 1362, 12 L.Ed.2d 506 (1964); Yick Wo v. Hopkins, 118 U.S. 356, 370, 6 S.Ct. 1064, 30 L.Ed. 220 (1886) (noting that the right to vote is "a fundamental political right, because preservative of all rights"). Further, the right to vote is a fundamental right protected by both the due process and equal protection clauses of the Fourteenth Amendment. Burdick v. Takushi, 504 U.S. 428, 433, 112 S.Ct. 2059, 119 L.Ed.2d 245 (1992) ("It is beyond cavil that `voting is of the most fundamental significance under our constitutional structure.'" (quoting Ill. Bd. of Elections v. Socialist Workers Party, 440 U.S. 173, 184, 99 S.Ct. 983, 59 L.Ed.2d 230 (1979))); Anderson v. Celebrezze, 460 U.S. 780, 787, 103 S.Ct. 1564, 75 L.Ed.2d 547 (1983) (the right to vote is one of the liberty interests protected by the due process clause); Harper, 383 U.S. at 665, 86 S.Ct. 1079 ("[O]nce the franchise is granted to the electorate, lines may not be drawn which are inconsistent with the Equal Protection Clause of the Fourteenth Amendment."). Thus, states may not enact laws that unduly burden the right to vote. No litmus test, however, neatly separates valid and invalid election laws. Crawford v. Marion Cnty. Election Bd., 553 U.S. 181, 189-90, 128 S.Ct. 1610, 170 L.Ed.2d 574 (2008). Rather, the Supreme Court has adopted a balancing test that courts must apply on a case-by-case basis. Id.

The test adopted by the Court recognizes that, "as a practical matter, there must be substantial regulation of elections if they are to be fair and honest and if some sort of order, rather than chaos, is to accompany the democratic process." Storer v. Brown, 415 U.S. 724, 730, 94 S.Ct. 1274, 39 L.Ed.2d 714 (1974). It further recognizes that an election regulation, "whether it governs the registration and qualification of voters ... or the voting process itself, inevitably affects — at least to some degree — the individual's right to vote and his right to associate with others for political ends." Anderson, 460 U.S. at 788, 103 S.Ct. 1564. Thus, courts applying the balancing test must weigh "`the character and magnitude of the asserted injury'" to the right to vote against "`the precise interests put forward by the State as justifications for the burden imposed by its rule,' taking into consideration `the extent to which those interests make it necessary to burden the plaintiff's rights.'" Burdick, 504 U.S. at 434, 112 S.Ct. 2059 (quoting Anderson, 460 U.S. at 789, 103 S.Ct. 1564). The rigor of the inquiry into the state's interests depends on the extent to which the challenged election law burdens the right to vote. Id. Even very slight burdens "must be justified by relevant and legitimate state interests `sufficiently weighty to justify the limitation.'" Crawford, 553 U.S. at 191, 128 S.Ct. 1610 (quoting Norman v. Reed, 502 U.S. 279, 288-89, 112 S.Ct. 698, 116 L.Ed.2d 711 (1992)).

In Crawford, the Supreme Court considered a claim similar to that of the Frank plaintiffs. The Crawford plaintiffs challenged an Indiana statute requiring citizens voting in person on election day, or casting a ballot in person at the office of the circuit court clerk prior to election day, to present a photo ID. 553 U.S. at 185, 128 S.Ct. 1610. A majority of the Court determined that the plaintiffs had failed to prove that the statute was invalid. Although no opinion expressed the rationale of a majority of the Court, six Justices agreed that the Anderson/Burdick balancing test applied to the plaintiffs' claim. See Crawford, 553 U.S. at 189-91, 128 S.Ct. 1610 (opinion of Stevens, J.); id. at 204-08, 128 S.Ct. 1610 (opinion of Scalia, J.). The opinions differed, however, with respect to how the balancing test was to be applied. Justice Scalia's view of the test was that a law could be evaluated only on the basis of its "reasonably foreseeable effect on voters generally," rather than on its effect on subgroups of voters. Id. at 206, 128 S.Ct. 1610 (emphasis in original). In contrast, Justice Stevens seemed to assume that a law could be invalid based on its effect on a subgroup of voters. Id. at 200-03, 128 S.Ct. 1610. Here, however, he concluded that the plaintiffs had failed to produce a record that enabled the Court to determine whether the law placed an excessive and/or unjustified burden on the rights of a subgroup of voters. Id. at 200, 128 S.Ct. 1610 ("[O]n the basis of the evidence in the record it is not possible to quantify either the magnitude of the burden on this narrow class of voters or the portion of the burden imposed on them that is fully justified."). Justice Stevens determined that this gap in the record left the Court with no choice but to weigh the state's justifications for the law against its "broad application to all Indiana voters." Id. at 202-03, 128 S.Ct. 1610. He and the Justices who joined his opinion concluded that because 99% of Indiana's voting-age population already possessed photo IDs that would allow them to comply with the new law, id. at 188 n. 6, 128 S.Ct. 1610, the state's general interests in the law were sufficient to justify the burdens it imposed on Indiana voters generally. Id. at 202-03, 128 S.Ct. 1610.

Because in Crawford a majority of the Court agreed that a photo ID requirement such as provided in Act 23 is to be evaluated under the Anderson/Burdick balancing test, I will apply that test here. However, because a majority of the Court could not agree on how to apply the test, Crawford is not binding precedent on that matter. "When a fragmented Court decides a case and no single rationale explaining the result enjoys the assent of five Justices, the holding of the Court may be viewed as that position taken by those Members who concurred in the judgment on the narrowest grounds." Marks v. United States, 430 U.S. 188, 193, 97 S.Ct. 990, 51 L.Ed.2d 260 (1977) (internal quotation marks and alteration omitted). Here, the opinion authored by Justice Stevens is the narrowest. Like Justice Scalia, Justice Stevens concluded that the Indiana law was valid because the state interests justified the law's burden on "all Indiana voters." Crawford, 553 U.S. at 202-03, 128 S.Ct. 1610. But Justice Stevens did not expressly answer the further constitutional question answered by Justice Scalia: whether a law could be invalidated based on the burdens imposed on a subgroup of voters. Justice Scalia answered "no" to this question, id. at 204-08, 128 S.Ct. 1610, while Justice Stevens determined only that the plaintiffs had not shown that the Indiana law imposed excessive burdens on a subgroup of voters, id. at 200-03, 128 S.Ct. 1610. Because Justice Stevens's opinion is narrowest, and because Justice Stevens did not determine whether a law could be invalidated based on the burdens it imposes on a subgroup of voters, Crawford is not precedential as to that question.

To find the rule of decision, then, I revert back to Anderson and Burdick, which are cases that produced majority opinions. And as I read these cases, they require invalidation of a law when the state interests are insufficient to justify the burdens the law imposes on subgroups of voters. Both cases emphasized that "[a] court considering a challenge to a state election law must weigh `the character and magnitude of the asserted injury to the rights protected by the First and Fourteenth Amendments that the plaintiff seeks to vindicate' against `the precise interests put forward by the State as justifications for the burden imposed by its rule,' taking into consideration `the extent to which those interests make it necessary to burden the plaintiff's rights.'" Burdick, 504 U.S. at 434, 112 S.Ct. 2059 (quoting Anderson, 460 U.S. at 789, 103 S.Ct. 1564) (emphasis added). The focus of this language is the rights of an individual plaintiff rather than the rights of "voters generally." Crawford, 553 U.S. at 206, 128 S.Ct. 1610 (opinion of Scalia, J.). This implies that an unjustified burden on some voters will be enough to invalidate a law, even if, because the law burdens other voters only trivially, the state's interests are sufficient to justify the burden placed on such other voters. Moreover, in Anderson, the Court explicitly framed the question presented as whether the Ohio law at issue placed an unconstitutional burden on the voting rights of a subgroup of the state's voters — namely, the subgroup composed of Anderson's supporters. 460 U.S. at 782, 103 S.Ct. 1564 ("The question presented by this case is whether Ohio's early filing deadline placed an unconstitutional burden on the voting and associational rights of Anderson's supporters."). For these reasons, I conclude that a law like Act 23 is invalid if it imposes burdens on a subgroup of a state's voting population that are not outweighed by the state's justifications for the law.

Given the above legal standards, I will proceed as follows. First, I will identify the state interests the defendants put forward to justify Act 23 and assess the extent to which Act 23 is necessary to serve those interests. Second, I will identify and assess the magnitude of the burdens Act 23 imposes on the right to vote. Finally, I will determine whether the state's interests are sufficiently weighty to justify those burdens.

A. The State's Justifications for Act 23

The defendants claim that Act 23's identification scheme serves four state interests: (1) detecting and preventing in-person voter-impersonation fraud; (2) promoting public confidence in the integrity of the electoral process; (3) detecting and deterring "other types of voter fraud;" and (4) promoting orderly election administration and accurate recordkeeping. Defs.' Post-Trial Br. at 8.

1. Detecting and preventing in-person voter-impersonation fraud

The defendants claim that Act 23 will deter or prevent voter fraud by making it harder to impersonate a voter and cast a ballot in his or her name without detection. Detecting and preventing in-person voter-impersonation fraud is a legitimate state interest, see Crawford, 553 U.S. at 196, 128 S.Ct. 1610, and the photo ID requirement does, to some extent, serve that interest by making it harder to impersonate a voter at the polls. However, as explained below, because virtually no voter impersonation occurs in Wisconsin and it is exceedingly unlikely that voter impersonation will become a problem in Wisconsin in the foreseeable future, this particular state interest has very little weight.

The evidence at trial established that virtually no voter impersonation occurs in Wisconsin. The defendants could not point to a single instance of known voter impersonation occurring in Wisconsin at any time in the recent past. The only evidence even relating to voter impersonation that the defendants introduced was the testimony of Bruce Landgraf, an Assistant District Attorney in Milwaukee County. Landgraf testified that in "major elections," by which he means gubernatorial and presidential elections, his office is asked to investigate about 10 or 12 cases in which a voter arrives at the polls and is told by the poll worker that he or she has already cast a ballot. Tr. 2056-57. However, his office determined that the vast majority of these cases — approximately 10 each election — have innocent explanations, such as a poll worker's placing an indication that a person has voted next to the wrong name in the poll book. Tr. 2057. Still, about one or two cases each major election remain unexplained, and the defendants contend that these one or two cases could be instances of voter-impersonation fraud. I suppose that's possible, but most likely these cases also have innocent explanations and the District Attorney's office was simply unable to confirm that they did.5 Moreover, the most Landgraf's testimony shows is that cases of potential voter-impersonation fraud occur so infrequently that no rational person familiar with the relevant facts could be concerned about them. There are over 660,000 eligible voters in Milwaukee County,6 and if the District Attorney's office finds two unexplained cases each major election, that means that there is less than one questionable vote cast each major election per 330,000 eligible voters. The rate of potential voter-impersonation fraud is thus exceedingly tiny.

The evidence introduced by the plaintiffs confirms that voter-impersonation fraud does not occur in Wisconsin. The plaintiffs offered the testimony of Lorraine Minnite, a professor at Rutgers University who specializes in the study of the incidence of voter fraud in contemporary American elections. Professor Minnite studied elections in Wisconsin during the years 2004, 2008, 2010 and 2012 to determine whether she could identify any incidents of voter fraud. She consulted a variety of sources of information, including newspaper databases, news releases by the Wisconsin Attorney General, criminal complaints, decisions by state courts, and documents issued by the GAB. From these sources, Minnite was able to identify only one case of voter-impersonation fraud. Tr. 1036-42. And the single case of voter-impersonation fraud did not involve in-person voter impersonation. Rather, that case involved a man who applied for and cast his recently deceased wife's absentee ballot.7 Tr. 1041. Thus, from Minnite's work, it appears that there have been zero incidents of in-person voter-impersonation fraud in Wisconsin during recent elections.

Some have suggested that voter fraud might be more widespread than the low number of prosecutions indicates because the laws that prohibit voter fraud are underenforced. See Crawford, 472 F.3d at 953. However, the defendants do not suggest that there is any underenforcement of such laws in Wisconsin. And the evidence at trial indicates that such laws are vigorously enforced. In 2004, a Joint Task Force was created to investigate and prosecute voter fraud that occurred in Milwaukee during the 2004 presidential election. LULAC Ex. 68 ¶ 28. The task force included the United States Attorney, the Milwaukee County District Attorney, the Milwaukee City Attorney and a representative of the Milwaukee Police Department. In 2002, the United States Department of Justice started the Ballot Access and Voting Integrity Initiative in response to allegations of voter fraud across the country. LULAC Ex. 68 ¶¶ 20, 25. From 2002 to 2005, one of the goals of this initiative was to identify and prosecute individuals who committed voter fraud. Previously, the Department had only brought charges against conspiracies to corrupt the political process and not against individuals acting alone. One of the cities the Department focused on was Milwaukee. And, in September 2008, the Wisconsin Attorney General announced that his office was partnering with the Milwaukee County District Attorney to form an "Election Fraud Task Force" to detect, investigate and prosecute election fraud crimes in Milwaukee County. LULAC Ex. 812 ¶ 4. Before the 2010 general election, the Election Fraud Task Force expanded to include the district attorneys of 11 more counties. Id. ¶ 5. The task force not only followed-up on complaints about voter fraud, but it also dispatched teams of assistant attorneys general and special agents for the Division of Criminal Investigation to polling places across Wisconsin during the 2008, 2010 and 2012 elections, including the special June 2012 recall election. Accordingly, the lack of prosecutions for voter-impersonation fraud in Wisconsin cannot be attributed to underenforcement.

The defendants contend that the absence of known instances of voter-impersonation fraud could be explained by the fact that such fraud is difficult to detect. However, the witnesses called by the defendants to testify about their efforts to investigate voter fraud did not indicate that voter-impersonation fraud is difficult to detect. When Michael Sandvick, a former Milwaukee police officer, was asked at trial whether or not voter fraud was difficult to detect, he answered, "There are different types of voter fraud. Some of them are hard to detect and some of them are not." Tr. 2036. When asked what types are hard to detect, he gave only one example: someone using a fake address to vote. He did not mention voter impersonation.

Moreover, if voter impersonation is occurring often enough to threaten the integrity of the electoral process, then we should be able to find more evidence that it is occurring than we do. If, for example, voter impersonation is a frequent occurrence, then we should find more than two unexplained cases per major election in which a voter arrives at the polls only to discover that someone has already cast a ballot in his or her name. Another way to determine whether voter impersonation is occurring is a method suggested by the defendants' expert witness, M.V. Hood III, a professor of political science at the University of Georgia. See M.V. Hood III & William Gillespie, They Just Do Not Vote Like They Used To: A Methodology to Empirically Assess Election Fraud, 93 Social Science Quarterly 76 (March 2012). Professor Hood and his coauthor explain that one way to commit voter-impersonation fraud is to impersonate a registered voter who is recently deceased. Obviously, the deceased voter cannot show up at the polls, and thus a person who wanted to cast an illegal ballot could appear at the place where the deceased voter was registered and give the deceased voter's name. Hood's method for detecting this type of fraud involves comparing a database of deceased registered voters to a database of persons who had cast ballots in a recent election. If the researcher is able to match entries in both databases, then further investigation could be undertaken to determine whether voter impersonation had occurred. Hood and his coauthor applied this methodology to the 2006 elections in Georgia and found no evidence of ballots being illegally cast in the name of deceased voters. Id. at 81-92.

Thus, although voter-impersonation fraud may be difficult to detect, it is not invisible. If it is occurring in Wisconsin to any significant extent, then at trial the defendants should have been able to produce evidence that it is. The absence of such evidence confirms that there is virtually no voter-impersonation fraud in Wisconsin.

The defendants also contend that even if there currently is no voter impersonation in Wisconsin, the state has an interest in taking steps to prevent voter-impersonation fraud from becoming a problem in the future. In support of this contention, the defendants point out that the Supreme Court has stated that legislatures "should be permitted to respond to potential deficiencies in the electoral process with foresight rather than reactively, provided that the response is reasonable and does not significantly impinge on constitutionally protected rights." Munro v. Socialist Workers Party, 479 U.S. 189, 195-96, 107 S.Ct. 533, 93 L.Ed.2d 499 (1986). However, the Supreme Court has also stated that states cannot burden the right to vote in order to address dangers that are remote and only "theoretically imaginable." Williams v. Rhodes, 393 U.S. 23, 33, 89 S.Ct. 5, 21 L.Ed.2d 24 (1968). In the present case, no evidence suggests that voter-impersonation fraud will become a problem at any time in the foreseeable future. As the plaintiffs' unrebutted evidence shows, a person would have to be insane to commit voter-impersonation fraud. The potential costs of perpetrating the fraud, which include a $10,000 fine and three years of imprisonment, are extremely high in comparison to the potential benefits, which would be nothing more than one additional vote for a preferred candidate (or one fewer vote for an opposing candidate), a vote which is unlikely to change the election's outcome. Tr. 1017-19, 1342. Adding to the cost is the fact that, contrary to the defendants' rhetoric, voter-impersonation fraud is not "easy" to commit. To commit voter-impersonation fraud, a person would need to know the name of another person who is registered at a particular polling place, know the address of that person, know that the person has not yet voted, and also know that no one at the polls will realize that the impersonator is not the individual being impersonated. Tr. 1341. The defendants offered no evidence at trial to support the notion that it is easy to obtain this knowledge. Thus, given that a person would have to be insane to commit voter-impersonation fraud, Act 23 cannot be deemed a reasonable response to a potential problem.8

2. Promoting public confidence in the integrity of the electoral process

The defendants claim that the photo ID requirement serves the state's interest in promoting confidence in the integrity of the electoral process. It is true that the state has an interest in protecting the public's confidence in the integrity of elections so that citizens are encouraged to participate in the democratic process. Crawford, 553 U.S. at 197, 128 S.Ct. 1610. However, the defendants produced no empirical support for the notion that Act 23's photo ID requirement actually furthers this interest. In contrast, one of the plaintiffs' expert witnesses, Barry Burden, a professor of political science at the University of Wisconsin-Madison, testified that the available empirical evidence indicates that photo ID requirements have no effect on confidence or trust in the electoral process. He described a study conducted by Stephen Ansolabehere and Nathaniel Persily and published in the Harvard Law Review which looked at the relationship between photo ID laws and voter confidence in the electoral process. See Stephen Ansolabehere & Nathaniel Persily, Vote Fraud in the Eye of the Beholder: The Role of Public Opinion in the Challenge to Voter Identification Requirements, 121 Harv. L.Rev. 1737, 1756 (2008). Burden explained that this study employed multivariate analysis of survey data and found "zero relationship" between voter ID laws and a person's level of trust or confidence in the electoral process. Tr. 1385.

Perhaps the reason why photo ID requirements have no effect on confidence or trust in the electoral process is that such laws undermine the public's confidence in the electoral process as much as they promote it. As Professor Minnite testified, the publicity surrounding photo ID legislation creates the false perception that voter-impersonation fraud is widespread, thereby needlessly undermining the public's confidence in the electoral process:

Tr. 1019-20. Burden likewise testified that unsubstantiated allegations of voter fraud made by public officials undermine confidence in the electoral system. Tr. 1388-89. And Kevin Kennedy, the director of the GAB, in a letter to the Speaker of the Wisconsin State Assembly, offered the same opinion: "Speaking frankly on behalf of our agency and local election officials, absent direct evidence I believe continued unsubstantiated allegations of voter fraud tend to unnecessarily undermine the confidence that voters have in election officials and the results of the elections." Tr. 1389.

Another way that photo ID laws undermine confidence in the electoral process is by causing members of the public to think that the photo ID requirement is itself disenfranchising voters and making it harder for citizens to vote, thus making results of elections less reflective of the will of the people. See Tr. 578-79, 582-83 (testimony that Act 23 will exacerbate the lack of trust that the Black and Latino communities already have in the system); Tr. 951 (Lorene Hutchins, a Wisconsin voter, testified that she believes Act 23 is designed to keep certain people from voting); Tr. 396 (testimony that many voters believe Act 23 was designed to confuse voters).

For these reasons, I conclude that Act 23 does not further the state's interest in promoting confidence in the electoral process.

3. Detecting and deterring other types of fraud

The defendants contend that the photo ID requirement will help detect and deter forms of voter fraud other than voter impersonation. However, the defendants do not adequately explain how that could be so. The first type of unlawful voting the defendants cite is "voting under invalid voter registrations." Defs.' Post-Trial Br. at 12-13. The examples the defendants give of this kind of voter fraud are voting by a registered voter who has been convicted of a felony and voting by a non-citizen who has managed to register to vote. However, the defendants do not explain how the requirement to present an ID at the polls will prevent these types of unlawful voting, and I cannot think of any way that it could. If a person is registered and has a valid ID, that person will be allowed to vote. No evidence in the record indicates that persons convicted of a felony or non-citizens will be unable to present qualifying forms of ID. The defendants also claim that the photo ID requirement will help prevent unlawful voting by registered Wisconsin voters who no longer maintain residency in the state but who have not yet been removed from the poll list and unlawful double voting by individuals who register to vote in more than one state. Again, however, the defendants fail to explain how the requirement to present a photo ID will prevent these forms of unlawful voting, and I cannot think of any way that it could. Thus, I find that Act 23 does not serve the state's interest in preventing types of voting fraud other than in-person voter-impersonation fraud.

4. Promoting orderly election administration and accurate recordkeeping

The final state interest cited by the defendants is the state's interest in promoting orderly election administration and accurate recordkeeping. Again, there is no question that this is an important state interest. See Crawford, 553 U.S. at 196, 128 S.Ct. 1610. However, the defendants have not identified any way in which Act 23's photo ID requirement serves this interest that is distinct from the state's interest in detecting and preventing voter fraud. See id. (mentioning the state's interest in promoting orderly election administration and accurate recordkeeping in the course of a discussion of the state's interest in detecting and preventing voter fraud). Thus, Act 23 serves the state's interest in orderly election administration and accurate recordkeeping only to the extent that it serves the state's interest in detecting and preventing voter fraud. For the reasons already discussed, Act 23 only weakly serves the latter interest.

B. The Burdens Imposed by Act 23

Act 23 applies to all Wisconsin residents. However, the burdens it imposes on the right to vote fall primarily on individuals who do not currently possess a photo ID. For those who already have a qualifying ID, such as a driver's license, the barrier to voting that Act 23 creates is extremely low: such individuals must simply remember to bring their IDs to the polls. But, as I will discuss, many eligible voters do not currently have a photo ID. And the daily lives of many of these individuals are such that they have not had to obtain a photo ID for purposes such as driving.9 For these eligible voters, the requirement that they obtain a photo ID in order to vote erects a more substantial barrier. They must do whatever it takes to gather the necessary documents and make a special trip to the DMV in order to procure an ID that they will expect to use for no purpose other than to vote.

Although it is true that those individuals who already have IDs must have at one time experienced the burdens and inconveniences of obtaining them (and must continue to experience the burdens and inconveniences of keeping their IDs valid), the photo ID requirement creates a unique barrier for those who would not obtain a photo ID but for Act 23. The individuals who obtained their IDs before the photo ID requirement went into effect (or who would today obtain an ID for reasons unrelated to voting) expect to derive benefits from having those IDs that are unrelated to voting. For example, a person who obtains a driver's license receives a daily benefit — the ability to drive — from having experienced the burden of gathering the necessary documents and visiting the DMV. Once the photo ID requirement was adopted, that person received the benefit of being able to vote at no additional cost. In contrast, a person whose daily life did not require possession of a photo ID prior to the imposition of the photo ID requirement is unlikely to derive any benefit from possessing a photo ID other than the ability to continue voting. Yet that person must pay the same costs — in the form of the hassle of obtaining the underlying documents and making a trip to the DMV — as the person who obtained the ID for driving. This difference in expected benefits results in Act 23 imposing a unique burden on those who need to obtain an ID exclusively for voting, with the result that these individuals are more likely to be deterred from voting than those who already possess an ID for other reasons.

Based primarily on the testimony of plaintiff's expert, Leland Beatty, a statistical marketing consultant with extensive experience in business and politics, I find that approximately 300,000 registered voters in Wisconsin, roughly 9% of all registered voters, lack a qualifying ID.10 To put this number in context, in 2010 the race for governor in Wisconsin was decided by 124,638 votes, and the race for United States Senator was decided by 105,041 votes. See LULAC Ex. 2 ¶ 10 & Table 2. Thus, the number of registered voters who lack a qualifying ID is large enough to change the outcome of Wisconsin elections. In addition to these registered voters without an ID, there are a number of persons who are eligible to vote but not yet registered who lack an ID. Because Wisconsin permits same-day registration at the polls, any eligible voter may become a registered voter on election day. One of the plaintiffs' expert witnesses, Matthew Barreto, a professor at the University of Washington and an expert on voting behavior, survey methods and statistical analysis, conducted a telephonic survey of eligible voters in Milwaukee County. Professor Barreto found that there were 63,085 eligible voters in Milwaukee County alone who lack a qualifying ID.11

A substantial number of the 300,000 plus eligible voters who lack a photo ID are low-income individuals who either do not require a photo ID to navigate their daily lives or who have encountered obstacles that have prevented or deterred them from obtaining a photo ID. At trial, I heard from eight witnesses who intend to vote in Wisconsin elections but who do not currently possess a qualifying photo ID. Seven of these witnesses are low income. Alice Weddle testified that she is unemployed, receives Social Security and Medicare/Medicaid benefits and has no bank accounts or credit cards. She attempted to obtain an ID but was unable to do so because she does not have a birth certificate. Eddie Holloway testified that he would be homeless if his sister did not agree to take him in, and that he is on various forms of public assistance. He testified that he attempted to obtain an ID but was unable to do so because of an error on his birth certificate that he cannot afford to have corrected. Rickey Davis testified that he is unemployed, has no bank accounts and attempted to obtain a photo ID but could not get one because he does not have a birth certificate. Shirley Brown testified that she lives on Social Security disability and attempted to obtain an ID but was unable to do so because she does not have a birth certificate. Melvin Robertson testified that he has no education beyond grade school and that he would like to obtain an ID but cannot because he lacks a birth certificate. Rose Thompson testified that after Act 23 was enacted, she attempted to obtain an ID but could not afford to pay the fees associated with obtaining her birth certificate from Mississippi. Sim Newcomb testified that he does not drive, relies on public transportation, has not recently traveled outside the United States, does not travel on airplanes, and that to the extent he needs a photo ID for banking, he is able to use his Veteran's ID card, which is not an acceptable ID under Act 23. He testified that he attempted to obtain a Wisconsin ID card but could not satisfy the DMV's documentation requirements.12

Professor Barreto's research sheds additional light on the demographic makeup of those who lack an ID and lends further support to the conclusion that a substantial number of the 300,000 plus voters who lack an ID are low income. Barreto found that between 20,494 and 40,511 eligible voters in Milwaukee County who lack an ID earn less than $20,000 per year. Frank Ex. 600 at 31. As already noted, Barreto found that the total number of eligible voters in Milwaukee County who lack an ID is 63,085. Thus, individuals who make less than $20,000 per year comprise between 32% and 64% of the population of eligible voters without an ID. Barreto also found that 80.5% of the eligible voters without an ID have no education past the high-school level. Frank Ex. 600 at 29. Because individuals with less education are likely to be lower income,13 this finding also shows that a substantial number of voters who lack an ID are low income.

In light of the fact that a substantial number of the 300,000 plus voters who lack an ID are low income, Act 23's burdens must be assessed with reference to them rather than with reference to a typical middle- or upper-class voter. Although the latter voter may have little trouble obtaining an ID, he or she is not the type of voter who will need to obtain one in order to comply with Act 23. Thus, in the discussion that follows, I identify the burdens associated with obtaining a qualifying photo ID and explain how they will impact low-income voters.

For almost all low-income voters who lack an ID, the easiest ID to obtain will be the free state ID card, which is issued by the DMV. To obtain a state ID card, a person generally must present documents that satisfy four requirements: (1) proof of name and date of birth, (2) proof of United States citizenship or legal presence in the United States, (3) proof of identity, and (4) proof of Wisconsin residency. See Wis. Admin. Code § Trans 102.15. The DMV will only accept certain documents to satisfy each of these requirements.14 However, if a person has a Wisconsin driver's license or state ID card that has been expired for fewer than eight years, the person will be allowed to renew using a procedure that generally requires only proof of a social security number. Tr. 1092-94; Defs.' Ex. 1074.

To prove name, date of birth and United States citizenship, most people will need to produce a birth certificate. The evidence at trial showed that a substantial number of eligible voters who lack Act 23-qualifying IDs also lack birth certificates. Professor Barreto, in his survey of Milwaukee County eligible voters, found that 25,354 persons lacked both a qualifying ID and a birth certificate.15 Tr. 301-02. Seven of the witnesses who testified about their own lack of a qualifying ID stated that it was the lack of a birth certificate that was preventing them from obtaining an ID. Tr. 37-38, 93-94, 209-11, 401, 418-19, 708-09; Frank Ex. 606 at 7-12.

To obtain a Wisconsin birth certificate, a person must produce either a driver's license or a state ID card or two documents from the following list: (1) a government-issued ID with photograph, (2) a United States passport, (3) a checkbook or bankbook, (4) a major credit card, (5) a health-insurance card, (6) a recent, signed lease, or (7) a utility bill or traffic ticket. Tr. 1663; Frank Ex. 138. The person must also pay a fee of $20. Wis. Stat. § 69.22(1)(a).16 Those who were not born in Wisconsin will need to determine how to obtain a birth certificate from their place of birth. It generally takes more time and expense to obtain a birth certificate from outside one's state of residence than it does to obtain a birth certificate from within the state. See LULAC Ex. 811 ¶ 60. Professor Barreto found that 46.9% of eligible voters in Milwaukee County who lack both an accepted photo ID and a valid birth certificate were born outside Wisconsin. Frank Ex. 600 at 24.

Individuals who need a free state ID card must also produce a document that the DMV will accept as proof of identity. Professor Barreto found that there are approximately 1,640 eligible voters in Milwaukee County alone who do not have qualifying photo IDs and do not have any of the documents the DMV accepts to prove identity. Frank Ex. 600 at 37. Newcomb, one of the eight witnesses who testified about their inability to obtain an ID, testified that when he tried to obtain a state ID card he was unsuccessful because he lacked proof of identity. Tr. 845-46. Other witnesses, Dewayne Smith and Carl Ellis, testified that they did not have proof of identity when Act 23 first passed and had to obtain such proof before they could apply for state ID cards. Tr. 562-63, 566-67, 856-58.

Most voters who do not have proof of identity will need to procure a social security card, as this is the most commonly available document to use to prove identity. Defs.' Ex. 1077; Tr. 467, 1819. To obtain a social security card, a person must visit the Social Security Office and show "convincing documentary evidence of identity." 20 C.F.R. § 422.107(c). Such evidence "may consist of a driver's license, identity card, school record, medical record, marriage record, passport, Department of Homeland Security document, or other similar document serving to identify the individual." Id. Voters who need free state ID cards to vote will not have driver's licenses, state ID cards or passports, so they will need to present one of the other items on the list. If they do not have one of these items, they will need to procure one by visiting a school, hospital or another governmental agency, where they may again be asked for an ID, and the document may cost money. See Tr. 857 (Smith had to ask his sister to show the hospital her photo ID so he could get his medical records to apply for a social security card); Tr. 121 (marriage certificate from the State of Illinois costs $11).

The remaining documentary requirement to obtain a state ID card is proof of residence. For most voters, this requirement will be easy to satisfy, as the DMV accepts a variety of documents that most individuals are likely to have on hand. Still, homeless voters who do not have a relationship with a social-service agency will be unable to prove residency. Tr. 1889 (homeless people can only prove residence by getting a letter from a social service agency). And they will be unable to provide the DMV with a physical address where it can send their ID cards once they are ready. Id. This will make it impossible for them to obtain a state ID card because the DMV does not allow individuals to pick up ID cards in-person. Id.

Having explained the general legal requirements for obtaining a free state ID card and identified the necessary underlying documents, I consider the practical obstacles a person is likely to face in deciding whether to obtain an ID for voting purposes. Again, because most individuals who lack ID are low income, I consider these obstacles from the perspective of such an individual.

The first obstacle to obtaining an ID will be to identify the requirements for obtaining a free state ID card. I am able to summarize the requirements for obtaining an ID because I have access to the Wisconsin Statutes and Administrative Code and heard testimony on the topic at trial. A typical voter who needs an ID, however, must educate him or herself on these requirements in some other way. Although this may be easy for some, for others, especially those with lower levels of education, it will be harder. Moreover, a person who needs to obtain one or more of the required documents to obtain an ID, such as a birth certificate, must determine not only the DMV's documentation requirements, but also the requirements of the agency that issues the missing document. This adds a layer of complexity to the process. See, e.g., Tr. 93-94 (Davis testified that the DMV told him he needs to order his birth certificate from Tennessee but he has no idea how to go about ordering it).

Assuming the person is able to determine what he or she needs to do to obtain an ID, the person must next consider the time and effort involved in actually obtaining the ID. This will involve at least one trip to the DMV. There are 92 DMV service centers in the state. Defs.' Ex. 1071. All but two of these close before 5:00 p.m. and only one is open on weekends. Tr. 1083-84, 1806-07. So, it is likely that the person will have to take time off from work. The person will either need to use vacation time if it's available or forego the hourly wages that he or she could have earned in the time it takes to obtain the ID. See Tr. 845 (Newcomb was unable to take paid time off from work to obtain an ID). The person will also have to arrange for transportation. Since this person does not have a driver's license and is low income, most likely he or she must use public transportation or arrange for another form of transportation. See Tr. 845-46 (Newcomb does not have a car and had to take a 45-minute bus ride to get to the DMV); Tr. 211 (Brown paid $3.00 each way to a driver from Medicare so she could get to the DMV); Tr. 562, 566-67 (Ellis walked to the DMV, which took 45 minutes each way, because he does not have a car and could not afford bus fare); Tr. 151-52 (Adams testified that the DMV in Kenosha is "out in the county," which means people who live in the inner-city and do not have cars must take the bus to get there); Tr. 430-33 (Lumpkin stated that the location of the DMV in Racine County is a problem because it is 3-5 miles away from the inner-city where the majority of the city's population lives, and cabs do not serve the inner-city); see also Frank Ex. 635 at 50-51 (GAB received a lot of complaints from voters who were having a hard time getting to the DMV, even from people in the City of Milwaukee, which has a "pretty good" public transportation system). Further, for some individuals public transportation will be of no help because not all of the DMV's service centers are accessible by public transit. Tr. 1848.

If the person does not have all of the documents the DMV requires to obtain an ID, then the person will most likely have to visit at least one government agency in addition to the DMV. If that is the case, then the person will likely have to take even more time off of work and pay additional transportation costs. Tr. 856-58 (Smith testified that he had to take the bus and ask for rides from others in order to visit the DMV, the Social Security Office, and other locations). Perhaps it is possible for a person to obtain a missing underlying document by mail, but even so that will require time and effort.

A person who needs to obtain a missing underlying document is also likely to have to pay a fee for the document. For some low-income individuals, it will be difficult to pay even $20.00 for a birth certificate. See Tr. 1988-89 (Robert Spindell, a member of the Board of Election Commissioners for the City of Milwaukee, stated that he personally knows individuals who will cannot pay even $20.00 for a birth certificate); see also Tr. 431-32 ("[W]hen the choice is made whether or not to pay $33 for an ID or to put some food on the table, I think any of us can kinda guess which way people will go."). Three witnesses, Thompson, Davis and Ellis, testified that they could barely afford to pay for a birth certificate. Tr. 88, 564-66, 704-05. And Raymond Ciszewski testified that he has met many low-income individuals in Milwaukee who have trouble paying for their birth certificates. Ciszewski is a volunteer at St. Benedict's Church in Milwaukee. Tr. 530-31. He works in the church's birth-certificate program, which helps low-income individuals obtain birth certificates by paying the birth-certificate fee to the extent it exceeds $5.00. Tr. 534-35. The program primarily serves homeless individuals, persons recently out of jail and persons in rehabilitation programs. Tr. 532-33. Ciszewski testified that over the last seven years he has helped over 600 people acquire birth certificates who would not otherwise have been able to afford them, and many of these people could barely afford the $5.00 co-pay the church requires. Tr. 532, 534-36.17

An additional problem is whether a person who lacks an ID can obtain one in time to use it to vote. For many who need an ID, it will take longer than a day or two to gather the necessary documents and make a trip to the DMV. Indeed, if a person needs to obtain a birth certificate, especially from another state, it might take weeks or longer to obtain it. Tr. 1114, 1660-61. If an election is imminent, a person may be unable to procure an ID in time to vote or to validate a provisional ballot by the Friday after the election.

Another problem that arises is a person's having errors or discrepancies in the documents needed to obtain an ID. For example, the DMV requires the name on a person's social security card and birth certificate to match. If there is an error in a person's social security record, the person must visit the Social Security Office and correct the record. Tr. 1884.18 If there is an error in a person's birth certificate, the person must get it amended.19 Making additional trips to government agencies to resolve discrepancies will require more time off work and additional transportation costs.

The defendants contend that the burden on those with errors or discrepancies in their underlying documents is mitigated by the fact that the DMV has discretion to grant exceptions. Although it is true that the DMV will sometimes make exceptions for such persons, this fact is not made known to applicants, Tr. 1121-24, 1891-94, and thus those who might benefit from the exception procedure are unlikely to learn of it. Consequently, those with errors in their underlying documents are more likely to give up trying to get an ID than to be granted an exception. The testimony of Genevieve Winslow illustrates this problem. Winslow is eligible to vote in Wisconsin. She testified that she did not have a qualifying photo ID when Act 23 went into effect, so she visited the DMV service center in Milwaukee County on Grange Avenue to apply for a free state ID card for voting purposes. Tr. 111. She brought with her a certified copy of her birth certificate, a certified copy of her marriage certificate, her social security card, her Medicare card, her property tax bill and her expired passport. Tr. 106. But the DMV employee who reviewed her application told her she could not get an ID because her name is misspelled on her birth certificate. Tr. 99-100. Her maiden name was Genevieve Kujawski, but her birth certificate says "Ganava Kujansky" (Ganava is the Polish version of Genevieve). Tr. 95-96. The employee told Winslow she would need to get her birth certificate amended. Tr. 106-07. Winslow and her son asked to speak with two different supervisors, who both agreed that Winslow would need to get an amended birth certificate. Tr. 107. Her son was frustrated by this experience and decided to call Winslow's state senator, Senator Tim Carpenter. Tr. 100-01. An aide in the senator's office told Winslow's son to contact James Miller, an official at the DMV. Tr. 100-01, 109-10. Miller said Winslow should return to the same DMV service center with the same documentation and ask for a particular supervisor. Tr. 110-11. When she did this, the DMV issued her an ID. Tr. 111-12. No one ever explained to Winslow why she was able to get an ID. They just told her it was a "special deal." Tr. 101.20

Given the obstacles identified above, it is likely that a substantial number of the 300,000 plus voters who lack a qualifying ID will be deterred from voting. Although not every voter will face all of these obstacles, many voters will face some of them, particularly those who are low income. And the evidence at trial showed that even small obstacles will be enough to deter many individuals who lack an ID from voting. Professor Burden testified about the "calculus of voting," which is "the dominant framework used by scholars to study voter turnout." LULAC Ex. 811 at 811; Tr. 1278-83. Under this framework, even small increases in the costs of voting can deter a person from voting, since the benefits of voting are slight. Tr. 1279-80. As Burden explained:

Tr. 1279-80; see also Tr. 1220-21 (Professor Levine also testified about the calculus of voting). Thus, for many voters who lack an ID, even minor burdens associated with obtaining one will be enough to deter them from voting. Cf. Crawford, 472 F.3d at 951 ("[E]ven very slight costs in time or bother or out-of-pocket expense deter many people from voting, or at least from voting in elections they're not much interested in."). But in light of the evidence presented at trial, it is also clear that for many voters, especially those who are low income, the burdens associated with obtaining an ID will be anything but minor. Therefore, I conclude that Act 23 will deter a substantial number of eligible voters from casting a ballot.

C. Weighing the Burdens Against the State Interests

In the previous section I determined that Act 23's burdens will deter or prevent a substantial number of the 300,000 plus voters who lack an ID from voting. "Substantial" is of course not a precise quantity, but a more precise measurement is impracticable. There is no way to determine exactly how many people Act 23 will prevent or deter from voting without considering the individual circumstances of each of the 300,000 plus citizens who lack an ID. But no matter how imprecise my estimate may be, it is absolutely clear that Act 23 will prevent more legitimate votes from being cast than fraudulent votes. Cf. Crawford, 472 F.3d at 953-54 (assessing whether "there are fewer impersonations than there are eligible voters whom the [Indiana photo ID] law will prevent from voting"). Thus, Act 23's burdens are not justified by the state's interest in detecting and preventing in-person voter impersonation. Moreover, because the state's interest in safeguarding confidence in the electoral process is evenly distributed across both sides of the balance — a law such as Act 23 undermines confidence in the electoral process as much as it promotes it — that interest cannot provide a sufficient justification for the burdens placed on the right to vote. Accordingly, the burdens imposed by Act 23 on those who lack an ID are not justified.

Having found a violation of the Fourteenth Amendment, I turn to the appropriate remedy. The lead opinion in Crawford noted that, even if the Indiana photo ID law placed an unjustified burden on some voters, the plaintiffs had not demonstrated that the proper remedy would be to invalidate the entire statute. 553 U.S. at 203, 128 S.Ct. 1610. In the present case, however, invalidating Act 23 is the only practicable way to remove the unjustified burdens placed on the substantial number of eligible voters who lack IDs. The plaintiffs suggest that I could order the defendants to allow eligible voters without photo IDs to vote without showing an ID or by signing an affidavit affirming their identities and lack of an ID. However, ordering such relief would be the functional equivalent of enjoining the current law and replacing it with a new law drafted by me rather than the state legislature. It is not clear that this approach would amount to a narrower remedy than simply enjoining the current law. Moreover, the Supreme Court has instructed the federal courts to avoid "judicial legislation," United States v. Nat'l Treasury Employees Union, 513 U.S. 454, 479, 115 S.Ct. 1003, 130 L.Ed.2d 964 (1995), and this is an apt term for the remedy envisioned by the plaintiffs. To grant this remedy, I would need to make a policy judgment as to whether eligible voters who do not have IDs should be required to sign affidavits of identity before receiving a ballot. And, if I found that an affidavit was required, I would need to decide what language the affidavit should contain. Once I issued this relief, I would have to supervise the state's election-administration officials to ensure that they were properly implementing my instructions. These tasks are outside the limited institutional competence of a federal court, and therefore I may not rewrite the photo ID requirement to conform it to constitutional requirements. See Ayotte v. Planned Parenthood, 546 U.S. 320, 329-30, 126 S.Ct. 961, 163 L.Ed.2d 812 (2006). I conclude that the only practicable remedy is to enjoin enforcement of the photo ID requirement.21

III. Section 2 of the Voting Rights Act

Both the LULAC plaintiffs and the Frank plaintiffs contend that Act 23's photo ID requirement violates Section 2 of the Voting Rights Act. Before addressing the merits of this claim, I address the defendants' argument that the LULAC plaintiffs lack standing to sue under the Voting Right Act.

A. Standing of LULAC plaintiffs

The defendants contend that the four LULAC plaintiffs lack standing to pursue a claim for injunctive relief under Section 2 of the Voting Rights Act. Whether they do has little practical significance, as the plaintiffs in the Frank case unquestionably have standing to pursue a claim for injunctive relief under Section 2, and only one plaintiff with standing is needed. See Crawford, 472 F.3d at 951. Nonetheless, because one or more of the plaintiffs with standing might drop out of this case before it is finally resolved, I will determine whether all four of the LULAC plaintiffs have standing to seek injunctive relief under Section 2.

The defendants argue that the LULAC plaintiffs lack Article III standing and also lack what is known as "statutory standing." I will begin with Article III standing, which requires a plaintiff to show that he or she has suffered an injury in fact that is fairly traceable to the challenged acts of the defendant and that is likely to be redressed by a favorable judicial decision. See, e.g., Lujan v. Defenders of Wildlife, 504 U.S. 555, 560-61, 112 S.Ct. 2130, 119 L.Ed.2d 351 (1992). Each element of standing must be supported in the same way as any other matter on which the plaintiff bears the burden of proof, i.e., with the manner and degree of evidence required at the successive stages of the litigation. Id. at 561, 112 S.Ct. 2130. We are at the trial stage of this case, and so the elements of standing must be supported by the evidence adduced at trial. Id. The only element of Article III standing that is in dispute is whether the LULAC plaintiffs have suffered an injury in fact. For this reason, I will not discuss the traceability or redressability elements.

The LULAC plaintiffs contend that they have established standing in two ways. First, they contend that they have standing to seek redress for their own injuries. Second, they contend that they have "associational" standing, which allows an organizational plaintiff to bring suit to redress an injury suffered by one or more of its members, even if the organization itself has not been injured. See, e.g., Hunt v. Washington State Apple Advertising Comm'n, 432 U.S. 333, 343, 97 S.Ct. 2434, 53 L.Ed.2d 383 (1977).

Turning first to the question of whether the LULAC plaintiffs have suffered their own injuries, I conclude that they have. It is well-established that an organization suffers a cognizable injury in fact when it devotes resources, however minimal, to dealing with effects of a law that are adverse to its interests. See, e.g., Havens Realty Corp. v. Coleman, 455 U.S. 363, 379, 102 S.Ct. 1114, 71 L.Ed.2d 214 (1982); Crawford, 472 F.3d at 951. I find based on the evidence adduced at trial that all four LULAC plaintiffs have devoted resources to dealing with the effects of Act 23 and would devote additional resources to dealing with those effects if the state-court injunctions were lifted. Each plaintiff devoted resources to educating its members and others whose interests it serves about the law and to helping individuals obtain qualifying forms of photo ID, and each plaintiff would do so again if Act 23 were reinstated. Tr. 146-49, 185-88 (LULAC); Tr. 375, 386 (Cross Lutheran Church); Tr. 343-47, 357-58 (Milwaukee Area Labor Council); Tr. 489-92, 519-20 (Wisconsin League of Young Voters). Accordingly, all four plaintiffs have standing to seek injunctive relief to redress their own injuries.

The defendants advance two reasons why the LULAC plaintiffs do not have standing in their own right. First, relying on a case from the Fifth Circuit, the defendants point out that not every diversion of resources establishes an injury in fact. See NAACP v. City of Kyle, Texas, 626 F.3d 233, 238 (5th Cir.2010). But Kyle does not suggest that the diversion of resources demonstrated by the plaintiffs in this case fails to qualify as an injury in fact. The resources found insufficient in that case were resources spent litigating the very claim at issue in the suit.22 Id. at 238. In the present case, no plaintiff is claiming litigation expenses as an injury in fact. Rather, they point to resources expended on educating their members and others about the requirements of Act 23 and on ensuring that those members and others obtain forms of identification that would allow them to vote. This is precisely the kind of expenditure of resources that the Seventh Circuit deemed sufficient to support standing in Crawford, 472 F.3d at 951.

Second, the defendants contend that the LULAC plaintiffs lack standing because they voluntarily spent resources in response to Act 23 and were not compelled to do so. This argument, as another court has recognized, "finds no support in the law." Florida State Conference of N.A.A.C.P. v. Browning, 522 F.3d 1153, 1166 (11th Cir.2008). If a voluntary as opposed to compelled expenditure of resources were insufficient to confer standing, then Crawford was wrongly decided, as Indiana's photo-identification law did not "compel" the Democratic Party to expend resources on getting its supporters to the polls.23 Crawford, 472 F.3d at 951. The only support the defendants can find for their argument is a single sentence in a Seventh Circuit opinion, which the defendants take out of context. The sentence is "No one has standing to object to a statute that imposes duties on strangers." Freedom From Religion Foundation v. Obama, 641 F.3d 803, 805 (7th Cir.2011). Taken out of context, this sentence implies that a person lacks standing to challenge a statute unless the statute imposes a legal duty on him or her, and that therefore a voluntary expenditure of resources made in response to the effects of the statute would not qualify as an injury in fact. But the law at issue in that case was a law requiring the President of the United States to issue each year a proclamation designating the first Thursday in May as a National Day of Prayer. Id. at 805. This law imposed no duties on anyone other than the President, and in addition it could not have prompted the plaintiff or anyone other than the President to expend any resources at all, voluntarily or not. Thus, placed in its proper context, the sentence cited by the defendants stands for the simple proposition that a person does not have standing to challenge a law that causes him or her no injury in fact. It does not stand for the proposition that a voluntary expenditure of resources does not qualify as an injury in fact.

Having found that the LULAC plaintiffs have standing to sue to redress their own injuries, I need not decide whether they also have standing to sue on behalf of their members. However, in the event that it becomes a relevant question on appeal, I will determine whether the LULAC plaintiffs also have standing to sue on behalf of their members. An association has standing to bring suit on behalf of its members when: (1) its members would otherwise have standing to sue in their own right; (2) the interests it seeks to protect are germane to the association's purpose; and (3) neither the claim asserted nor the relief requested requires the participation of individual members in the lawsuit. Hunt, 432 U.S. at 343, 97 S.Ct. 2434.

With respect to the first Hunt element, a member of one of the plaintiffs would have standing in his or her own right if that member is suffering an injury in fact. The defendants argue that the only way a member of the plaintiffs — that is, an individual voter — could be suffering an injury as a result of Act 23 is if that member currently lacks an acceptable form of photo ID and is unable to obtain an acceptable form of photo ID. However, the part of Act 23 that the plaintiffs challenge is the provision requiring a voter to present a photo ID at the polls. It is the need to present such an ID that injures a voter and confers standing to sue. See Common Cause/Georgia v. Billups, 554 F.3d 1340, 1351-52 (11th Cir.2009) (holding that "[r]equiring a registered voter either to produce photo identification to vote in person or to cast an absentee or provisional ballot is an injury sufficient for standing"). This means that even those members of the plaintiffs who currently possess an acceptable form of ID have standing to sue. Id. at 1352 ("[T]he lack of an acceptable photo identification is not necessary to challenge a statute that requires photo identification to vote in person.").24 Thus, every member of the plaintiff organizations who is a Wisconsin voter has suffered an injury in fact. As the defendants do not dispute that each plaintiff has members who intend to vote in Wisconsin elections, I conclude that all four LULAC plaintiffs have members who are injured by Act 23.

Moreover, even if the lack of an acceptable photo ID were a prerequisite to standing, at least one of the LULAC plaintiffs, Cross Lutheran Church, has members who lack such an ID. Weddle, an African American member of the Church, testified at trial that she currently does not possess an acceptable form of photo identification. Tr. 35-36. I find her testimony credible and conclude that she does not, in fact, possess an acceptable form of photo identification. The defendants contend that Weddle could if she tried hard enough obtain an acceptable form of identification, but this has no bearing on her standing to sue. The premise of this lawsuit is that voters should not have to bear the burdens associated with obtaining and presenting identification in order to vote. A plaintiff who must bear those burdens in order to vote is necessarily injured by Act 23, whether or not he or she would be successful in obtaining and presenting an ID. Accordingly, I find that Cross Lutheran Church has members who have standing to challenge Act 23 on the ground that they lack acceptable forms of ID.25

The second Hunt element requires that the lawsuit be "germane" to the organization's purpose. I find that this lawsuit is germane to each LULAC plaintiff's purpose. LULAC's mission is to "advance the economic condition, educational attainment, political influence, housing, health, and civil rights of the Hispanic population of the United States." Tr. 158-59. It is hard to imagine a suit that is more germane to this mission than the present suit, which seeks to remove a barrier to minority participation in the political process and thus advance the political influence of Hispanics. Cross Lutheran Church believes that God requires it to fight for the civil rights of its members. Tr. 365-66, 377-78. Again, it is hard to imagine a suit that is more germane to this purpose than the present suit. One of the purposes of the Milwaukee Area Labor Council is "[t]o organize for social and economic justice, to propose and support legislation that is beneficial to working families, and to oppose legislation that harms working people." Tr. 342-43. Again, the present suit is germane to this purpose. Finally, this lawsuit is obviously germane to one of the purposes of the League of Young Voters Education Fund, which is to encourage young people of color to vote. Tr. 518.

The third Hunt element asks whether the claim asserted or the relief requested requires the participation of the organization's members in the lawsuit. I conclude that the participation of members is not required. The claims were tried without substantial participation by the plaintiffs' members, and nothing about the relief requested — an injunction — requires their participation. Accordingly, this element is satisfied.

Having concluded that the four LULAC plaintiffs have Article III standing, I turn to the defendants' remaining standing argument, which is that the plaintiffs lack "statutory standing." As I noted in a prior opinion, LULAC ECF No. 84, "statutory standing" is not a matter of standing in the Article III sense but a question of substantive law. The question is whether the statute under which the plaintiffs sue, here Section 2 of the Voting Rights Act, authorizes the plaintiffs to sue. See Steel Co. v. Citizens for a Better Environment, 523 U.S. 83, 92, 118 S.Ct. 1003, 140 L.Ed.2d 210 (1998).

With respect to that question, Section 2 allows suits to be instituted by "aggrieved person[s]." See 42 U.S.C. § 1973a. The Supreme Court has determined that similar language in Title VII of the Civil Rights Act of 1964 incorporates the "zone of interests" test. Thompson v. North American Stainless, LP, 562 U.S. 170, 131 S.Ct. 863, 870, 178 L.Ed.2d 694 (2011). Under this test, a plaintiff may not sue unless he falls within the zone of interests sought to be protected by the statutory provision whose violation forms the legal basis for his complaint. Id. at 870. The test denies a right to sue where "the plaintiff's interests are so marginally related to or inconsistent with the purposes implicit in the statute that it cannot reasonably be assumed that Congress intended to permit the suit." Id. (quoting Clarke v. Sec. Indus. Ass'n, 479 U.S. 388, 399-400, 107 S.Ct. 750, 93 L.Ed.2d 757 (1987)).

The defendants contend that only individuals seeking to enforce their right to vote are within the zone of interests of Section 2, and that organizations seeking to protect the voting rights of individuals are not within the zone of interests. I disagree. The word "person" in an act of Congress is presumed to include organizations, see 1 U.S.C. § 1, and thus the text of the statute does not suggest that a cause of action under Section 2 is limited to individuals. Moreover, the Senate Report on the bill that added the "aggrieved persons" language to the Voting Rights Act confirms that Congress intended to confer a right to sue on organizations seeking to protect the voting rights of their members and others. See S.Rep. No. 94-295, at 40 (1975), reprinted in 1975 U.S.C.C.A.N. 774, 806-07 ("An `aggrieved person' is any person injured by an act of discrimination. It may be an individual or an organization representing the interests of injured persons."). The evidence adduced at trial establishes that all four LULAC plaintiffs are organizations representing the interests of individuals whose voting rights are burdened by Act 23. Therefore, I find that all four LULAC plaintiffs fall within the zone of interests of Section 2 and are aggrieved persons within the meaning of Section 2.

In support of their argument that the plaintiffs are not aggrieved persons, the defendants cite Roberts v. Wamser, 883 F.2d 617, 621 (8th Cir.1989), and various district court cases that rely on Roberts.26 In Roberts, the Eighth Circuit held that "an unsuccessful candidate attempting to challenge election results does not have standing under the Voting Rights Act." 883 F.2d at 621. Neither this holding nor the reasoning that led to it supports the defendants' argument that organizations representing the interests of injured voters cannot be aggrieved persons under Section 2. In fact, the Eighth Circuit implied that had the plaintiff in Roberts been suing to protect the rights of other voters, he would have been an aggrieved person. Id. ("Nor does Roberts allege that he is suing on behalf of persons who are unable to protect their own rights."). Accordingly, the defendants' reliance on Roberts and the district court cases decided in its wake is misplaced.

In sum, I find that all four LULAC plaintiffs have Article III standing in two ways: they have standing to seek redress for their own injuries and also associational standing. I also find that all four plaintiffs have statutory standing.

B. Merits

Section 2 of the Voting Rights Act prohibits states from imposing or applying "any voting qualification or prerequisite to voting or standard, practice, or procedure" that "results in a denial or abridgement of the right of any citizen of the United States to vote on account of race or color." 42 U.S.C. § 1973(a). To prove a Section 2 violation, a plaintiff does not need to prove discriminatory intent. See Chisom v. Roemer, 501 U.S. 380, 394 & n. 21, 111 S.Ct. 2354, 115 L.Ed.2d 348 (1991). Rather, a Section 2 violation is established "if, based on the totality of circumstances, it is shown that the political processes leading to nomination or election in the State or political subdivision are not equally open to participation by members of a class of citizens protected by [§ 1973(a)] in that its members have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice." 42 U.S.C. § 1973(b). In the present case, the plaintiffs claim that the requirement to show a photo ID is a voting practice that results in Blacks and Latinos having less opportunity to participate in the political process and to elect representatives of their choice.

Before going further, I must determine how to apply Section 2 in the context of a challenge to a voting practice like the requirement to present a photo ID at the polls. Much of the Section 2 jurisprudence was developed in the context of so-called "vote dilution" cases. The term "vote dilution" — which is contrasted with the term "vote denial" — describes cases involving structural devices, such as at-large elections and redistricting plans, that can be used to minimize or cancel out the effect of minority votes. At-large elections can be used to minimize or cancel out the effect of minority votes because they submerge a minority group that would likely constitute a majority in a single-member district within a larger white majority. Redistricting plans can be used to minimize or cancel out the effect of minority votes because they scatter a minority voting bloc that would likely constitute a majority in a properly drawn district among several irregular districts, with the result that the minority voting bloc within any single district is too small to constitute a majority. The present case does not involve at-large elections, redistricting plans, or similar structural devices, and the legal standards developed for dealing with those devices do not necessarily apply here. For example, the so-called "Senate factors" or "Gingles factors," see Thornburg v. Gingles, 478 U.S. 30, 106 S.Ct. 2752, 92 L.Ed.2d 25 (1986), play a central role in vote-dilution cases. However, those factors were developed to assist courts in resolving the tension between, on the one hand, ensuring that structural practices such as at-large elections and redistricting plans are not used to dilute minority voting power, and, on the other, the Congressional directive that Section 2 does not require proportional representation. See Baird v. City of Indianapolis, 976 F.2d 357, 359 (7th Cir. 1992); Daniel P. Tokaji, The New Vote Denial: Where Election Reform Meets the Voting Rights Act, 57 S.C. L.Rev. 689, 722 (2006). Factors developed for this purpose are not necessarily relevant to cases, like this one, that do not present that tension, and in any event the federal courts have largely disregarded the Senate factors in Section 2 cases that do not involve challenges to at-large elections, redistricting plans, and the like. See Tokaji, supra, at 720-21 (arguing that the Senate factors do not help courts decide cases that do not involve vote dilution and observing that the lower courts have mostly disregarded those factors in vote-denial cases). Thus, I cannot resolve the present issue by applying the legal standards developed for vote-dilution cases.27

Although the vast majority of Section 2 cases involve vote dilution, appellate courts have extensively discussed Section 2 in the context of felon disenfranchisement, which does not involve vote dilution and falls into the category of "vote denial." See, e.g., Farrakhan v. Gregoire, 623 F.3d 990 (9th Cir.2010) (en banc); Simmons v. Galvin, 575 F.3d 24 (1st Cir.2009); Hayden v. Pataki, 449 F.3d 305 (2d Cir.2006) (en banc); Johnson v. Governor of Florida, 405 F.3d 1214 (11th Cir.2005) (en banc). However, the consensus that has emerged in those cases is that laws disenfranchising felons do not violate Section 2 because those laws existed when the Voting Rights Act was enacted in 1965 and the legislative history of the Act supports the conclusion that Congress did not intend to invalidate them. See, e.g., Farrakhan, 623 F.3d at 993 (finding that "[f]elon disenfranchisement laws have a long history in the United States," and that "Congress was no doubt aware of these laws when it enacted the VRA in 1965 and amended it in 1982, yet gave no indication that felon disenfranchisement was in any way suspect"). This reasoning obviously does not apply to voter photo identification requirements, which are a recent phenomenon. See Kathleen M. Stoughton, A New Approach to Voter ID Challenges: Section 2 of the Voting Rights Act, 81 Geo. Wash. L.Rev. 292, 296-98 (2013) (describing history of voter ID legislation, which begins in the year 2000). Thus, the felon-disenfranchisement cases are not helpful.

Because the cases contain only limited guidance,28 I will focus on the text of the statute. See Gonzalez v. City of Aurora, 535 F.3d 594, 597 (7th Cir.2008) (emphasizing the importance of considering the text of Section 2). The key language states that a violation of Section 2 is established if the totality of the circumstances shows that the challenged voting practice results in a political process that is not "equally open to participation by members [of a minority group]," in that the members of that group "have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice." 42 U.S.C. § 1973(b). The meaning of this language is clear: "Section 2 requires an electoral process `equally open' to all, not a process that favors one group over another." Gonzalez, 535 F.3d at 598. Justice Scalia, in a dissent in a vote-dilution case, provided the following illustration of the meaning of Section 2: "If, for example, a county permitted voter registration for only three hours one day a week, and that made it more difficult for blacks to register than whites, blacks would have less opportunity `to participate in the political process' than whites, and Section 2 would therefore be violated...." Chisom, 501 U.S. at 407-08, 111 S.Ct. 2354 (Scalia, J., dissenting). Based on the text, then, I conclude that Section 2 protects against a voting practice that creates a barrier to voting that is more likely to appear in the path of a voter if that voter is a member of a minority group than if he or she is not. The presence of a barrier that has this kind of disproportionate impact prevents the political process from being "equally open" to all and results in members of the minority group having "less opportunity" to participate in the political process and to elect representatives of their choice.

The next question is whether the evidence adduced at trial shows that Wisconsin's photo ID requirement creates a barrier to voting that is more likely to appear in the path of a voter if that voter is Black or Latino. The photo ID requirement applies to all voters, regardless of race. However, as explained in Section II.B, above, the requirement places a unique and heightened burden on those who must obtain an ID if they wish to continue voting in Wisconsin. These individuals are more likely to be deterred from voting than those who obtained their photo IDs for other reasons, such as driving. The evidence adduced at trial demonstrates that this unique burden disproportionately impacts Black and Latino voters. As the defendants concede, the plaintiffs' evidence "shows that minorities are less likely than whites to currently possess qualifying ID." Defs.' Post-Trial Brief at 1. Because the defendants concede that minorities are less likely than whites to currently possess a photo ID, it is not necessary for me to discuss the evidence adduced at trial in support of this point and make explicit findings of fact. Nonetheless, because the parties presented substantial evidence on this question at trial and explicit findings might prove useful in the event of an appeal, I will explain how the evidence adduced at trial leads to the conclusion that, in Wisconsin, Blacks and Latinos are less likely than whites to possess a qualifying form of photo identification.

Three of the plaintiffs' expert witnesses offered testimony supporting the conclusion that Blacks and Latinos in Wisconsin are less likely than whites to possess a qualifying ID. First, the plaintiffs presented the testimony of Leland Beatty. As discussed in more detail in Appendices A and C,29 Beatty compared a list of Wisconsin registered voters to a list of individuals holding a Wisconsin driver's license or state ID card and attempted to determine how many registered voters could be matched to a corresponding driver's license or state ID card. Then, using the assistance of a third party, Beatty determined the likely race of the Wisconsin registered voters who could not be matched to a driver's license or state ID card and computed the percentage of registered voters of each race who lacked such forms of ID. After performing this analysis, Beatty concluded that minority registered voters in Wisconsin "were substantially more likely to be without a matching driver's license or state ID than white voters." Tr. 645. Specifically, he found that data for the year 2012 showed that African American voters in Wisconsin were 1.7 times as likely as white voters to lack a matching driver's license or state ID and that Latino voters in Wisconsin were 2.6 times as likely as white voters to lack these forms of identification. Tr. 646-47, 658; LULAC Ex. 2. He also found that data for the year 2013 showed that African American voters in Wisconsin were 1.4 times as likely as white voters to lack a matching driver's license or state ID and that Latino voters were 2.3 times as likely as white voters to lack these forms of identification. Tr. 686; LULAC Ex. 817 ¶¶ 4, 9. I consider Beatty's findings and opinions credible and have given them significant weight in making my findings of fact.

Before moving on, I note that Professor Hood performed a matching analysis that was similar to Beatty's, except that he did not attempt to identify the race of the registered voters who did not possess an ID. As discussed in more detail in Appendix A, under some of Hood's criteria for determining whether a given registered voter could be matched to an ID, Hood found that the number of registered voters who could not be matched was smaller than the number found by Beatty: Beatty found that about 317,000 registered voters lacked an ID, while under the loosest of Hood's criteria only about 167,000 registered voters lacked an ID. For the reasons explained in Appendix A, I find Beatty's number more reliable than Hood's. But it is worth noting that when Beatty analyzed the racial breakdown of the voters Hood deemed to be without IDs, he found that the disproportionate impact on Blacks and Latinos was even greater: using Hood's numbers, Beatty found that both Blacks and Latinos were more than twice as likely as whites to lack driver's licenses or state ID cards. Tr. 682-84.

Next, the plaintiffs presented the testimony of Professor Barreto. Like Beatty, Barreto offered opinions on the existence of racial disparities in the possession of photo identification. However, the scope of Barreto's opinions differ from Beatty's in three ways. First, while Beatty examined ID possession by individuals who are registered to vote, Barreto examined ID possession by individuals who are eligible to vote. Second, while Beatty examined statewide possession rates, Barreto focused on Milwaukee County. Third, while Beatty focused on possession of driver's licenses and state ID cards, Barreto investigated possession rates of all forms of Act 23-qualifying ID.

As indicated in Section II.B and Appendix B, Barreto's opinions were based on a telephonic survey of Milwaukee County residents. The results of the survey showed that a sizable portion of the population of eligible voters in Milwaukee County do not possess either a qualifying form of ID or the documents needed to obtain a qualifying form of ID. Frank Ex. 600 at 16-17. Moreover, the results showed that Black and Latino eligible voters are less likely than white voters to possess a qualifying form of ID. Specifically, Barreto found that while only 7.3% of eligible white voters lack a qualifying form of ID, 13.2% of eligible African American voters and 14.9% of eligible Latino voters lack a qualifying form of ID. Tr. 304.

The defendants offer several reasons why I should give Barreto's findings limited weight. I have already discussed these reasons somewhat in Appendix B, in the context of determining the number of Wisconsin voters who lack IDs and the burdens they will face. Here I will discuss these reasons in the context of determining whether those who lack qualifying IDs are disproportionately likely to be Black or Latino.

First, the defendants contend that Barreto's findings are outdated. The survey was conducted in January of 2012, and the trial of this matter was held in November of 2013. The defendants note that between the time of the survey and the time of trial, Wisconsin's free ID program was in effect, that a significant number of people obtained IDs through this program during that time, and that a disproportionate share of free IDs were issued to Black and Latino voters. Thus, argue the defendants, it is possible that the free ID program mitigated somewhat the disparity in possession rates by the time of trial. I agree that this is possible. But the defendants do not suggest that the free ID program eliminated the disparity in possession rates identified in Barreto's survey. Moreover, at the time of Barreto's survey, the free ID program had been in effect for six months,30 and thus to some extent the survey results do account for the issuance of free IDs. Finally, it would be speculative to conclude that those who obtained free IDs since the time of Barreto's survey are individuals who previously lacked a qualifying form of ID. Many of the free IDs could have been issued as replacement IDs to individuals who already possessed IDs at the time of Barreto's survey, or as duplicate IDs to individuals who already possessed another form of ID at the time of the survey, such as a driver's license.31 And looking at the number of free IDs issued in isolation fails to take into account possible changes in the population of eligible voters: perhaps there has been an increase in the population of eligible voters, and although many new voters have obtained free IDs, many others have not obtained any form of ID. The defendants' own expert witness agreed that it would be speculative to draw conclusions about the disparity in possession rates based on the issuance of free IDs alone. Tr. 1559-61. Finally, Beatty updated his matching analysis just prior to trial, and he found that racial disparities in possession rates persist. Thus, despite the age of Barreto's survey, I remain convinced that his results support the conclusion that Blacks and Latinos are less likely than whites to possess qualifying forms of ID.

The defendants also point out that Barreto studied possession rates in Milwaukee County rather than statewide, and that therefore his findings do not prove that the disparities he found exist at the state level. This is a fair point, but it is weakened by the fact that Beatty studied statewide possession rates and found that the disparities Barreto identified in Milwaukee County do exist at the state level. Moreover, Milwaukee is the largest county in the state and has the state's largest populations of Blacks and Latinos, and thus findings based on a study of Milwaukee County alone are suggestive of what a statewide study would find. Tr. 284-85, 1517-20. Finally, there is no reason to think that in other parts of the state minorities possess IDs at such high rates and whites possess IDs at such low rates that the disparities found in Milwaukee County would be cancelled out if individuals from outside of Milwaukee were included in the study. To the contrary, a study of voting-age adults in Wisconsin published in 2005 found that Blacks and Latinos residing outside of Milwaukee County were less likely than whites to possess a valid driver's license. See John Pawasarat, The Drivers License Status of the Voting Age Population in Wisconsin, p. 22 (UW-Milwaukee Employment and Training Institute, June 2005); LULAC Ex. 58.32 Accordingly, I conclude that Barreto's findings, when added to the other evidence in this case, support the conclusion that minorities in Wisconsin are less likely than whites to possess a qualifying ID.

The remaining expert witness who offered testimony on the disparity in ID possession rates among minorities and white voters is Professor Burden. He identified a consensus in the literature showing that Black and Latino voters in Wisconsin and elsewhere in the United States are less likely than white voters to possess photo IDs. Tr. 1329-34. Burden cited the following studies: (1) a study performed by Professor Barreto and others showing that minorities in Indiana were less likely than whites to possess photo IDs, see Matt A. Barreto, et al., The Disproportionate Impact of Voter-ID Requirements on the Electorate — New Evidence from Indiana, 42 PS: Political Science & Politics 111 (2009); (2) an article coauthored by the defendants' expert witness, Professor Hood, which found that Blacks and Latinos in Georgia were less likely than whites to have driver's licenses, see M.V. Hood III & Charles S. Bullock III, Worth a Thousand Words? An Analysis of Georgia's Voter Identification Statute, 36 Am. Politics Research 555 (2008) (Def. Ex. 1005); (3) a study by the American Automobile Association showing that, in the United States, 18-year-old whites are significantly more likely than 18-year-old Blacks and Latinos to have driver's licenses, see AAA Foundation for Traffic Safety, Timing of Driver's License Acquisition and Reasons for Delay among Young People in the United States, 2012, at 11, table 3 (August 2013), available at www.aaafoundation.org/research/ completed-projects (last viewed April 28, 2014); and (4) the study by Pawasarat, discussed above, finding that in 2005 Black and Latino adults in Wisconsin were much more likely than white adults to lack valid driver's licenses, LULAC Ex. 58. Burden's testimony and the literature he cites reinforce the conclusion that Black and Latino voters in Wisconsin are more likely than white voters to lack qualifying IDs.

The defendants have pointed to no evidence introduced at trial or studies performed by others showing that Blacks and Latinos in Wisconsin or elsewhere possess IDs at the same or nearly the same rates as whites. To the contrary, as noted, they concede that "minorities are less likely than whites to currently possess qualifying ID." Defs.' Post-Trial Br. at 1. Thus, in light of the evidence presented at trial and the defendants' admission, the conclusion that Blacks and Latinos disproportionately lack IDs is inescapable.33

Although the defendants concede that Blacks and Latinos disproportionately lack IDs, they argue that the plaintiffs have not shown that Blacks and Latinos are incapable of obtaining qualifying IDs. This argument depends on the premise that a violation of Section 2 cannot be found unless the challenged voting practice makes it impossible for affected minorities to vote. As defense counsel argued in his closing:

Tr. 2134, 2142. Under the defendants's view of the law, the example given by Justice Scalia in Chisom — a county's permitting voting registration for only three hours one day a week and thereby making it more difficult for Blacks to register than whites — would not involve a violation of Section 2, since it would of course be possible for every Black person in the county to register during the one three-hour window per week. However, no authority supports the defendants' view of the law. The cases the defendants cite state that "a bare statistical showing" of disproportionate impact is not enough to prove a Section 2 violation. See Tr. 2134-35, citing Smith v. Salt River Project Ag. Improvement & Power Dist., 109 F.3d 586, 595 (9th Cir. 1997). But what these cases mean is that beyond showing a disproportionate impact on minorities, a Section 2 plaintiff must show that the disproportionate impact is tied in some way to the effects of discrimination. There is nothing in these cases indicating that a Section 2 plaintiff must show that the challenged voting practice makes it impossible for minorities to vote or that minorities are incapable of complying with the challenged voting procedure. Therefore, I reject the defendants' argument that Act 23 could violate Section 2 only if minorities who currently lack IDs are incapable of obtaining them.34

The defendants also argue that the plaintiffs' have not shown that minorities "face different considerations than whites in obtaining qualifying ID." Defs.' Post-Trial Br. at 1-2. In making this argument, the defendants imply that the burden of having to obtain an ID is not, by itself, a burden that could result in the denial or abridgment of the right to vote, and that the plaintiffs must point to some more serious burden that disproportionately impacts Black and Latino voters before they could establish a violation of Section 2. I disagree. Even if the burden of obtaining a qualifying ID proves to be minimal for the vast majority of Blacks and Latinos who will need to obtain one in order to vote, that burden will still deter a large number of such Blacks and Latinos from voting. As discussed in Section II.C, the plaintiffs' expert witnesses testified that, under the dominant framework used by scholars to study voter turnout, even small increases in the costs of voting can deter a person from voting, since the benefits of voting are slight and can be elusive. Tr. 1279-80, 1220-21. Under this framework, the need to obtain an ID is likely to deter a substantial number of individuals who lack IDs from voting, even if most of these individuals could obtain an ID without much trouble. These individuals, who prior to Act 23 were unwilling to pay the costs necessary to obtain an ID, are unlikely to pay those costs in order to comply with Act 23 when the expected benefits of voting are slight. Act 23 thus creates a political process in which white voters, who are more likely to already possess qualifying IDs than Black and Latino voters, will not face the deterrent effect of having to obtain an ID that they would not obtain but for the requirement to present it at the polls, while Blacks and Latinos who wish to vote and who lack qualifying IDs must pay the cost, in the form of time or bother or out-of-pocket expense, to obtain what is essentially a license to vote. This is not a political process that is "equally open to participation" by Blacks and Latinos. 42 U.S.C. § 1973(b). It is one in which a disproportionate share of the Black and Latino populations must shoulder an additional burden in order to exercise the right to vote.

But even if the defendants were correct that the plaintiffs needed to show that Blacks and Latinos face different considerations than whites in obtaining qualifying IDs, the plaintiffs would still have shown that Act 23 violates Section 2. There are additional hurdles that Blacks and Latinos who lack IDs are more likely to have to overcome than whites who lack them. First, as Professor Barreto's survey indicates, Black and Latino voters who lack a qualifying ID are more likely than white voters to also lack one or more of the underlying documents they would need to obtain a qualifying ID as a first-time applicant. In Milwaukee County, only 2.4% of white eligible voters lack both a qualifying ID and one or more of the underlying documents needed to obtain an ID, while 4.5% of Black and 5.9% of Latino eligible voters lack both an ID and at least one underlying document.35 Frank Ex. 600 at 23-24; Tr. 307-08. The defendants note that Barreto did not determine whether it would be impossible for those who lack both an ID and an underlying document to obtain the underlying document, but this misses the point. The point is that Barreto's survey shows that even among the pool of white and minority voters who lack IDs, Black and Latino voters are more disadvantaged than whites because they are more likely to have to overcome two hurdles in order to vote rather than one. First, they will have to obtain the missing underlying document, which will likely involve some time (such as a trip to the office of vital records) and expense (such as the fee for obtaining a birth certificate). Then, they will have to obtain state ID cards, which will involve the time and expense of going to the DMV. The need to overcome two hurdles instead of one makes the burden more substantial for a disproportionate share of Blacks and Latinos.

Another reason why it will be more difficult for many Blacks and Latinos to obtain IDs is that Blacks and Latinos are more likely to have been born outside of Wisconsin than whites. Professor Burden identified survey results showing that for the 5-year period ending in 2011, 75% of white residents were born in Wisconsin, yet only 59% of Blacks and 43% of Latino residents were born in the state. LULAC Ex. 811 ¶ 60. As discussed in Section II.B, it generally takes more time and expense to obtain a birth certificate from outside one's state of residence than it does to obtain a birth certificate from within the state. See also id. Therefore, Blacks and Latinos who need to obtain a birth certificate are likely to find themselves facing a more daunting task than their white counterparts.36 Moreover, Latino voters who speak primarily Spanish will face additional difficulties as they try to navigate a process that was designed to accommodate those who speak English. See Tr. 171 (witness testified that she did not see Spanish forms at DMV and could not get help from bilingual personnel); Tr. 133 (witness testified that she has worked with Latinos who encountered language barriers at the DMV).37

Up to this point, I have only discussed the evidence establishing that Act 23 has a disproportionate impact on Blacks and Latinos. But courts have stated that, to succeed on a Section 2 claim, a plaintiff must do more than establish that the challenged voting practice results in a disproportionate impact. See, e.g., Smith v. Salt River Project, 109 F.3d 586, 595 (9th Cir.1997) (noting that "a bare statistical showing of disproportionate impact on a racial minority" does not, by itself, prove a violation of § 2). Rather, the plaintiff must also show that the challenged voting practice produces a "discriminatory result." Id. What this seems to mean is that the plaintiff must show that the disproportionate impact results from the interaction of the voting practice with the effects of past or present discrimination and is not merely a product of chance. See Gingles, 478 U.S. at 47, 106 S.Ct. 2752 ("The essence of a § 2 claim is that a certain electoral law, practice, or structure interacts with social and historical conditions to cause an inequality in the opportunities enjoyed by black and white voters to elect their preferred representatives.").

I find that the plaintiffs have shown that the disproportionate impact of the photo ID requirement results from the interaction of the requirement with the effects of past or present discrimination. Blacks and Latinos in Wisconsin are disproportionately likely to live in poverty.38 Individuals who live in poverty are less likely to drive or participate in other activities for which a photo ID may be required (such as banking, air travel, and international travel),39 and so they obtain fewer benefits from possession of a photo ID than do individuals who can afford to participate in these activities. In addition, as explained in Section II.B, low-income individuals who would like to obtain an ID generally find it harder to do so than do those with greater resources. Cf. Texas v. Holder, 888 F.Supp.2d 113, 138 (D.D.C. 2012) (finding that "the burdens associated with obtaining an ID will weigh most heavily on the poor"), vacated on other grounds, ___ U.S. ___, 133 S.Ct. 2886, 186 L.Ed.2d 930 (2013). Thus, we find that Blacks and Latinos are less likely than whites to obtain a photo ID in the ordinary course of their lives and are more likely to be without one.

The reason Blacks and Latinos are disproportionately likely to live in poverty, and therefore to lack a qualifying ID, is because they have suffered from, and continue to suffer from, the effects of discrimination. At trial, Professor Levine of the University of Wisconsin-Milwaukee testified that residential segregation and housing discrimination are major causes of the socioeconomic disparities between whites and minorities in Wisconsin. By certain measures, Milwaukee ranks the worst of the 102 largest metropolitan areas in Black/white segregation and the ninth worst in Latino/white segregation. Tr. 1201-02. This level of segregation is, as Levine testified, "the cornerstone from which all of these other socioeconomic disparities flow." Tr. 1202-03. It prevents Black and Latino populations in central Milwaukee from accessing suburban employment opportunities. Tr. 1203. And there is a robust correlation between metropolitan areas that have high levels of segregation and low levels of Black male employment. Tr. 1208. Levine also testified that contemporary segregation can be traced in part to Milwaukee's history of housing discrimination. Tr. 1204-06.

The socioeconomic disparities between whites and minorities in Wisconsin are also traceable to the effects of discrimination in employment. Levine described one study of the Milwaukee labor market, conducted in the early 2000s, which showed that white job applicants received call-back interviews more than twice as frequently as Black applicants, and that even white applicants with criminal records received call-back interviews more frequently than Black applicants. Tr. 1211-13. Levine concluded that this study showed that "discrimination was alive and well in the Milwaukee labor market." Tr. 1212. Levine testified that racial disparities in education also contribute to the lower socioeconomic status of Blacks and Latinos in Wisconsin, and that these disparities are likewise a product of discrimination. Tr. 1214-16.

Professor Levine summarized his findings concerning the effects of discrimination on the socioeconomic status of Blacks and Latinos in Wisconsin as follows:

Tr. 1217.40 Similar testimony from Professor Burden, see Tr. 1298-1314, lends further support to the conclusion that the reason Blacks and Latinos are disproportionately likely to lack an ID is because they are disproportionately likely to live in poverty, which in turn is traceable to the effects of discrimination in areas such as education, employment, and housing. Based on this evidence, I conclude that Act 23's disproportionate impact results from the interaction of the photo ID requirement with the effects of past and present discrimination and is not merely a product of chance. Act 23 therefore produces a discriminatory result.

A remaining question is whether Section 2 requires or allows me to take the state's interest in the challenged voting practice into account. There is nothing in the text of Section 2 indicating that the state's interest is relevant, but one of the "unenumerated" Senate factors — whether the policy underlying the challenged voting practice is "tenuous" — suggests that it is. See Gingles, 478 U.S. at 37, 106 S.Ct. 2752. Moreover, it seems reasonable to understand Section 2 as allowing a state to maintain a voting practice despite any discriminatory result it produces if the practice is clearly necessary to protect an important state interest. However, as discussed in Section II.A., Act 23 only weakly serves the state interests put forward by the defendants. Accordingly, I conclude that those interests are tenuous and do not justify the photo ID requirement's discriminatory result.

To summarize my findings of fact and conclusions of law regarding the plaintiffs' Section 2 claim: Act 23 has a disproportionate impact on Black and Latino voters because it is more likely to burden those voters with the costs of obtaining a photo ID that they would not otherwise obtain. This burden is significant not only because it is likely to deter Blacks and Latinos from voting even if they could obtain IDs without much difficulty, but also because Blacks and Latinos are more likely than whites to have difficulty obtaining IDs. This disproportionate impact is a "discriminatory result" because the reason Black and Latino voters are more likely to have to incur the costs of obtaining IDs is that they are disproportionately likely to live in poverty, and the reason Black and Latino voters are disproportionately likely to live in poverty is connected to the history of discrimination against Blacks and Latinos in Wisconsin and elsewhere. Finally, Act 23 only tenuously serves the state's interest in preventing voter fraud and protecting the integrity of the electoral process, and therefore the state's interests do not justify the discriminatory result. Accordingly, the photo ID requirement results in the denial or abridgment of the right of Black and Latino citizens to vote on account of race or color.

A remaining matter is to identify the appropriate remedy.41 The plaintiffs request a permanent injunction against enforcement of the photo ID requirement, and the defendants have not argued that this is not a proper remedy. Moreover, such an injunction is the only practicable remedy — surely it would make little sense to allow Blacks and Latinos to vote without showing IDs while continuing to require white voters to show IDs. Thus, I will enjoin the defendants from requiring voters to present photo IDs in order to cast a ballot.

The LULAC plaintiffs point out that the Wisconsin legislature might amend the photo ID provisions of Act 23 in response to this decision. They ask me to make clear that I will schedule expedited proceedings to address any claim that an amendment to Act 23 has cured the defects identified in this opinion and provides grounds for relief from the permanent injunction. I will do so. Should the State of Wisconsin enact legislation amending the photo ID requirement, and should the defendants believe that, as amended, the photo ID requirement no longer violates Section 2, they may file a motion for relief from the permanent injunction. If an election is imminent at the time that the defendants file their motion, I will schedule expedited proceedings on the motion. However, I also note that, given the evidence presented at trial showing that Blacks and Latinos are more likely than whites to lack an ID, it is difficult to see how an amendment to the photo ID requirement could remove its disproportionate racial impact and discriminatory result.

IV. Other Matters

There are two remaining procedural matters to consider. The first is the Frank plaintiffs' motion for class certification and the second is the defendants' motion to dismiss the claims of certain plaintiffs. Given that the relief granted in this case is a permanent injunction against enforcement of the requirement that eligible voters present a photo ID to cast a ballot, these matters are moot. The motion for class certification is moot because, as the defendants concede, all members of the proposed classes will benefit from the permanent injunction whether or not classes are certified, and there is no reason to formally certify a class. Defs.' Br. in Opp. to Mot. For Class Cert. at 8, 20 (arguing that if Act 23 were enjoined there would be "no need for any classes as the remedy would invalidate the entire photo identification requirement and cover all of the citizens and registered voters in the State of Wisconsin"). Similarly, the motion to dismiss the claims of certain plaintiffs is moot — those plaintiffs will benefit from the relief requested regardless of whether they are dismissed as plaintiffs. Cf. Crawford, 472 F.3d at 951 (noting that as long as one plaintiff has standing to seek the injunctive relief requested, question of standing of additional parties can be ignored).

V. Conclusion

For the reasons stated, IT IS ORDERED that the named Defendants and Defendants' officers, agents, servants, employees, and attorneys, and all those acting in concert or participation with them, or having actual or implicit knowledge of this Order by personal service or otherwise, are hereby permanently enjoined from conditioning a person's access to a ballot, either in-person or absentee, on that person's presenting a form of photo identification.

IT IS FURTHER ORDERED that the Frank plaintiffs' motion for class certification is DENIED as MOOT.

IT IS FURTHER ORDERED that the defendants' motion for judgment on partial findings is DENIED as MOOT.

FINALLY, IT IS ORDERED that the Clerk of Court shall enter final judgment consistent with this opinion.

Appendix A: Wisconsin Voters Who Lack A Qualifying ID

I base my finding that approximately 300,000 registered voters in Wisconsin lack a qualifying ID primarily, but not exclusively, on the testimony of plaintiffs' expert, Leland Beatty. Beatty is a statistical marketing consultant with extensive experience both in business and politics. He sought to determine the number of registered voters who, as of September 2013, did not possess either a driver's license or state ID card, which matched the information maintained in the list of registered voters. Drivers' licenses and state ID cards are the two most common forms of Act-23 identification.1 To do this, Beatty obtained databases from both the DMV and GAB. The DMV database contained information about individuals with driver's licenses or state ID cards with expiration dates in September 2013 or later. The GAB database contained information about individuals who were registered to vote as of September 2013. Beatty compared the information in the databases to determine how many registered voters could be "matched" to a DMV product.

Initially, Beatty created three definitions of a "match." First, he counted a pair of entries as a match if a person having the same first name, last name, date of birth, residence county and zip code could be found in both the GAB and DMV databases. Second, he counted a pair of entries as a match if a person having the same first name, last name and date of birth could be found in both databases. Third, he counted a pair of entries as a match if a person having the same last name, date of birth and zip code could be found in both databases. This latter definition of a match was designed to account for individuals who were identified by a nickname in one database and their formal first name in the other.

After running these matches, Beatty attempted to "recover" or "reclaim" some of the non-matches by determining whether circumstances justified deeming them a match. First, he attempted to match people with multiple-word first or last names such as individuals with names like Mary Ann or Maryann or those with hyphenated last names. Beatty isolated the records of individuals with multi-word names and accepted a pair of entries as a match if either word in the multi-word name matched. Second, he attempted to match individuals who may have recently changed their last name as indicated by a field in the DMV database. He did this by treating any person listed in the DMV database as having a former last name as a registered voter with either a driver's license or a state ID card if he could find a person in the GAB database having the same first name, middle initial and date of birth.2

After reclaiming as many unmatched registered voters as he could, Beatty determined that 317,735 registered voters possessed neither a driver's license nor a state ID card. The total number of registered voters in Wisconsin was 3,395,688.3 Thus, 9.4% of registrants lacked a matching driver's license or state ID card.

In response to Beatty, the defendants offered the testimony of M.V. Hood III, a University of Georgia professor of political science. Hood also attempted to match registered voters in the GAB database with individuals in the DMV database in order to identify the number of registrants who possessed a qualifying ID. Hood, however, used different criteria than Beatty for determining what counted as a match, and he concluded that between 167,351 (4.9% of registrants) and 368,824 (10.9% of registrants) did not possess a driver's license or state ID card.4

The significant differences between the criteria employed by Beatty and Hood involve the use of the identification number associated with an entry. In the DMV database, the identification number is the number that appears on a person's driver's license or state ID card. In 2006, as required by the Help America Vote Act, Wisconsin began asking voters to write down this number when they registered to vote. Thus, for post-2006 registrants, the entries in the GAB database include an identification number. Hood used these numbers in two ways. First, he assumed that two entries qualified as a match if they had matching identification numbers. Second, employing some of his more "relaxed" criteria, he assumed that if a person had an identification number associated with his or her entry in the GAB database, that person also possessed a driver's license or a state ID card, even if the person could not be matched to a specific driver's license or state ID card by other means. Hood's use of identification numbers in these two ways caused him to find a greater number of matches between registered voters and DMV products than did Beatty. And as explained below, I find that his use of identification numbers in these ways renders his conclusions about the number of registered voters without an ID suspect. Therefore, I give greater weight to Beatty's conclusions than I do Hood's.

Regarding Hood's automatically counting a pair of entries as a match if they contained the same identification number: In the course of his work, Beatty noticed a large number of cases in which two individuals with the same identification number had different names or dates of birth. As an example, Beatty points to a case in which an identification number in the GAB database was assigned to a person with the first name Damon who was born in 1980 and resides in Milwaukee County, while the same identification number in the DMV database was assigned to a person with the first name Danielle who was born in 1971 and resides in Marinette County.5 LULAC Ex. 202 ¶ 6. On the basis of these observations, Beatty concluded that identification numbers were not unique. The defendants argue that Hood was right to assume that identification numbers are unique, but the only evidence they provide in support of this contention is Hood's testimony, which in turn is based on an interview he conducted with Debra Kraemer, a DMV employee who told Hood that DMV identification numbers are unique. Kraemer did not testify at trial, and the defendants have not explained how she determined that identification numbers are unique. Thus, I will not credit her hearsay statement. Besides interviewing Kraemer, Hood made no effort to verify whether identification numbers are unique, such as examining his matches to determine whether the names, dates of birth, etc., matched.6 Tr. 1546. Moreover, the defendants have not attempted to explain why, if identification numbers are unique, Beatty was able to find instances in which the same identification number was assigned to two different individuals, as in the case of Damon and Danielle. For these reasons, I conclude that Hood's decision to automatically count a pair of entries as a match if they had matching identification numbers renders his conclusions about the number of registered voters without an ID suspect.7

Regarding Hood's decision to deem a person with an identification number in the GAB database as possessing an ID: During his interview with Kraemer, Hood learned that DMV identification numbers are not permanent and that they are generated using an algorithm based on a person's name, sex, and date of birth. Any changes or corrections to a person's name or date of birth will cause the DMV to issue a new identification number. In light of this information, Hood hypothesized that some of the individuals with identification numbers in the GAB database who did not match an entry in the DMV database had informed the DMV of changes or corrections to their names or dates of birth. This would have caused the DMV to issue a new identification number, and this identification number would be different than the GAB identification number and also could have explained why Hood was unable to generate a match using the person's name and date of birth. Hood also hypothesized that some of his unmatched voters with identification numbers may have been in possession of a DMV product that expired before September 11, 2013, the latest date on which an ID could expire and still appear in the database Hood received from the DMV. If the ID expired after the date of the last general election, the person holding it could use it to comply with Act 23 until the date of the next general election. See Wis. Stat. § 5.02(6m)(a). On the basis of these hypotheses, which he did not meaningfully test,8 Hood, under his more relaxed criteria, counted every person in the GAB database with an identification number associated with his or her name as a registered voter who possessed a driver's license or a state ID card that could be used for voting. Hood determined that, under his relaxed criteria, the number of voters without a DMV product ranged from 167,351 to 285,425. See Defs.' Ex. 1001 at 6-7 & Table 2.

In general, I think it is reasonable to assume that a person with an identification number in the GAB database at one time possessed either a driver's license or a state ID card. After all, if a number appears in GAB database, it means that the person had a DMV identification number at the time he or she registered to vote and wrote that number down on the registration form. But the fact that a person at one time had a matching DMV product tells us little about whether that person currently has a matching DMV product. Possibly a significant number of people with identification numbers in the GAB database still have the driver's licenses or state ID cards they used when they registered, but assuming Hood's hypotheses are true, those cards will either be expired or they will have names on them that differ from the names that appear in the poll books. If they are expired, they could not be used to comply with Act 23 unless they expired between November 6, 2012 and September 12, 2013, and there is no evidence indicating that the number of IDs with expiration dates within this range is likely to be significant. If the IDs have different names on them, then it is unlikely that they could be used to comply with Act 23 because the names on the IDs will likely not conform to the names that appear in the poll books. See Wis. Stat. § 6.79(2)(a). Thus, I do not agree that individuals with identification numbers in the GAB database can, on that basis alone, be counted as individuals who currently possess a driver's license or state ID card that could be used to comply with Act 23. As Beatty did not automatically count such individuals as possessing a DMV product, I give greater weight to his opinion on the number of registered voters lacking IDs than I do to Hood's.

The defendants point out that Beatty did not investigate whether those who lack a valid driver's license or a valid state ID card nonetheless possess some other form of qualifying ID, such as a passport or a military ID. They then note that it is possible that the percentage of voters who possess only a form of ID other than a driver's license or a state ID card could be large. While this is possible, the defendants have pointed to no evidence suggesting that it is likely that a large percentage of the 317,735 voters who lack a valid driver's license or a valid state ID card possess some other form of ID. And plaintiffs' evidence is to the contrary. As discussed in greater detail in Appendix B, Matthew Barreto, an associate professor of political science at the University of Washington, conducted a telephonic survey of eligible voters in Milwaukee County and asked the survey respondents about the forms of ID they possessed. One of his findings was that the percentage of Milwaukee County eligible voters who had only a form of ID other than a driver's license or a state ID card was 0.3%. Tr. 300. Although Barreto's survey was conducted in Milwaukee County rather than statewide and targeted eligible voters rather than registered voters, there is no reason to think that the percentage of registered voters in the state who possess only a form of ID other than a driver's license or a state ID card is much higher than 0.3%. Applying this percentage to the number of registered voters in the GAB database provided to Beatty (3,395,688), we can estimate that about 10,000 voters in Wisconsin possess only a form of qualifying ID other than a driver's license or state ID card. If we subtract this estimate from the number of registered voters without a valid driver's license or state ID card, the estimated number of voters without any form of Act 23-qualifying ID in the state becomes 307,735. As this is an estimate rather than a precise measurement, I will round down to 300,000 and find that this is the number of registered voters in Wisconsin who, at the time of trial, did not possess a qualifying form of ID. This is approximately 9% of the population of registered voters in Wisconsin.

Appendix B: Expert Opinions of Matthew Barreto

Professor Matthew Barreto is an expert on voting behavior, survey methods and statistical analysis who, in January 2012, conducted a telephonic survey of eligible voters in Milwaukee County. Barreto designed the survey in collaboration with Professor Gabriel Sanchez of the University of New Mexico. The survey asked voters whether they had a qualifying photo ID as defined in Act 23. It also asked voters whether they had all of the primary documents required to obtain a free state ID card as a first-time applicant. The results showed that, of 661,958 eligible voters in Milwaukee County, 9.53% or 63,085 voters did not possess an acceptable form of photo ID, and 34.1% of these voters — 21,512 people — also lacked the primary documents required to get a free state ID card as a first-time applicant. Frank Ex. 600 at 16-17, 34, 37. Barreto concluded that the most common problem for individuals who lack primary documents is the requirement that they show proof of citizenship and name and date of birth. The survey results showed that 32% of the eligible voters in Milwaukee County who lack a photo ID — 20,162 people — do not have certified copies of their birth certificates or any of the other documents necessary to prove citizenship. Frank Ex. 600 at 37. Barreto also found that approximately 2.6% of the eligible voters in Milwaukee County who lack a qualifying photo ID — approximately 1,640 people — do not have any of the documents necessary to prove identity. Id.

The defendants argue first that Barreto's data it is outdated because the survey was conducted in January 2012 and the trial took place in November 2013. They suggest that, between the time of the survey and the time of trial, many of the individuals who lacked an ID at the time of Barreto's survey might have obtained one through the state's free ID program. They offer evidence showing that, between July 2011 and September 2013, the DMV issued 74,030 free state ID cards to Milwaukee County residents for voting purposes. Defs.' Ex. 1001 at 19 (Table 9).

I agree that it is possible that some of those who lacked an ID at the time of Barreto's survey have obtained one, but I find it unlikely that the free ID program substantially reduced the number of eligible voters without an ID. First, at the time of Barreto's survey, the free ID program had already been in effect for six months. Thus, the survey results account for the issuance of some of the free IDs. Second, some of the free IDs the DMV issued were replacement or renewal IDs that went to individuals who already had an ID. See Tr. 1818 (noting that a person can get a replacement or renewal card as part of the free ID program). Third, it is very likely that some of the free IDs were issued to individuals who already had driver's licenses. Although the DMV is not supposed to issue state ID cards to individuals who already possess a driver's license, data from the DMV shows that many individuals in Wisconsin possess both a driver's license and a state ID card. In April 2012, one of the plaintiffs' experts, Leland Beatty, reviewed the DMV's records and found 112,397 duplicate records in the driver's license and state ID card databases. LULAC Ex. 2. In these cases, the driver's license holder and the ID card holder had the same first name, last name, date of birth, gender, ethnicity, county of residence and zip code. LULAC Ex. 2. When Beatty updated his work in September 2013, he found that the number of duplicates had increased. Tr. 736, 739. Overall he found that about 30% of state ID card holders also have a driver's license. Tr. 690.1 One reason for the high number of duplicates might be that there is a common misconception that under Act 23 a person must obtain a special ID card from the DMV in order to vote even if he or she already has a driver's license. Tr. 1814. Fourth, looking at the number of free IDs issued in isolation fails to take into account possible changes in the population of eligible voters: possibly that population increased. The defendants' own expert agreed that it would be speculative to draw conclusions about current possession rates based on the issuance of free IDs alone. Tr. 1559-61. For all of these reasons, I find it unlikely that the free ID program has significantly changed the number of eligible voters who lack an ID.2

Alternatively, the defendants argue that Barreto's survey results should be given little weight because it will be easy for most of those who lack photo IDs to get free state ID cards. As evidence of this, they point to the testimony of Professor Hood. Hood reviewed Barreto's survey and found that it "was conducted in a professional manner using commonly accepted survey research practices," and he agreed that the survey results show that only 90.5% of eligible voters in Milwaukee County have a qualifying photo ID under Act 23. Defs.' Ex. 1003 ¶¶ 20, 26. However, he noted that the survey results show that an additional 6.9% of survey respondents who do not currently have a qualifying photo ID stated that they have had a Wisconsin driver's license or state ID card at some point in their lives. Id. ¶¶ 27-28. He believes that all of these individuals should be able to easily obtain free state ID cards for voting purposes because they successfully obtained them in the past. Id. Thus, he concludes that 97.4% of eligible voters in Milwaukee County either have a qualifying ID or could easily obtain one.

I reject Hood's conclusion that it will be easy for all 6.9% of the voters who held IDs at some point in the past to obtain a state ID card because I do not know anything about the circumstances of these voters. I do not know how long ago they held their IDs, and I do not know if they currently possess all of the primary documents required to obtain a state ID card. Even if a voter at some point had all of the documents required to get a state ID card, he or she could have lost some of the necessary documents. This is especially true for low-income voters, who Barreto found are more likely to lack a qualifying ID. Frank Ex. 600 at 28-31. It is also important to note that the DMV's documentation requirements have changed and become more strict over time. For example, the DMV used to accept a baptismal certificate or hospital birth certificate as proof of citizenship, but now it will only accept a certified copy of a birth certificate. Tr. 1848-49. Thus, a person who was able to meet the documentation requirements at some point in the past may not be able to do so today even if they still have all of the documents they used to obtain their first ID card.

Hood's analysis does, however, raise a question about how many of the eligible voters in Milwaukee County who currently lack IDs will be treated as first-time applicants by the DMV. The DMV treats anyone who had a Wisconsin driver's license or state ID card that expired within the last eight years as a renewal applicant, and it only requires renewal applicants to show proof of identity and, if the person has moved, proof of residence to get a state ID card. Tr. 1092-94; Defs.' Ex. 1074. Barreto's survey data shows that approximately 9.53% of eligible voters in Milwaukee County — approximately 63,085 people — do not have qualifying photo IDs under Act 23. The DMV will treat approximately 17,210 of these voters as first-time applicants because they are part of the 2.6% of people identified by Hood who have never had a Wisconsin driver's license or state ID card. Defs.' Ex. 1003 ¶ 27. But it is unclear how many of the remaining approximately 45,875 voters will be treated as first-time applicants because these voters have had Wisconsin driver's licenses or state ID cards at some point in the past. Anyone in this group who has had an ID that expired within the last eight years will be treated as a renewal applicant.

Because of the uncertainty about who will be treated as a first-time applicant, the record does not indicate exactly how many eligible voters in Milwaukee County lack a qualifying photo ID and the primary documents required to get one. I know from Barreto's report that 21,512 voters lack an ID and the documents required to get an ID if they are first time applicants, but I do not know how many of these voters will actually be treated as first-time applicants because Barreto did not consider this question. Barreto's data does, however, prove three things: (1) approximately 9.53% of the eligible voters in Milwaukee County, or 63,085 voters, do not have qualifying IDs under Act 23, (2) the DMV will treat at least 17,210 of these voters as first-time applicants if they apply for a state ID card because they have never had a Wisconsin driver's license or state ID card, and (3) there are approximately 1,640 eligible voters in Milwaukee County alone who lack qualifying IDs and proof of identity, which the DMV will require them to show regardless of whether they are a first-time or renewal applicant.

Appendix C: Beatty's Methodology in Determining Disproportionate Impact

Beatty performed a "matching" analysis of databases maintained by the GAB and the DMV. The GAB database contained a list of Wisconsin registered voters, and the DMV databases contained lists of Wisconsin residents who have a current driver's license and/or state ID card. Beatty's methodology proceeded in two major steps. First, he determined how many registered voters in the GAB database could be matched to either a driver's license or a state ID card in the DMV database. I have explained how he performed this step in Appendix A. Second, Beatty attempted to identify the race of the remaining unmatched voters. I explain this step of his methodology below.

Beatty submitted certain information about the unmatched voters to a third party, Ethnic Technologies, to determine their likely race. The reason Beatty did this is that the GAB database did not include information about race, and thus he had to determine the race of the voters who could not be matched to a DMV product through some other process. Ethnic Technologies is a firm that uses information about the name of a person and where that person lives to determine his or her likely race and ethnicity (among other characteristics). Typical clients of the firm include companies and organizations that engage in direct marketing in which it is important to know the race or ethnicity of the individuals receiving the company's marketing materials. Tr. 598-601.

Beatty explained the general process that a firm like Ethnic Technologies uses to identify race and ethnicity as follows:

Tr. 636-37.

John Mas, a former employee of Ethnic Technologies, provided examples to illustrate Ethnic Technologies' general methodology: "So the premise is that you look at a person's first name and you can infer a little bit about their culture or their background or you can look at their last name. If you heard Alex Rodriguez you wouldn't think that he's a Chinese person playing baseball." Tr. 606. The firm then refines its analysis by drawing inferences based on the neighborhood in which the person lives: "So you could have people with the last name Lee, like Bruce Lee or Stan Lee and help decipher if he's Chinese or Jewish. So if you know where they live you can know that Bruce Lee if he lived in Chinatown more than likely would be Chinese, or Stan Lee if he lived in Riverdale, New York was Jewish." Tr. 608.

Using the data Beatty provided, Ethnic Technologies was able to identify the likely race of 91.6% of the unmatched voters in 2012 and 93.1% of the unmatched voters in 2013. Beatty then computed the percentage of registered voters of each race that lacked a matching driver's license or state ID. This produced the following results: In 2012, 9.5% of white voters, 16.2% of Black voters, and 24.8% of Hispanic voters lacked a matching ID. In 2013, 8.3% of white voters, 11.5% of Black voters, and 19.2% of Hispanic voters lacked a matching ID.

The defendants offer two criticisms of Beatty's methodology. Their first criticism is that Beatty's analysis failed to account for the possibility that unmatched voters might possess a form of qualifying ID other than a driver's license or a state ID card. This is a fair point, but as other evidence in this case establishes, only an extremely small number of people possess a form of qualifying ID other than a driver's license or a state ID card and do not also possess either a driver's license or a state ID card. Tr. 300 (testimony of Professor Barreto reporting that only 0.3 percent of Milwaukee County residents had only a form of ID other than a driver's license or state ID card). Thus, although some unmatched voters will possess a form of qualifying ID other than a driver's license or state ID card, it is highly unlikely that these unmatched voters are so numerous that they would affect Beatty's ultimate conclusion that minorities are substantially more likely than whites to lack a qualifying form of ID.

The defendants' second criticism of Beatty's methodology has to do with his use of Ethnic Technologies to determine the race of the unmatched voters. The defendants point out that Ethnic Technologies determined the likely race of the unmatched voters by inputting the voters' names and locations into its proprietary software program, and that no witness gave precise details about the algorithm on which that software is based. However, the general principles underlying the software are known. As Beatty testified, the software first attempts to match a person's first name to an ethnicity, then examines the last name and possibly middle name, and finally uses information about the neighborhood in which the person lives to estimate the person's race and ethnicity. Ethnic Technologies also explains that this is the general principle underlying its software on its website. See LULAC Ex. 211; www.ethnictechnologies.com (last viewed April 28, 2014).

Moreover, even though we do not know the precise details surrounding Ethnic Technologies' software, there is ample evidence in the record indicating that Ethnic Technologies' software is reliable enough for the purposes it was used in this case, which is to estimate the racial makeup of a population. First, there is a consensus in the academic literature that although the general principles employed by Ethic Technologies — known as onomastics, Tr. 662-63 — do not perfectly determine a person's race, they "provide[] a sufficient level of classification confidence to be used in the measurement of inequalities and in the design and delivery of services that meet the needs of ethnic minorities." Pablo Mateos, A Review of Name-based Ethnicity Classifications Methods and their Potential in Population Studies, Population, Space and Place, July/August 2007, at 243; LULAC Ex. 213 at 26. Here, we are attempting to measure a racial inequality, and thus software based on onomastics is a proper tool to use. Second, a study supported by a grant from the National Cancer Institute found that Ethnic Technologies' software was "nearly perfect in estimating white race" — meaning that the software almost never identified a person as nonwhite when the person self-identified as white. Jessica T. DeFrank et al., Triangulating Differential Nonresponse by Race in a Telephone Survey, Preventing Chronic Disease, July 2007, at 1, 5; LULAC Ex. 212. It is true that the software misidentified a large number of self-identified Black individuals as white, id., but this does not undermine Beatty's conclusion that Black voters are more likely than white voters to lack photo ID. If anything, it indicates that the disparity in possession rates is even greater, as it implies that many of the unmatched voters whom Ethnic Technologies identified as white are actually Black.1 Moreover, as Beatty testified, Ethnic Technologies has improved its software since the time of the CDC study, and today the software has less of a tendency to misidentify Blacks as whites. Tr. 661-62.

A final factor indicating that Ethnic Technologies' software is reliable is the fact that Ethnic Technologies has been able to remain in business since 1995. Tr. 597. Marketers would not continue to hire Ethnic Technologies to estimate the race and ethnicity of their target audiences if its software were unreliable. And Beatty himself testified that he has been using Ethnic Technologies in his work for many years and has found their results to be very reliable. Tr. 634-35, 662-63.

In sum, I conclude that Beatty's methods, and the conclusions he reached after applying those methods, are reliable and should be given significant weight.

DECISION AND ORDER

In this decision and order, I address the defendants' motion for a stay pending appeal of my order of April 29, 2014, in which I permanently enjoined the defendants from conditioning a person's access to a ballot on that person's presenting a form of photo identification. See Fed.R.Civ.P. 62(c) & Fed. R.App. P. 8(a)(1). The standard for granting a stay pending appeal mirrors that for granting a preliminary injunction. In re A & F Enterprises, Inc. II, 742 F.3d 763, 766 (7th Cir.2014). Stays, like preliminary injunctions, are necessary to mitigate the damage that can be done during the interim period before a legal issue is finally resolved on the merits. The goal is to minimize the costs of error. Id. To determine whether to grant a stay, I must consider the moving party's likelihood of success on the merits, the irreparable harm that will result to each side if the stay is either granted or denied in error, and whether the public interest favors one side or the other. Id. As with a motion for a preliminary injunction, a "sliding scale" approach applies; the greater the moving party's likelihood of success on the merits, the less heavily the balance of harms must weigh in its favor, and vice versa. Id.

A. Likelihood of Success on the Merits

On appeal, the defendants argue that I misinterpreted the law applicable to the plaintiffs' claims under the Fourteenth Amendment and Section 2 of the Voting Rights Act. In general, the law applicable to such claims is unsettled, and thus I acknowledge that the defendants have some likelihood of success on the merits. However, having considered the specific arguments that the defendants raise in their motion for a stay, I conclude that their likelihood of success on the merits is low. I discuss these arguments below, as well as the defendants' argument that the scope of the injunction is too broad.

1. Fourteenth Amendment

In their motion for a stay pending appeal, the defendants initially argued that my disposition of the plaintiffs' Fourteenth Amendment claim is likely to be reversed because I made three errors: (1) deciding the claim when my disposition of the plaintiffs' claim under the Voting Rights Act made it unnecessary to do so, Mot. to Stay at 7; (2) enjoining the law as to all voters when I found that the law placed an unjustified burden on only a subgroup of voters, id. at 8; and (3) giving insufficient weight to the state's interest in preventing or deterring voter-impersonation fraud, id. at 9-10. In a recent letter, the defendants raise a fourth argument: that the decision of the Wisconsin Supreme Court in Milwaukee Branch of the NAACP v. Walker, 2014 WI 98, ___ Wis.2d ___, 851 N.W.2d 262, indicates that "Act 23 is lawful." Letter of Aug. 1, 2014, at 1. I address each argument in turn.

First, in my original order, I acknowledged that, given my resolution of the plaintiffs' claim under the Voting Rights Act, I could have declined to resolve the Fourteenth Amendment claim. Dec. & Order at 842-43. But as I explained, the two claims overlap substantially, in that many of the factual findings I made at the conclusion of a nearly two week trial were relevant to both claims, and therefore it would have been inefficient to resolve the claim under the Voting Rights Act but not the claim under the Fourteenth Amendment. Id. The defendants do not fully develop their argument that this approach constituted reversible error; instead, they merely assert in conclusory fashion that the decision to address the Fourteenth Amendment claim was error. Mot. to Stay at 7. Thus, I see no reason to think that the defendants are likely to succeed on this argument on appeal.

The defendants' second argument is that I should not have enjoined the photo-ID requirement in its entirety because I found that many Wisconsin voters already have a qualifying ID and thus will not experience unjustified burdens. The defendants suggest that I should have fashioned some other remedy that was limited to the voters who will experience the unjustified burdens that I identified. Mot. to Stay at 8. But as I explained in my original order, there is no practicable way to remove the unjustified burdens on the voters who do not currently possess an ID without enjoining the photo ID requirement as to all voters. Dec. & Order at 862-63. Indeed, the defendants did not, in their post-trial brief, identify any practicable remedy short of enjoining Act 23 in its entirety, and they do not, in their motion for a stay pending appeal, identify any such remedy. Thus, I conclude that the defendants are not likely to succeed on this argument.

The defendants' third argument is that my "application of the Anderson/Burdick balancing test was incorrect"1 because I "gave insufficient weight to the legitimate and important state interests that the Supreme Court recognized in Crawford."2 Mot. to Stay at 9. Specifically, the defendants contend that I gave insufficient weight to the state's interest in preventing and deterring voter-impersonation fraud. They argue that Crawford establishes that a state has a "legitimate and important interest" in preventing and deterring voter-impersonation fraud even in the absence of evidence that such fraud has occurred. Mot. to Stay at 9. I agree that Crawford generally establishes that a state has a legitimate and important interest in preventing or deterring voter-impersonation fraud, even in the absence of evidence that such fraud has occurred. But Crawford does not hold that a court may not consider the evidence (or lack thereof) that such fraud has occurred when deciding how much weight to assign to that particular interest under the Anderson/Burdick balancing test. If the evidence shows that voter-impersonation fraud is prevalent, then the state's interest in preventing and deterring such fraud may be "sufficiently weighty to justify" the burdens placed on the rights of individual voters. Crawford, 553 U.S. at 191, 128 S.Ct. 1610 (quoting Norman v. Reed, 502 U.S. 279, 288-89, 112 S.Ct. 698, 116 L.Ed.2d 711 (1992)). But if the evidence shows that voter-impersonation fraud is rare or nonexistent, then the state's interest is assigned less weight. In the present case, the evidence showed that virtually no voter-impersonation fraud occurs in Wisconsin and that it is unlikely to become a problem in the foreseeable future. For these reasons, I determined that the state's interest in preventing or deterring voter-impersonation fraud was insufficiently weighty to justify the burdens Act 23 placed on a substantial number of voters. The defendants have not shown that this application of the Anderson/Burdick balancing test was erroneous.

The defendants remaining argument concerning the Fourteenth Amendment claim is that the Wisconsin Supreme Court's decision in Milwaukee Branch of the NAACP v. Walker, establishes that Act 23 is lawful. In their letter to me, the defendants do not expand on this argument, but in their appellate filings, they argue that the state supreme court's construction of an administrative regulation "will eliminate the potential financial burden that many voters who lack a birth certificate might experience when obtaining a free ID card from the DMV." Consol. Reply Br. of Defendants-Appellants at 5. To explain the defendants' argument, I must first briefly discuss the relevant part of NAACP v. Walker.

The Wisconsin Supreme Court began its discussion of the administrative regulation at issue here by noting that, at the state trial-court level, the plaintiffs provided evidence that they were required to make payments to government agencies to obtain certain primary documents, such as birth certificates, that the division of motor vehicles ("DMV") requires them to produce in order to obtain free state ID cards for voting. NAACP, 2014 WI 98, ¶¶ 50 & 52, ___ Wis.2d ___, 851 N.W.2d 262. The court then determined that the DMV's requiring a person to produce a document that he or she cannot obtain without making a payment to a government agency resulted in a severe burden on the right to vote. Id. ¶¶ 60-62. In an effort to eliminate this severe burden, the court construed a regulation of the Wisconsin Department of Transportation that granted the administrator of the DMV discretion to issue state ID cards to persons who could not produce a birth certificate or other specifically identified document as proof of name and date of birth. Id. ¶ 66-71. That regulation states that if a person is "unable" to provide a birth certificate or other specifically identified document, and such documents are "unavailable" to the person, the person may make a written petition to the administrator of the DMV for an exception to the requirement to produce a birth certificate or similar document. Wis. Admin. Code § Trans 102.15(3)(b) & (c).3 Under the Wisconsin Supreme Court's construction of this regulation, a person is "unable" to provide a birth certificate or similar document, and such documents are "unavailable" to the person, "so long as [the person] does not have the documents and would be required to pay a government agency to obtain them." NAACP, 2014 WI 98, ¶ 69.

Under the Wisconsin Supreme Court's construction of this regulation, then, a person is entitled to petition for an exception to the birth-certificate requirement if the person cannot obtain a birth certificate without paying a fee to a government agency. But this does not mean that the person will be able to obtain a free state ID for voting without producing a birth certificate. This is so because, under the regulation at issue, a person must still provide "[w]hatever documentation is available which states the person's name and date of birth," Wis. Admin. Code § Trans 102.15(3)(b)3., and then the administrator, in his or her discretion, may accept or reject "such extraordinary proof of name and date of birth," id. § Trans 102.15(3)(c). There is no guidance in the regulation that indicates what kind of documentation might constitute "extraordinary proof of name and date of birth" or what factors the administrator should consider when exercising his or her discretion to determine whether the documentation the person has produced constitutes extraordinary proof. In NAACP, the Wisconsin Supreme Court did not provide any guidance to the administrator concerning the meaning of "extraordinary proof" or set forth any standard that might guide the exercise of the administrator's discretion. Instead, the court offered this cryptic instruction: "the administrator, or his or her designee, shall exercise his or her discretion in a constitutionally sufficient manner." 2014 WI 98, ¶ 70.

Another problem is that it is not clear how members of the public who need to obtain a free state ID will learn that the DMV now has discretion to issue IDs to persons who cannot obtain birth certificates without paying fees to government agencies. At the trial in this case, the plaintiffs demonstrated that the DMV does not publicize its exception procedure, which involves using Form MV3002, because the DMV wants to minimize exceptions. Dec. & Order at 858-59 n. 17. In light of this evidence, I concluded that a person who might need an exception is more likely to give up trying to get an ID than to be granted an exception. Id. at 858-61 & n. 17. Nothing in the supreme court's decision requires the DMV to do a better job of informing the public that the MV3002 procedure exists.

Thus, from the mere fact that a person may petition for an exception to the birth certificate requirement if the person cannot obtain a birth certificate without paying a fee to a government agency, it does not follow that the person will actually obtain a free ID card without producing a birth certificate, and so the defendants have not shown that the Wisconsin Supreme Court's construction of § Trans 102.15(3)(b) and (c) "will eliminate the potential financial burden that many voters who lack a birth certificate might experience when obtaining a free ID card from the DMV." Consol. Reply Br. of Defendants-Appellants at 5. In any event, having to pay a fee to obtain a birth certificate is only one of many burdens that a person who needs to obtain an ID for voting purposes might experience. See Dec. & Order at 857-61. So even if I assumed that the supreme court's construction of § Trans 102.15(3)(b) and (c) eliminates the burden of having to pay a fee to obtain a birth certificate or similar document, I could not conclude that the burdens Act 23 places on the right to vote have been lessened to such a degree that the state's interests are now sufficient to justify them. Accordingly, the state supreme court's decision does not significantly increase the defendants' likelihood of success on the Fourteenth Amendment claim.

2. Section 2 of the Voting Rights Act

The defendants argue that my disposition of the plaintiffs' Section 2 claim is likely to be reversed for two reasons: (1) I incorrectly determined that the LULAC plaintiffs have "statutory standing," and (2) my interpretation of how Section 2 applies in a "vote denial" case was erroneous. I have already addressed the statutory-standing argument twice and will not discuss it further, except to note that even if the defendants prevail on this argument on appeal, they will not succeed in reversing my disposition of the Section 2 claim, as the Frank plaintiffs unquestionably have statutory standing.4

With respect to my interpretation of Section 2, the defendants argue that I am likely to be reversed because my interpretation "would potentially invalidate other laws not reasonably subject to challenge, such as voter registration laws." Mot. to Stay at 11. The defendants argue that my interpretation has the potential to invalidate voting practices that are unquestionably legitimate (or, in their words, "not reasonably subject to challenge") if they have disproportionate impacts on the poor, as a greater percentage of minorities than whites are poor, and this is due to the history of discrimination against minorities. The suggestion is that my interpretation will lead to an absurd result by invalidating laws that Congress, in passing Section 2, could not have intended to invalidate. But what unquestionably legitimate voting practice is likely to have a disproportionate impact on the poor? In their motion to stay, the defendants point to voter registration, but they do not explain how voter registration is likely to have a disproportionate impact on the poor. In their appellate filings, the defendants point to "all existing voting practices that require in-person voting" by giving the following example:

Consol. Reply Br. at 18. Here, however, the final premise of the defendants argument — that in-person voting is more difficult without an automobile — is likely false. Based on the evidence presented at trial, I can conclude that lower-income minorities, especially those who do not own automobiles, are likely to live in urban areas, where it is easier to walk to a polling place than to drive. So it is very difficult to envision a plaintiff using disparities in rates of automobile ownership as a basis for challenging an existing voting practice that requires a person's presence at the polls.

In any event, even if it could be shown that an unquestionably legitimate voting practice would have a disproportionate impact on the poor, and therefore on minorities, that practice would not necessarily be invalidated under my interpretation of Section 2. As I noted in my original decision, if the voting practice was clearly necessary to protect an important state interest — an interest that is not "tenuous" — that voting practice could be sustained even if it has a disproportionate impact on minorities. Dec. & Order at 878-79. Any voting practice that could be described as "unquestionably legitimate" or "not reasonably subject to challenge" will almost certainly be clearly necessary to protect an important state interest. Consequently, I conclude that there is no merit to the defendants' argument that my interpretation of Section 2 will lead to results that Congress did not intend.

The defendants also argue that it was error for me to cite Justice Scalia's dissenting opinion in Chisom v. Roemer, 501 U.S. 380, 111 S.Ct. 2354, 115 L.Ed.2d 348 (1991), in the course of interpreting Section 2. The defendants make the obvious point that a dissent has no precedential value. But I did not cite Justice Scalia's dissent for its precedential value. I cited it because it illuminates the plain meaning of Section 2: it shows that I am not the only jurist to have read the text of Section 2 and come to the conclusion that it means that a state may not adopt a voting practice that makes it more difficult for minorities to vote than whites. The defendants also note that Justice Scalia's dissent was part of a "vote dilution" case, not a vote-denial case. But even though that is true, it does not alter the fact that the example Justice Scalia gave — a county's making it more difficult for Blacks to register to vote than whites — involved vote denial rather than vote dilution. Indeed, Justice Scalia himself described his example as involving a "nondilution § 2 violation[]." Chisom, 501 U.S. at 408, 111 S.Ct. 2354 (emphasis in original). So the dissent is instructive on the meaning of Section 2 as applied to a vote-denial claim.

The defendants also argue that I should have upheld Act 23 under Section 2 because Act 23 does not cause any of the racial disparities I identified, such as the disparity in poverty rates for whites and minorities and the resulting disparity in ID-possession rates. But although Act 23 does not cause these disparities, it clearly interacts with them in a way that makes it harder for minorities to vote. And it is this interaction with the effects of discrimination that produces a discriminatory result.5 Thornburg v. Gingles, 478 U.S. 30, 47, 106 S.Ct. 2752, 92 L.Ed.2d 25 (1986) ("The essence of a § 2 claim is that a certain electoral law, practice, or structure interacts with social and historical conditions to cause an inequality in the opportunities enjoyed by black and white voters to elect their preferred representatives." (Emphasis added)).6 The Ninth Circuit has specifically rejected the idea that a voting practice that produces a disproportionate racial impact may survive scrutiny under Section 2 so long as the voting practice, by itself, is not responsible for any underlying racial disparities. See Farrakhan v. Washington, 338 F.3d 1009, 1016-20 (9th Cir.2003). As the court stated, "demanding `by itself' causation would defeat the interactive and contextual totality of the circumstances analysis repeatedly applied by [other] circuits in Section 2 cases, as they also require a broad, functionally-focused review of the evidence to determine whether a challenged voting practice interacts with surrounding racial discrimination in a meaningful way or whether the practice's disparate impact `is better explained by other factors independent of race.'" Id. at 1018 (quoting Smith v. Salt River Project Agric. Improvement & Power Dist., 109 F.3d 586, 591 (9th Cir.1997)). The court continued:

Id. at 1019 (emphasis added).7 In other words, the question in a Section 2 case is whether the challenged practice magnifies or exacerbates an existing racial disparity caused by discrimination in other areas, thereby importing the effects of that discrimination into the electoral process. In the present case, the evidence showed that discrimination in areas such as employment, housing, and education caused higher poverty rates for minorities than for whites, with the result that a greater percentage of the minority population lacks a photo ID and will have more difficulty obtaining an ID. Act 23 thus imports the effects of discrimination in these areas into the electoral process and produces a discriminatory result.

The defendants also contend that my interpretation of Section 2 is incorrect because my interpretation "focuses not on causation but on mere likelihood." Mot. to Stay at 12. This is in reference to the following language in my opinion: "Section 2 protects against a voting practice that creates a barrier to voting that is more likely to appear in the path of a voter if that voter is a member of a minority group than if he or she is not." Dec. & Order at 870. The defendants argue that my interpretation focuses on whether a voting practice "could potentially create more difficulty for minorities to vote than non-minorities," rather than on whether it actually creates more difficulty for minorities to vote than non-minorities. Mot. to Stay at 13 (emphasis in original). This is wrong. Under my interpretation, a voting practice violates Section 2 only when it actually creates more difficulty for minorities to vote. And I found that Act 23 actually creates more difficulty for minorities to vote than non-minorities, in that Act 23 will prevent or deter a greater percentage of minorities from voting than whites. Dec. & Order at 875-76. The phrase "more likely than whites" (and related phrases) refers to the fact that although not every minority will be deterred or prevented from voting, a greater percentage of minorities will be deterred or prevented from voting than whites. Perhaps what the defendants mean to argue is that a voting practice does not violate Section 2 unless it prevents or deters every member of a racial group from voting. But there is no support for this narrow view of Section 2. The text states that a violation of Section 2 occurs when the political process is not "equally open to participation" by members of a minority group in that its members "have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice." 42 U.S.C. § 1973(b). If a voting practice will prevent or deter a greater percentage of minorities from voting than whites, the political process is not "equally open to participation" by minorities, in that they will not have the same opportunity as whites to participate in the political process and to elect representatives of their choice.

Finally, the defendants argue that because my conclusion that Act 23 violates Section 2 depends on the premise that a greater percentage of minorities than whites are poor, I have turned income or wealth into a protected class. That is incorrect. I concluded that Act 23 produces a discriminatory result because it interacts with the effects of racial discrimination, including higher poverty rates for minorities than for whites. If the reason a greater percentage of minorities than whites in Wisconsin are poor were unrelated to racial discrimination, then showing that minorities are more likely than whites to be poor would not have been sufficient to show that Act 23 produces a discriminatory result. Thus, the root cause of Act 23's disproportionate impact is discrimination on account of race, not income or wealth.

3. Scope of Injunction

The defendants' final argument is that I issued a permanent injunction that is impermissibly broad, in that I enjoined the defendants from enforcing any requirement to produce a photo ID to gain access to a ballot, not simply the photo-ID requirement embodied in Act 23. Importantly, however, I stated that if the State of Wisconsin enacted a new photo-ID law, the defendants could file a motion for relief from the permanent injunction, and that I would schedule expedited proceedings on any such motion, if necessary. Dec. & Order at 879-80. The defendants contend that I would have no jurisdiction to hear such a motion while an appeal from the order granting the original injunction was pending, because ordinarily a district court loses jurisdiction over a case between the time a notice of appeal is filed and the time the mandate issues. However, Federal Rule of Civil Procedure 62(c) states that while an appeal from an order granting an injunction is pending, the district court may modify the injunction. Moreover, to the extent there were any doubt over whether I would have jurisdiction to consider a motion to modify the injunction, the procedure outlined in Federal Rule of Civil Procedure 62.1 and Federal Rule of Appellate Procedure 12.1 would apply. These rules provide that if a timely motion is made in the district court for relief that the district court lacks authority to grant because an appeal is pending, the district court may inform the court of appeals that it would grant the motion (or that the motion raises a substantial issue), and then the court of appeals may remand the case to the district court for a ruling on the motion. Thus, if the State of Wisconsin enacts a new photo-ID law, the defendants are not precluded from seeking relief from the present injunction.

With respect to the question of whether I erred in enjoining the defendants from enforcing any photo-ID requirement, not just Act 23, I first note that even if this were error, it would not be grounds for staying my order pending appeal. If the court of appeals concludes that the injunction is impermissibly broad, the court will not reverse my order in its entirety. Rather, the court will vacate the injunction and remand with instructions to enter an injunction limited to Act 23. See PepsiCo, Inc. v. Redmond, 54 F.3d 1262, 1272 (7th Cir.1995) (stating that court of appeals will "restrict the breadth" of an overbroad injunction). Thus, the argument that the injunction is too broad does not support the defendants' motion for a stay pending appeal. At best, it is an argument that supports modifying the injunction pending appeal, which the defendants have not asked me to do.

In any event, the defendants have not shown that I erred in enjoining them from enforcing any photo-ID requirement, not just Act 23. An injunction "must ... be broad enough to be effective, and the appropriate scope of the injunction is left to the district court's sound discretion." Russian Media Group, LLC v. Cable America, Inc., 598 F.3d 302, 307 (7th Cir.2010). To make an injunction effective, a district court may enjoin "similar conduct reasonably related to" the violation established in the litigation. EEOC v. AutoZone, Inc., 707 F.3d 824, 841 (7th Cir.2013). In the present case, I concluded that to render the injunction effective, it was necessary to enjoin similar conduct reasonably related to the enforcement of Act 23. While the present case was under consideration, the Wisconsin Assembly adopted an amendment to Act 23, and the state's governor announced that he would call a special session of the legislature to modify Act 23 in the event that the courts did not uphold it. Now, it is possible that the state could make changes to Act 23 that result in its surviving scrutiny under the Voting Rights Act and the Fourteenth Amendment. However, given the evidence presented during the trial, it seemed doubtful that the kind of changes being discussed at the time would have had that result. Thus, to prevent the defendants from circumventing the injunction by enforcing a new photo-ID requirement that continued to place unjustified burdens on a substantial number of voters and that produced a discriminatory result, I enjoined the defendants from enforcing any photo-ID requirement, not just Act 23 as it then existed, until such time as it could be determined whether the new law removed the unjustified burdens and discriminatory result. In my discretion, I determined that an injunction of this breadth was necessary to render the relief afforded to the plaintiffs effective.

The defendants contend that, although a district court has authority to enjoin a defendant from engaging in conduct that is similar to the conduct found to be unlawful in the litigation, I abused my discretion by imposing a remedy that could be likened to the preclearance requirement of Section 5 of the Voting Rights Act, which the Supreme Court addressed in Shelby County v. Holder, ___ U.S. ___, 133 S.Ct. 2612, 186 L.Ed.2d 651 (2013). But the preclearance requirement of Section 5 prevents a covered jurisdiction from enforcing any changes to state election law until they have been precleared by the federal government. Id. at 2624. The injunction I issued allows the State of Wisconsin to enforce any changes to its election law that it wants, so long as the law at issue does not require a person to present a photo ID as a condition to receiving a ballot. Nothing in Shelby County suggests that once a specific voting practice has been shown to be unlawful under Section 2, a court may not enjoin a state from adopting a similar voting practice without first seeking relief from the injunction. Thus, the defendants' reliance on Shelby County is misplaced.

B. Irreparable Harm and the Public Interest

The other factors that I must consider when deciding whether to stay an injunction pending appeal are the irreparable harm that will result to each side if the stay is either granted or denied in error, and whether the public interest favors one side or the other. The irreparable harm that will result to the defendants if the stay is denied in error is tied to the interests the defendants put forward to justify Act 23 in the first place: preventing inperson voter-impersonation fraud and promoting public confidence in the integrity of the electoral process. But as I found in deciding this case on the merits, there is virtually no in-person voter-impersonation fraud in Wisconsin, and there is no evidence that laws such as Act 23 promote public confidence in the integrity of the electoral process. Thus, if a stay pending appeal is denied in error, the defendants would suffer very little irreparable harm — almost certainly no in-person voter-impersonation fraud will have occurred during the time that the appeal was pending, and the public's confidence in the integrity of the electoral process will not have declined.

On the other side of the balance, the irreparable harm that the plaintiffs would suffer if a stay were granted in error would be significant. To begin with, some of the named individual plaintiffs, including Shirley Brown and Eddie Lee Holloway, Jr., would be unable to vote during any election that occurred while the stay was in effect, as they lack a photo ID and have been unable to obtain a photo ID. Similarly, many members and individuals represented by the organizational plaintiffs in the LULAC case would be prevented or deterred from voting because of Act 23. Finally, under the public-interest factor, I take into account the fact that a large number of individuals who are not parties to this case and who might not be represented by any of the LULAC organizations would also be prevented or deterred from voting if Act 23 were reinstated pending appeal.

In short, in balancing the potential for irreparable harm to each party, I reiterate my finding that "it is absolutely clear that Act 23 will prevent more legitimate votes from being cast than fraudulent votes." Dec. & Order at 862. Thus, the balance of the harms weighs against a stay pending appeal.

C. Sliding Scale

Having found that the defendants' likelihood of success on appeal is low, that the defendants would suffer very little irreparable harm if a stay pending appeal were denied in error, and that the plaintiffs and members of the public would suffer significant irreparable harm if a stay pending appeal were granted in error, I conclude that, under the sliding-scale approach, I should not stay the permanent injunction pending appeal.

CONCLUSION

For the reasons stated, IT IS ORDERED that the defendants' motion for a stay pending appeal is DENIED.