EASTERBROOK, Circuit Judge.

Since 2005 Indiana has required voters to present photographic identification at the polls. The Supreme Court held that this statute is compatible with the Constitution. Crawford v. Marion County Election Board, 553 U.S. 181, 128 S.Ct. 1610, 170 L.Ed.2d 574 (2008). In May 2011 Wisconsin enacted a similar statute, 2011 Wis. Act 23. A district court held that Act 23 is unconstitutional and enjoined its implementation. Frank v. Walker, ___ F.Supp.3d ___, 2014 WL 1775432 (E.D.Wis. Apr. 29, 2014), stay denied, 2014 U.S. Dist. LEXIS 111811 (E.D.Wis. Aug. 13, 2014). After receiving briefs and argument, we stayed that injunction. Order issued Sept. 12, 2014; reconsideration denied Sept. 26, 2014; opinions issued Sept. 30, 2014. We now reverse the injunction, because the district court's findings do not justify an outcome different from Crawford.

The Justices observed that a commission chaired by former President Carter had recommended the use of photo ID to verify a person's entitlement to vote. Commission on Federal Election Reform, Building Confidence in U.S. Elections 18 (2002). The Court added that the Help America Vote Act of 2002 (HAVA) requires states to verify a person's eligibility to vote, using photo ID, portions of Social Security numbers, or unique state-assigned-identifiers. 52 U.S.C. § 21083(a)(5)(A), formerly 42 U.S.C. § 15483(a)(5)(A). Many people register to vote when they get drivers' licenses (National Voter Registration Act of 1993, 52 U.S.C. § 20504, formerly 42 U.S.C. § 1973gg-3), which links registration and photo ID from the outset. The Justices concluded that both the prevention of voter impersonation on election day and the preservation of public confidence in the integrity of elections justify a photo ID requirement, even though persons who do not already have government-issued photo IDs must spend time to acquire necessary documents (such as birth certificates) and stand in line at a public agency to get one. "For most voters who need them, the inconvenience of making a trip to the [department of motor vehicles], gathering the required documents, and posing for a photograph surely does not qualify as a substantial burden on the right to vote, or even represent a significant increase over the usual burdens of voting." 553 U.S. at 198, 128 S.Ct. 1610. These observations hold for Wisconsin as well as for Indiana.

Wisconsin's law differs from Indiana's, but not in ways that matter under the analysis in Crawford. One difference is that Wisconsin requires photo ID for absentee voting as well as in-person voting; a person casting an absentee ballot must submit a photocopy of an acceptable ID. Another difference is that when a person who appears to vote in person lacks a photo ID but says that he has one, and therefore casts a provisional ballot, the state will count that ballot if the voter produces the photo ID by the next Friday; in Indiana the voter signs an affidavit of eligibility in one of the state's circuit courts (which usually means travel to the county seat) within 10 days. Offices of the Department of Motor Vehicles in Wisconsin (where most people get government-issued photo IDs) are open shorter hours than those in Indiana, but more than three years have passed since Act 23's adoption, which makes it difficult to conclude that people who want photo ID have been unable to find an open office in all that time; no one thinks that people who want drivers' licenses in Wisconsin are unable to get them because of limited office hours. Wisconsin's list of acceptable documents (drivers' licenses, Wisconsin state ID cards, passports, military ID of persons in active service, recent naturalization papers, photo ID issued by a recognized Indian tribe, or signed photo ID issued by a college or university) omits some documents that Indiana accepts (see 553 U.S. at 198 n. 16, 128 S.Ct. 1610) and includes some that Indiana omits. There are other differences in detail, but none establishes that the burden of voting in Wisconsin is significantly different from the burden in Indiana.

The district court concluded that Crawford is not controlling for three principal reasons. First, the judge estimated that 300,000 registered voters in Wisconsin lack a photo ID that the state will accept for voting. That is approximately 9% of the state's 3,395,688 registered voters. The district judge in Crawford, by contrast, estimated that only 43,000 persons eligible to vote lacked an acceptable photo ID. Indiana Democratic Party v. Rokita, 458 F.Supp.2d 775, 807 (S.D.Ind.2006). Second, the judge found that voter-impersonation fraud (a ringer pretending to be a registered voter) happens so rarely in Wisconsin that the desire to reduce its occurrence cannot justify any significant burden on voters. Third, the judge found that white persons who are eligible to vote are more likely than others to have in their possession either an acceptable photo ID or the documents (such as copies of birth certificates) that make it simple to get an acceptable photo ID. The judge found that in Milwaukee County (which the judge took as a proxy for the whole state) 97.6% of white eligible voters have a qualifying photo ID or the documents they need to get one. That figure is 95.5% for black eligible voters and 94.1% for Latino eligible voters. The judge concluded from the first two findings that Act 23 violates the Constitution and from the third that it violates the Voting Rights Act. The judge made many other findings, but these are the most important ones.

Before we address the significance of the findings the judge made, we mention a few things that the judge did not find. First, the judge did not find that substantial numbers of persons eligible to vote have tried to get a photo ID but been unable to do so. Eight people testified that they had been frustrated when trying to get photo IDs. Six of the eight testified that the state would not issue photo IDs because they lack birth certificates, but they did not testify that they had tried to get them, let alone that they had tried but failed. Only two testified that distance or poverty hindered them when trying to obtain birth certificates or correct records to remove an error from a birth certificate.

Nor did the judge find that the situation of these eight differed from the situation of many persons in Indiana. The record in Crawford contains evidence about the same kind of of frustration, encountered by persons born out of state, who are elderly and may have forgotten their birthplaces and birthdates (if their parents ever told them), who are uneducated (and thus may not grasp how to get documents from public agencies), or who are poor (and so may have trouble getting to a public agency, or paying fees for copies of documents). The district judge here made extensive findings demonstrating that the poor are less likely to have photo IDs than persons of average income. Yet the district judge in Crawford also discussed these problems; so did the Supreme Court, which deemed them an inadequate basis for holding Indiana's law unconstitutional. 553 U.S. at 199-203, 128 S.Ct. 1610.

The Court reached that conclusion even though Indiana charged for copies of birth certificates — as did Wisconsin, at the time of trial. Between the trial and the argument of this appeal, however, the Supreme Court of Wisconsin directed state officials to issue photo IDs without requiring applicants to present any document that must be paid for. Milwaukee Branch of NAACP v. Walker, 2014 WI 98 ¶¶ 66-70, 851 N.W.2d 262. Moreover, Wisconsin recently issued regulations requiring officials to get birth certificates (or other qualifying documents) themselves for persons who ask for that accommodation on the basis of hardship. Emergency Rule 14, Wis. Admin. Reg. 704b (August 31, 2014). So at the time of trial it was no harder to get supporting documents in Wisconsin than in Indiana, and today it is easier in Wisconsin than in Indiana.1

Second, the judge did not make findings about what happened to voter turnout in Wisconsin during the February 2012 primary, when Act 23 was enforced (before two state judges enjoined it). Did the requirement of photo ID reduce the number of voters below what otherwise would have been expected? Did that effect differ by race or ethnicity? The record does not tell us. This suit, like Crawford, therefore is a challenge to Act 23 as written ("on its face"), rather than to its effects ("as applied").

The record also does not reveal what has happened to voter turnout in the other states (more than a dozen) that require photo IDs for voting. If as plaintiffs contend a photo ID requirement especially reduces turnout by minority groups, students, and elderly voters, it should be possible to demonstrate that effect. Actual results are more significant than litigants' predictions. But no such evidence has been offered.

The lack of evidence about what has happened in other states (or even in Wisconsin itself in 2012) means that this case is in the same posture as Indiana's: the parties and the district court have tried to make predictions about the effects of requiring photo ID, but the predictions cannot be compared with results.

Plaintiffs want us to treat Crawford as a case in which there was no record, so that the Supreme Court had no facts to go on. That's not what happened. An extensive record was compiled in Crawford, and the district judge issued a lengthy opinion. The judge in Indiana thought, just as the judge in Wisconsin has found, that some voters would be unable, as a practical matter, to get photo IDs — because of age or infirmity, lack of ability to pay for birth certificates, or the difficulty of obtaining them from public-records bureaus thousands of miles away in other states — and therefore would have to travel to the county seat after every election to file an affidavit of eligibility, but could not ascertain how many people were in that category. The trial in Wisconsin produced the same inability to quantify.

The findings not made affect how to interpret the findings that were made. Take the conclusion (based on the testimony of a "marketing consultant") that 300,000 registered voters lack acceptable photo ID. The number is questionable; the district judge who tried the Indiana case rejected a large estimate as fanciful in a world in which photo ID is essential to board an airplane, enter Canada or any other foreign nation, drive a car (even people who do not own cars need licenses to drive friends' or relatives' cars), buy a beer, purchase pseudoephedrine for a stuffy nose or pickup a prescription at a pharmacy, open a bank account or cash a check at a currency exchange, buy a gun, or enter a courthouse to serve as a juror or watch the argument of this appeal. Could 9% of Wisconsin's voting population really do none of these things? (Some may have photo ID that is not accepted for elections, such as a veteran's card, but the record does not show how many people get through life with the sort of photo ID that Wisconsin does not accept for voting.) Nonetheless, we accept the district court's finding in this case. What is its legal significance?

Plaintiffs describe registered voters who lack photo ID as "disenfranchised." If the reason they lack photo ID is that the state has made it impossible, or even hard, for them to get photo ID, then "disfranchised" might be an apt description. But if photo ID is available to people willing to scrounge up a birth certificate and stand in line at the office that issues drivers' licenses, then all we know from the fact that a particular person lacks a photo ID is that he was unwilling to invest the necessary time. And Crawford tells us that "the inconvenience of making a trip to the [department of motor-vehicles], gathering the required documents, and posing for a photograph surely does not qualify as a substantial burden on the right to vote, or even represent a significant increase over the usual burdens of voting." 553 U.S. at 198, 128 S.Ct. 1610.

Registering to vote is easy in Wisconsin.2 Yet of those eligible, only 78% have registered. (In raw numbers, 4.247 million were eligible in 2012, and of that number only 3.318 million were registered. The difference is almost a million, vastly exceeding the number of registered voters who lack photo ID. U.S. Census Bureau, Reported Voting and Registration by Sex, Race and Hispanic Origin, for States: November 2012 (May 2013).) This proportion is lower than the 91% of registered voters who have qualifying photo ID. We know from registration data (and the fact that not all registered persons cast ballots) that any procedural step filters out some potential voters. No one calls this effect disfranchisement, even though states could make things easier by, say, allowing everyone to register or vote from a computer or smartphone without travel or standing in line. Yet if 22% of the eligible population does not perform even the easiest step, registration, it is difficult to infer from the fact that 9% have not acquired photo ID that that step is particularly difficult. A more plausible inference would be that people who do not plan to vote also do not go out of their way to get a photo ID that would have no other use to them. This does not imply that a need for photo ID is an obstacle to a significant number of persons who otherwise would cast ballots.

Some of the district court's other findings support the conclusion that for most eligible voters not having a photo ID is a matter of choice rather than a state-created obstacle. We have mentioned the court's finding that 2.4% of white adult residents in Milwaukee County do not now have in their possession either a qualifying photo ID or the documentation needed to get one. (This is the same thing as the proposition that 97.6% do have a photo ID or the qualifying documents.) The judge estimated that 4.5% of blacks and 5.9% of Latinos lack both. But if 9% of eligible voters lack a photo ID, this necessarily means that more than half of eligible voters who lack a photo ID do have a birth certificate or other qualifying documents among the family records. (One witness testified that, of persons who lack qualifying photo IDs, 32% also lack the documents needed to get one; this means that 68% of all persons who lack a photo ID could get one without hassle.) If people who already have copies of their birth certificates do not choose to get free photo IDs, it is not possible to describe the need for a birth certificate as a legal obstacle that disfranchises them.

Because the burden of getting a photo ID in Wisconsin is no greater than the burden in Indiana, the district court's constitutional holding must rest on its finding that photo IDs do not serve any important purpose — for if that's right, then under the constitutional standard laid out in Crawford even a modest burden is forbidden.

The district judge concluded that the only kind of fraud that photo IDs address is impersonation of voters at the polls, and he found that impersonation does not happen in Wisconsin. (He allowed that some frauds may go undetected but thought that the number is trivial.) Although the judge recognized that some voter-impersonation frauds had been detected — on one occasion, for example, a man cast an absentee ballot for his deceased wife — the judge thought that a photo ID would not necessarily prevent these. He observed that the man could have submitted a photocopy of his deceased wife's photo ID. The state also contended that requiring identification of voters at the polls promotes public confidence in the integrity of elections, but the judge found that there is no relation between voter-identification statutes and public confidence. It follows, the judge concluded, that Wisconsin's Act 23 serves no legitimate purpose.

One problem with relying on these findings is that the first of them — the conclusion that voter impersonation is rare if not nonexistent — is identical to a finding made in the Indiana litigation. The district judge in Indiana found that there had never been a documented instance of voterimpersonation fraud in that state. The Supreme Court recited this finding, 553 U.S. at 194-96, 128 S.Ct. 1610, yet found it inadequate to conclude that the statute does not serve any purpose. That's because the Supreme Court thought that a photo ID requirement has other benefits (id. at 191-97, 128 S.Ct. 1610): it deters fraud (so that a low frequency stays low); it promotes accurate record keeping (so that people who have moved after the date of registration do not vote in the wrong precinct); it promotes voter confidence. The Court took the last of these as almost self-evidently true. And the need for documentation such as a birth certificate to get a photo ID suggests another benefit: it will prevent some people who should not have registered (because they are too young or not citizens) from voting when they are unable to get a qualifying photo ID. Wisconsin allows registration on election day, and a photo ID can help to verify (or refute) representations a person makes when trying to register.

The dissenting Justices were not impressed by the benefits their colleagues touted. Justice Souter (joined by Justice Ginsburg) heaped scorn on them, deeming them unsubstantiated and at any event too modest to justify an appreciable burden. 553 U.S. at 223-37, 128 S.Ct. 1610 (dissenting opinion). (Justice Breyer, who also dissented, did so because in his view the photo ID requirement discouraged too many people from voting; he did not join Justice Souter's view that the law served no valid purpose.) In this litigation, plaintiffs produced the testimony of a political scientist who agrees with Justice Souter, and the district judge found as a fact that the majority of the Supreme Court was wrong about benefits such as better record keeping and promoting public confidence. Maybe that testimony will eventually persuade the Justices themselves, but in our hierarchical judicial system a district court cannot declare a statute unconstitutional just because he thinks (with or without the support of a political scientist) that the dissent was right and the majority wrong.

To put this in legalese, whether a photo ID requirement promotes public confidence in the electoral system is a "legislative fact" — a proposition about the state of the world, as opposed to a proposition about these litigants or about a single state. Judges call the latter propositions "adjudicative facts." On matters of legislative fact, courts accept the findings of legislatures and judges of the lower courts must accept findings by the Supreme Court. See, e.g., Armour v. Indianapolis, ___ U.S. ___, 132 S.Ct. 2073, 2080, 182 L.Ed.2d 998 (2012); A Woman's Choice-East Side Women's Clinic v. Newman, 305 F.3d 684 (7th Cir.2002).

The district judge heard from one political scientist, whose view may or may not be representative of the profession's. After a majority of the Supreme Court has concluded that photo ID requirements promote confidence, a single district judge cannot say as a "fact" that they do not, even if 20 political scientists disagree with the Supreme Court.

Photo ID laws promote confidence, or they don't; there is no way they could promote public confidence in Indiana (as Crawford concluded) and not in Wisconsin. This means that they are valid in every state — holding constant the burden each voter must bear to get a photo ID — or they are valid in no state. Functionally identical laws cannot be valid in Indiana and invalid in Wisconsin (or the reverse), depending on which political scientist testifies, and whether a district judge's fundamental beliefs (his "priors," a social scientist would say) are more in line with the majority on the Supreme Court or the dissent.

Wisconsin-specific findings do matter to some issues; if the burden of getting a photo ID in Wisconsin were materially greater than the burden in Indiana, then Wisconsin's law could indeed be invalid while Indiana's stands. But no one suggests that photo ID laws promote confidence in Indiana but not Wisconsin; the district court's finding concerns the nation as a whole. (The political scientist who testified at trial relied not on his own work, or even on work in a refereed scholarly journal, but on 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 (2008), which reported the results of one opinion poll of people living throughout the country.)

That photo IDs promote confidence, even if they have no other effect, is widely accepted outside the field of voting. Take the photo ID requirement for boarding an aircraft. As far as we are aware, a need to produce photo ID has never prevented a hijacking or act of terror ism; no one even argues that it has. Magnetometers, x-ray machines, and other technical resources find guns, knives, and explosives. (Find them frequently: many people who possess photo ID try to carry these items onto planes.) But the public feels safer when everyone must show a photo ID, which makes the requirement a rational one. Perhaps that is why both state and federal judiciaries require photo ID of people entering courthouses, even though it is the magnetometers and other technical gear, not the ID, that finds the weapons.

If the public thinks that photo ID makes elections cleaner, then people are more likely to vote or, if they stay home, to place more confidence in the outcomes. These are substantial benefits. One district judge's contrary view is not enough to condemn a state statute as unconstitutional. By contrast, a finding that a photo ID law has significantly reduced the turnout in a particular-state would imply that the requirement's additional costs outweigh any benefit in improving confidence in electoral integrity. As we have observed, however, the judge did not find that photo ID laws measurably depress turnout in the states that have been using them.

We have said enough to demonstrate that Crawford requires us to reject a constitutional challenge to Wisconsin's statute. (The Supreme Court of Wisconsin reached the same conclusion in Milwaukee Branch of NAACP and League of Women Voters v. Walker, 2014 WI 97, 851 N.W.2d 302 (2014), both of which reversed injunctions that had been issued by state judges.) In Crawford plaintiffs relied exclusively on the Constitution; in this suit plaintiffs also contended, and the district judge found, that the state law violates § 2 of the Voting Rights Act, 52 U.S.C. § 10301, formerly 42 U.S.C. § 1973:

The judge recognized that most case law concerning the application of § 2 concerns claims that racial gerrymandering has been employed to dilute the votes of racial or ethnic groups. See, e.g., Thornburg v. Gingles, 478 U.S. 30, 106 S.Ct. 2752, 92 L.Ed.2d 25 (1986); Chisom v. Roemer, 501 U.S. 380, 111 S.Ct. 2354, 115 L.Ed.2d 348 (1991). In Gingles the Justices borrowed nine factors from a Senate committee report (often called the "Gingles factors") as the standard for applying § 2. The judge found that line of cases unhelpful for situations involving eligibility to vote. The judge recognized that a separate line of § 2 cases does involve eligibility and has concluded that felon-disfranchisement statutes do not violate § 2 even though these laws have a disparate impact on minorities. (Both blacks and Latinos are more likely to have felony convictions than are whites.) See 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). But the judge deemed all of those decisions irrelevant too, because most felon-disfranchisement laws predate the Voting Rights Act.

The judge thought that § 2 offers the best guide to its own interpretation and emphasized the rule that laws must not "result[ ] in a denial" of the right to vote. Act 23 has such a result, the judge concluded, because white registered voters are more likely to possess qualifying photo IDs, or the documents necessary to get them. We have mentioned one statistical disparity: 97.6% of whites, 95.5% of blacks, and 94.1% of Latinos currently possess either qualifying photo IDs or the documents that would permit Wisconsin to issue them.3 (In other words, these registered voters have, or can get, photo IDs without asking any public-records office for any additional document, such as a birth certificate.) If instead of asking who has either photo IDs or the documents required to get them, we ask only who had qualifying photo IDs as of the trial, the district judge estimated that 92.7% of whites, 86.8% of blacks, and 85.1% of Latinos did. Finally, the judge found that it would be harder for blacks and Latinos, on average, to get the documents they need, because for the five years ending in 2011 some 75% of Wisconsin's white residents had been born in that state, while only 59% of blacks and 43% of Latinos had been born there. Getting birth certificates from other states is harder than getting them from Wisconsin, the judge found. The decision of the Supreme Court of Wisconsin and the state's new regulations may reduce that burden but cannot eliminate it; persons who rely on the waiver procedure still must apply for it, which means that on average black and Latino residents must file more paperwork than white residents.

Although these findings document a disparate outcome, they do not show a "denial" of anything by Wisconsin, as § 2(a) requires; unless Wisconsin makes it needlessly hard to get photo ID, it has not denied anything to any voter. Nor did the district court find that differences in economic circumstances are attributable to discrimination by Wisconsin. The judge explained his findings this way: "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." ___ F.Supp.3d at ___, 2014 WL 1775432 at *32. The judge did not conclude that the state of Wisconsin has discriminated in any of these respects. That's important, because units of government are responsible for their own discrimination but not for rectifying the effects of other persons' discrimination. See, e.g., Milliken v. Bradley, 418 U.S. 717, 94 S.Ct. 3112, 41 L.Ed.2d 1069 (1974). Section 2(a) forbids discrimination by "race or color" but does not require states to overcome societal effects of private discrimination that affect the income or wealth of potential voters.

Section 2(b)tells us that § 2(a) does not condemn a voting practice just because it has a disparate effect on minorities. (If things were that simple, there wouldn't have been a need for Gingles to list nine non-exclusive factors in vote-dilution cases.) Instead § 2(b) tells us: "A violation of subsection (a) is established if, based on the totality of circumstances, it is shown that the political processes leading to nomination or election in the State or political subdivision are not equally open to participation by members of a class of citizens protected by subsection(a) in that its members have less opportunity than other members of the electorate to participate in the political process" (emphasis added). Act 23 does not draw any line by race, and the district judge did not find that blacks or Latinos have less "opportunity" than whites to get photo IDs. Instead the judge found that, because they have lower income, these groups are less likely to use that opportunity. And that does not violate § 2. In voting-dilution cases, citizens lumped into a district can't extricate themselves except by moving, so clever district-line drawing can disadvantage minorities. But Act 23 extends to every citizen an equal opportunity to get a photo ID.

To the extent outcomes help to decide whether the state has provided an equal opportunity, we must look not at Act 23 in isolation but to the entire voting and registration system. If blacks and Latinos do not get photo IDs at the same frequency as whites, that will reduce their relative share of voting in Wisconsin. By how much? We don't know, because (for reasons we have covered) it may be that the people who do not get photo IDs are also those least likely to vote with or without photo IDs. Experience from other states would help to understand the full effect, but the record lacks that information. But we do know, from data published by the Census Bureau, that blacks do not seem to be disadvantaged by Wisconsin's electoral system as a whole. In 2012 79.6% of Wisconsin's eligible white non-Hispanic residents were registered to vote. That year, 81% of the state's eligible black residents were registered to vote. (Only 46.8% of Latino residents were registered; this might be caused by errors in the data; the Census Bureau provides an 18.4% margin of error for this figure.) In 2012 75% of the state's eligible white non-Hispanic registered voters went to the polls; 78.5% of the state's eligible black voters cast ballots. Even if Act 23 takes 2.1% off this number (the difference between the 97.6% of white voters who already have photo ID or qualifying documents, and the 95.5% of black voters who do), black turnout will remain higher than white turnout.

We are not saying that, as long as blacks register and vote more frequently than whites, a state is entitled to make changes for the purpose of curtailing black voting. Far from it; that would clearly violate § 2. Our point, rather, is that when the validity of the state's voting laws depends on disparate impact, as the district court held, it is essential to look at everything (the "totality of circumstances", § 2(b) says) to determine whether there has been such an impact. Otherwise § 2 will dismantle every state's voting apparatus.

No state has exactly equal registration rates, exactly equal turnout rates, and so on, at every stage of its voting system. At oral argument, counsel for one of the two groups of plaintiffs made explicit what the district judge's approach implies: that if whites are 2% more likely to register than are blacks, then the registration system top to bottom violates § 2; and if white turnout on election day is 2% higher, then the requirement of in-person voting violates § 2. Motor-voter registration, which makes it simple for people to register by checking a box when they get drivers' licenses, would be invalid, because black and Latino citizens are less likely to own cars and there fore less likely to get drivers' licenses. (The district judge cited with approval, ___ F.Supp.3d at ___ n. 32, 2014 WL 1775432 at *27 n. 32, a study concluding that in Milwaukee County 73% of white adults, 47% of black adults, and 43% of Hispanic adults have valid drivers' licenses; this implies an equally large difference in registration rates using the motorvoter protocol.) Yet it would be implausible to read § 2 as sweeping away almost all registration and voting rules. It is better to understand § 2(b) as an equal-treatment requirement (which is how it reads) than as an equal-outcome command (which is how the district court took it).

For the sake of argument, let us put all of the felon-disfranchisement cases to one side, even though they offer strong support for our reading of § 2, in voter-qualification situations, as an equal-treatment requirement. Three appellate opinions have applied § 2 to voter-qualification rules other than felon-disfranchisement statutes: Gonzalez v. Arizona, 677 F.3d 383, 404-10 (9th Cir.2012) (en banc); Ohio State Conference of NAACP v. Husted, No. 14-3877, 768 F.3d 524, 2014 WL 4724703 (6th Cir. Sept. 24, 2014), stayed under the name Husted v. NAACP, No. 14A336, ___ U.S.___, ___ S.Ct.___, ___ L.Ed.2d ___, 2014 WL 4809069 (S.Ct. Sept. 29, 2014); and League of Women Voters of North Carolina v. North Carolina, No. 14-1845, ___ F.3d ___, 2014 WL 4852113 (4th Cir. Oct. 1, 2014). Gonzalez held that Arizona's voter ID statute (which requires voters to present one qualifying photo ID or two qualifying non-photo IDs) is valid under § 2; the court cited Gingles but did not use most of its nine factors or establish an alternative approach. The Fourth Circuit and the Sixth Circuit, by contrast, found Gingles unhelpful in voter-qualification cases (as do we) and restated the statute as calling for two inquiries.

League of Women Voters, ___ F.3d at ___-___, 2014 WL 4852113. We are skeptical about the second of these steps, because it does not distinguish discrimination by the defendants from other persons' discrimination. In vote-dilution cases, the domain of Gingles, the government itself draws the district lines; no one else bears responsibility. But if we were to adopt this approach for the sake of argument, our plaintiffs would fail at the first step, because in Wisconsin everyone has the same opportunity to get a qualifying photo ID.

Photo ID laws have been politically contentious. Crawford remarked on the apparently partisan nature of the disagreement between those who favor and those who oppose these statutes. The lead opinion stated: "if a nondiscriminatory law is supported by valid neutral justifications, those justifications should not be disregarded simply because partisan interests may have provided one motivation for the votes of individual legislators.... The application of the statute to the vast majority of Indiana voters is amply justified by the valid interest in protecting `the integrity and reliability of the electoral process.'" 553 U.S. at 204, 128 S.Ct. 1610. That is true of Wisconsin as well.

One final comment. Even if Act 23 violated § 2 or the Constitution because of its disparate impact on economically disadvantaged voters, the district court's injunction could not be affirmed. It reads:

___ F.Supp.3d at ___, 2014 WL 1775432 at *34. The injunction is perpetual and unconditional. Even if Wisconsin offers a photo ID to everyone registered to vote, without the need for supporting documentation, it still cannot require anyone to present photo ID at a polling place. Under the injunction's language, it is irrelevant how well the changes required by Milwaukee Branch of NAACP or adopted by regulation work in alleviating difficulties that some persons encounter in getting photo IDs.

A district judge's remedial authority is limited to ending the illegal conduct — and the problem identified by the district court is not photo ID in the abstract, but how income and education affect the probability of having photo ID. The injunction should have allowed the state an opportunity to make photo ID more readily available.

Details of the injunction do not matter, however, given our conclusion that Act 23 does not violate either § 2 or the Constitution. The judgment of the district court is reversed.