dissenting from denial of rehearing en banc:
This is a dark day for the Voting Rights Act. In adopting a constitutionally ques*1117tionable interpretation of the Act, the panel lays the groundwork for the dismantling of the most important piece of civil rights legislation since Reconstruction. The panel also misinterprets the evidence, flouts our voting rights precedent and tramples settled circuit law pertaining to summary judgment, all in an effort to give felons the right to vote. The court should have taken this case en banc and brought order back into our caselaw. I dissent from the court’s failure to do so.
1. Plaintiffs’ case is based entirely on statistical disparities: They claim that disparities in the felony conviction rates of certain minority groups in relation to their presence in the general population lead to a disparity in the rate of disenfranchisement under Washington’s felon disenfranchisement law. They argue that these disparities alone prove that under the “totality of the circumstances” they have less of an opportunity to participate in the electoral process, because these disparities interact with the felon disenfranchisement provision and result in denial of the right to vote on account of race.
Though the panel hints otherwise, plaintiffs never produced a shred of evidence of intentional discrimination in Washington’s criminal justice system. The studies and expert testimony plaintiffs rely upon establish only disparities. For example, the Washington State Minority and Justice Commission study of bail and pre-trial detention practices admits that “it would be inappropriate to conclude that racial and ethnic differences in pre-trial release necessarily reflect overt racial bias or discrimination in the decisions of Superior Court judges or staff.” S.E.R. at 221. The same is true of charging and sentencing of felony drug offenders; the report plaintiffs rely upon attributes any disparities there to concerns about resources, not race: “charges are routinely changed between initial filing and conviction” but “these changes are, for the most part, not related to race.” S.E.R. at 243.
Studies based on statistical disparities are notoriously unreliable. See McCleskey v. Kemp, 481 U.S. 279, 107 S.Ct. 1756, 95 L.Ed.2d 262 (1987). Plaintiffs’ own expert, Dr. Bridges, concedes that “[disparities have complex causes and among them are important qualitative differences among defendants in the types of crimes they have committed.” S.E.R. at 221. The bottom line is plaintiffs have produced no evidence that Washington’s criminal justice system is infected with racial bias.
This is significant because the record is settled. The panel admits that “[pjlaintiffs did not dispute the State’s statement of material facts.” Slip op. at 10134. No triable issues of fact remain. The legal question presented is therefore quite simple: Can plaintiffs survive a summary judgment motion in a section 2 vote denial case if they have produced only evidence of statistical disparities in an area external to voting, which then result in statistical disparities in voting? Under Smith v. Salt River Project Agricultural Improvement and Power District, 109 F.3d 586 (9th Cir.1997), the answer is clearly “no.” Yet the panel reverses the district court’s grant of summary judgment. In so doing, it effectively holds that such disparities could be enough to establish vote denial on account of race, in violation of section 2 of the Voting Rights Act (VRA), 42 U.S.C. § 1973. This holding plainly contradicts the law of our circuit and four others. See Salt River, 109 F.3d at 595; Ortiz v. City of Philadelphia Office of the City Comm’rs, 28 F.3d 306, 314-15 (3d Cir.1994); Salas v. Southwest Texas Junior Coll. Dist., 964 F.2d 1542, 1556 (5th Cir.1992); Irby v. Virginia State Bd. of Elections, 889 F.2d 1352, 1358-59 (4th Cir.1989); Wesley v. Collins, 791 F.2d 1255, 1262 (6th Cir.1986).
*1118In Salt River, we held that statistical disparities were not enough to establish vote denial under section 2. We explained that “a bare statistical showing of disproportionate impact on a racial minority does not satisfy the § 2 ‘results’ inquiry” because causation cannot be inferred from impact alone. Id. at 595. We upheld a land-owner voting system against a section 2 challenge because it did not result in discrimination “on account of race or col- or,” id. at 596, even though whites were more likely to have a vote under that system because their rate of home' ownership was much higher than that of blacks, id. at 590. Evidence of racial disparities in the rate of land ownership, which were then mapped directly onto the voter registration rolls, could not support a violation of the VRA. Salt River therefore stands for the principle that plaintiffs cannot prove a section 2 violation without substantial evidence other than a statistical disparity in some area unrelated to voting. There is nothing in the record here beyond statistical disparities, and the facts are settled. Summary judgment for Washington is therefore the only possible outcome.
The Sixth Circuit rejected a section 2 challenge of Tennessee’s felon disenfranchisement law that was based primarily on statistical differences between minority and white convictions. Wesley, 791 F.2d at 1262. The court upheld Tennessee’s statute under the “totality of the circumstances” test, even as it recognized “the presence in Tennessee of certain factors enumerated in the legislative history of Section 2, such as a history of racial discrimination, the effects of which continue to the present day.” Id. at 1261. The court held that “the disproportionate impact suffered by black Tennesseans does not ‘result’ from the state’s qualification of the right to vote on account of race or color.” Id. at 1262. It concluded that section 2 was not violated because the denial of the right to vote was not on account of race but rather on account of the felon’s decision to commit a crime. Id.
The Third Circuit has also made it clear that statistical disparities aren’t enough to support a section 2 vote denial claim. Ortiz upheld a Pennsylvania statute that purged voters’ names from the rolls if they failed to vote for two years, disparately impacting minority voters. 28 F.3d at 307. The court held that statistical disparities weren’t enough; plaintiffs in section 2 vote denial cases must demonstrate a “causal connection between the challenged electoral practice and the alleged discrimination that results in a denial or abridgment of the right to vote.” Id. at 310.
Section 2 vote denial claims based only on disparities are also dead on arrival in the Fourth Circuit. Irby upheld Virginia’s system of appointing members of its school boards. 889 F.2d at 1358-59. Despite the existence of a “significant disparity” between the percentage of blacks in the population and the percentage of blacks on the school board, id. at 1358, the court held that “[t]he evidence cast considerable doubt on the existence of a causal link between the appointive system and black underrepresentation in Buckingham and Halifax counties,” id. at 1359. Rather, the disparity existed because, “although blacks comprise a large portion of the population, they are not seeking school board seats in numbers consistent with their percentage of the population.” Id. at 1358 (internal quotation marks and citations omitted).
Salas also rejected a section 2 vote denial challenge because it was based entirely on statistical disparities. Plaintiffs challenged an at-large voting system based on the disparities in turnout rates of white and minority voters. 964 F.2d at 1556. The evidence revealed that minority voter turnout was about seven percentage points below white voter turnout, but the court held that “a protected class is not entitled *1119to § 2 relief merely because it turns out in a lower percentage than whites to vote.” Id. at 1556. Plaintiffs’ section 2 claim failed because the cause of the disparity in turnout rates wasn’t intentional race discrimination.
2. Under section 2, courts must evaluate the “totality of the circumstances” in which the challenged voting law operates to determine whether plaintiffs have been denied the right to vote based on their race. 42 U.S.C. § 1973(b); Salt River, 109 F.3d at 594. The panel reverses summary judgment although plaintiffs haven’t demonstrated that the “totality of the circumstances” point to vote denial on account of race. Plaintiffs have not produced any evidence relating to the nine Senate factors we must weigh under that test, S.Rep. No. 97-417, at 28-29 (1982), reprinted in 1982 U.S.C.C.A.N. 177, 206-07, beyond statistical disparities in the criminal justice system. They have no evidence of a history of official discrimination in voting, no evidence of racially polarized voting, no evidence of voting practices or procedures often used to discriminate against minorities, no evidence of discrimination in candidate slating, no evidence of discrimination in health, education or employment, no evidence of racial appeals in campaigns, no evidence that minorities have a harder time winning elections, no evidence that representatives are unresponsive to minority communities and no evidence that felon disenfranchisement is an unjustified policy. Plaintiffs have utterly failed to meet their burden of producing evidence showing vote denial on account of their race.
Plaintiffs in other cases have lost section 2 vote denial claims even though they produced much stronger evidence than the plaintiffs here. In Ortiz, the Third Circuit upheld under the “totality of the circumstances” Pennsylvania’s voter purge statute though it disproportionately impacted minority voters. The Ortiz plaintiffs had evidence of racially polarized voting, racial appeals in elections, unresponsiveness by elected officials, lower minority voter turnout and disparities in employment and other non-voting areas. Id. at 312. Nonetheless, the fact that “there was no evidence of historical voting-related discrimination ... no evidence of discrimination in the candidate slating process that denied minority candidates equal access to the political process .... [and no] evidence that minorities experience difficulty in electing representatives of their choice,” id., led the court to conclude that section 2 had not been violated, id. at 313. See also Burton v. City of Belle Glade, 178 F.3d 1175 (11th Cir.1999) (rejecting a section 2 challenge even in the face of past intentional discrimination).
The panel tries to pass off evidence of disparities as evidence of intentional discrimination, but the two have entirely different consequences for vote denial claims. Intentional discrimination in the criminal justice system, if it interacts with a standard, practice or procedure with respect to voting, could amount to illegal vote denial on account of race. See Johnson v. Governor of the State of Florida, 353 F.3d 1287 (11th Cir.2003). To the extent the district court’s decision granting defendants’ motion for summary judgment was based on its misunderstanding of section 2 and belief that evidence of intentional discrimination external to voting could never be taken into account, it was wrong. However, the result it reached was correct because, even under the correct standard, plaintiffs have not produced evidence of intentional discrimination to survive summary judgment.
3. Because the district court’s decision granting summary judgment was correct, even if for the wrong reasons, it must be affirmed. The panel’s decision to reverse and remand' — on a settled record contain*1120ing nothing more than evidence of disparities' — flies in • the face of a solid wall of circuit precedent holding that “[i]f the decision below is correct, it must be affirmed, even if the district court relied on the wrong grounds or wrong reasoning.” Jackson v. S. California Gas Co., 881 F.2d 638, 643 (9th Cir.1989) (emphasis added). Accord Cigna Prop. & Cas. Ins. Co. v. Polaris Pictures Corp., 159 F.3d 412, 418 (9th Cir.1998); Unigard Sec. Ins. v. Lakewood Eng’g & Mfg. Corp., 982 F.2d 363, 367 (9th Cir.1992). Moreover, the panel’s remand order is hopelessly ambiguous, directing the district court to “make any requisite factual findings following an appropriate evidentiary hearing, if necessary, and assess the totality of the circumstances.” Slip op. at 10147-48. There is no authority for such a sloppy remand order, which tells the district court it may reconsider summary judgment or make findings after an evidentiary hearing, without any instruction as to when such a hearing might be appropriate. The panel misunderstands its task at this stage. Under Celotex Corp. v. Catrett, 477 U.S. 317, 106 S.Ct. 2548, 91 L.Ed.2d 265 (1986), a party is entitled to summary judgment if the opposing party has failed to present a genuine dispute of fact on a material issue. Id. at 322-23, 106 S.Ct. 2548. Because plaintiffs have failed to present evidence of intentional discrimination, summary judgment is the only option.
4. The panel’s decision suffers from a more fundamental flaw: It ignores the fact that the VRA was never intended to "reach felon disenfranchisement laws. When Congress enacted the VRA in 1965, it was careful to carve out an exception for felon disenfranchisement laws. The House and Senate Judiciary Committee Reports clearly bear this out. The Senate Report explains that the ban on “good moral character” tests contained in section 1973b(c) of the statute “would not result in the proscription of the frequent requirement of States and political subdivisions that an applicant for voting or registration for voting be free of conviction of a felony.... It applies where lack of good moral character is defined in terms of a conviction of lesser crimes.” S.Rep. No. 89-162, reprinted in 1965 U.S.C.C.A.N. 2508, 2562 (joint views of Senators Dodd, Hart, Long, Kennedy, Bayh, Burdick, Tydings, Dirksen, Hruska, Fong, Scott and Javits). The House Report also explains that the Act does “not proscribe a requirement of a State or any political subdivision of a State that an applicant for voting or registration for voting be free of conviction of a felony.” H.R.Rep. No. 89-439, reprinted in 1965 U.S.C.C.A.N. 2437, 2457. This legislative history demonstrates that Congress recognized the long tradition of felon disenfranchisement laws when it. enacted the VRA. Moreover, it distinguishes those provisions from other electoral qualifications that the VRA has been held to reach, such as literacy tests, which do not in and of themselves violate the Constitution. See, e.g., Katzenbach v. Morgan, 384 U.S. 641, 645 n. 3 & 646-47, 86 S.Ct. 1717, 16 L.Ed.2d 828 (1966) (upholding ban on literacy tests targeted to deal with disenfranchisement of Puerto Ricans in New York even after upholding their constitutionality in Lassiter v. Northampton Elections Board, 360 U.S. 45, 79 S.Ct. 985, 3 L.Ed.2d 1072 (1959)); see also Baker v. Pataki, 85 F.3d 919, 928 (2d Cir.1996) (opinion of Mahoney, J.).
When Congress enacted the “results” test as part of the 1982 amendments to the VRA, it was primarily concerned with whether section 5 of the Act — the provision placing special requirements on jurisdictions with a history of past discrimination — was still necessary. The debate surrounding section 2 focused mainly on whether the results test would require proportional representation by race. See generally S.Rep. No. 97-417 (1982), re*1121printed, in 1982 U.S.C.C.A.N. 177; Jennifer G. Presto, The 1982 Amendments to Section 2 of the Voting Rights Act: Constitutionality After City of Boerne, 59 N.Y.U. Ann. Surv. Am. L. (forthcoming 2004). There was no evidence that felon disenfranchisement laws were being used in a discriminatory manner, nor any discussion of felon disenfranchisement at all. There is thus no evidence that Congress had changed its mind about the legitimacy of felon disenfranchisement when it enacted section 2. See Johnson, 353 F.3d at 1316-18 (Kravitch, J., dissenting) (demonstrating that Congress never intended the VRA to reach felon disenfranchisement laws).
In fact, Congress has since enacted laws making it easier for states to keep felons off the voting roster. In 1993, Congress enacted the National Voter Registration Act (NVRA), Pub. L. No. 103-31, 107 Stat. 77 (1993), which clearly lists conviction of a felony as a justification for cancellation of a voters’ registration. 42 U.S.C. § 1973gg-6(a)(3)(B). Congress even instructed federal prosecutors, “[o]n the conviction of a person of a felony in a district court of the United States,” to “give written notice of the conviction to the chief State election official.” 42 U.S.C. § 1973gg-6(g). The NVRA also provides that “[o]n request of the chief State election official of a State or other State official with responsibility for determining the effect that a conviction may have on an offender’s qualification to vote, the United States attorney shall provide such additional information as the United States may have concerning the offender and the offense of which the offender was convicted.” 42 U.S.C. § 1973gg-6(g)(3).
Congress drafted federal prosecutors to help states disenfranchise felons even as it enacted the NVRA to extirpate certain registration practices that dampen minority participation in the political process. 42 U.S.C. § 1973gg(a)(3). It’s thus crystal clear that felon disenfranchisement wasn’t one of the practices about which Congress was concerned. We are bound to respect that legislative judgment — not override it.
5. There is yet a more fundamental problem with extending the VRA to reach felon disenfranchisement laws: Doing so seriously jeopardizes its constitutionality.
“[A]ny attempt by Congress to subject felon disenfranchisement provisions to the ‘results’ methodology of [the VRA] would pose a serious constitutional question concerning the scope of Congress’s power to enforce the Fourteenth and Fifteenth Amendments,” Baker, 85 F.3d at 930, because “felon disenfranchisement is a very widespread historical practice that has been accorded explicit constitutional recognition in § 2 of the Fourteenth Amendment,” id. at 928; see also Richardson v. Ramirez, 418 U.S. 24, 54, 94 S.Ct. 2655, 41 L.Ed.2d 551 (1974).
Unlike any other voting qualification, felon disenfranchisement laws are explicitly endorsed by the text of the Fourteenth Amendment. Id.; see also Johnson, 353 F.3d at 1314-16 (Kravitch, J., dissenting); Baker, 85 F.3d at 930. They are presumptively constitutional. Only a narrow subset of them- — those enacted with an invidious, racially discriminatory purpose — is unconstitutional. Hunter v. Underwood, 471 U.S. 222, 233, 105 S.Ct. 1916, 85 L.Ed.2d 222 (1985). If Congress had intended the VRA to reach felon disenfranchisement laws, the only constitutional violations it had authority to remedy are the type recognized in Hunter: the purposeful, invidious use of those laws to deprive minorities of the right to vote.
Hunter violations, however, are exceedingly rare. See Developments in the Law — The Law of Prisons, 115 Harv. L. Rev. 1939, 1951-52 (2002). Hunter involved a provision of the Alabama Consti*1122tution of 1901 that denied the right to vote to persons who had committed crimes of “moral turpitude.” The 1901 Constitution was an egregious example of post-Reconstruction disenfranchising constitutions. Accordingly, plaintiffs were able to produce a mountain of evidence of discriminatory intent in its passage, together with evidence of a continuing discriminatory effect in its application, and the Court struck it down as a violation of the Fourteenth Amendment. Alabama’s provision, with its dubious provenance, stands far apart from the legitimate, neutral provision of the Washington Constitution at issue in this case. Washington does not share Alabama’s long history of race discrimination. It has never been a covered state under section 5 of the VRA. Its felon disenfranchisement provision dates back to 1866, see State v. Collins, 69 Wash. 268, 270-71, 124 P. 903 (1912), four years before the Fifteenth Amendment extended the right to vote to blacks. Therefore Washington’s felon disenfranchisement provision couldn’t have been enacted with the intent of depriving minorities of the vote. In fact, plaintiffs’ constitutional claim was so weak that it was dropped shortly after it was made.
In City of Boerne v. Flores, 521 U.S. 507, 117 S.Ct. 2157, 138 L.Ed.2d 624 (1997), the Supreme Court warned that legislation enforcing constitutional rights— such as the VRA — can do no more than remedy or prevent actual constitutional violations, as defined by the Supreme Court. Id. at 519, 117 S.Ct. 2157. Congress cannot alter the substantive constitutional right in an attempt to enforce it. Id. To ensure that Congress is merely enforcing — not altering — 'the substantive right, courts must police the “congruence and proportionality between the injury to be prevented or remedied and the means adopted to that end.” Id. at 520, 117 S.Ct. 2157.
Time and again, the Supreme Court has insisted that section 5 enforcement legislation be supported by a record of constitutional violations. In Oregon v. Mitchell, 400 U.S. 112, 91 S.Ct. 260, 27 L.Ed.2d 272 (1970), the Court invalidated an amendment to the VRA which would have extended the vote to 18-year-olds in state and local elections. Justice Black, who announced the judgment of a fractured Court, explained: “Congress had before it a long history of the discriminatory use of literacy tests to disenfranchise voters on account of their race,” id. at 132, 91 S.Ct. 260, but “Congress made no legislative findings that the 21-year-old vote requirement was used by the States to disenfranchise voters on account of race,” id. at 130, 91 S.Ct. 260. Securing the right to vote for 18-year-olds required passage of the Twenty-Sixth Amendment in 1971.
More recently, in Board of Trustees of the University of Alabama v. Garrett, 531 U.S. 356, 374, 121 S.Ct. 955, 148 L.Ed.2d 866 (2001), the Court held that the Americans with Disabilities Act was an invalid attempt to enforce the Fourteenth Amendment against the states because Congress did not make findings establishing a pattern of unconstitutional irrational job discrimination by the states. Likewise, Kimel v. Florida Board of Regents, 528 U.S. 62, 91, 120 S.Ct. 631, 145 L.Ed.2d 522 (2000), held that Congress failed to amass sufficient evidence of widespread irrational age discrimination in violation of the Fourteenth Amendment to justify application of the Age Discrimination in Employment Act to the states. See also Fla. Prepaid Postsecondary Educ. Expense Bd. v. Coll. Savs. Bank, 527 U.S. 627, 119 S.Ct. 2199, 144 L.Ed.2d 575 (1999). The Court has demonstrated that it’s serious about policing the limits on Congress’s enforcement powers, and we should be, too.
*1123Congress could legitimately remedy Hunter-type violations, but it would have to have evidence that states were using felon disenfranchisement in a purposeful, invidious manner. However, “not only has Congress failed ever to make a legislative finding that felon disenfranchisement is a pretext ... for racial discrimination[,] it has effectively determined that it is - not.” Baker, 85 F.3d at 929. This sharply distinguishes felon disenfranchisement provisions from literacy tests because “[t]he legislative history of the 1970 Amendments [banning literacy tests] contains substantial information upon which Congress could have based a finding that the use of literacy tests ... deny[ ] the vote to racial minorities.” Oregon v. Mitchell, 400 U.S. at 234, 91 S.Ct. 260 (Brennan, J., concurring).
Here, there is no “pattern of unconstitutional discrimination on which § 5 [enforcement] legislation must be based.” Garrett, 531 U.S. at 370, 121 S.Ct. 955; cf. Nev. Dept. of Human Res. v. Hibbs, 538 U.S. 721, 123 S.Ct. 1972, 1981, 155 L.Ed.2d 953 (2003) (upholding the FMLA because Congress had a sufficient evidentiary basis for that legislation). Congress was apparently confident enough in the legitimacy of state laws disenfranchising felons to expressly sanction cancellation of a voter’s registration based on conviction of a felony and direct federal prosecutors to help states in that endeavor in the 1993 National Voter Registration Act. See p. 1122 supra. Without a record proving that most of the state felon disenfranchisement laws prohibited by section 2 would be unconstitutional on their own, the vast overinclu-siveness of section 2 as interpreted by the panel dooms it. See Boerne, 521 U.S. at 532, 117 S.Ct. 2157. The theoretical, undocumented threat of unconstitutional felon disenfranchisement laws -simply doesn’t justify such a broad remedy.
In contrast to the expansive reading of section 2 adopted by the panel today, our parsimonious interpretation of that provision in Salt River preserved its constitutionality. A state whose criminal justice system is infected with actual intentional race discrimination might be more likely to have an unconstitutional felon disenfranchisement law. “[PJurposeful governmental discrimination outside the electoral system might play out within the electoral system, where it would be observed in the disparate impact of otherwise acceptable policies.... Congress could ... remedy the effects of that impermissible prior discrimination.” Pamela S. Karlan, Two Section Twos and Tivo Section Fives: Voting Rights and Remedies After Flores, 39 Wm. & Mary L. Rev. 725, 728-29 (1998) (emphasis added). Section 2 is therefore a more congruent and proportional remedy if plaintiffs are required to produce evidence of intentional discrimination in an area external to voting which interacts with a voting practice to result in the denial of the right to vote on account of race. By allowing plaintiffs to survive summary judgment on a settled record containing nothing but disparities in the criminal justice system, and absolutely no evidence of intentional discrimination, the panel destroys section 2’s congruence and proportionality as a remedy for the kind of constitutional violations recognized in Hunter.
■Section 2 might also be on firmer constitutional footing if it were tailored to areas with a history of discriminatory voting practices. For example, its congruence and proportionality might be stronger if plaintiffs had to produce evidence of a history of discrimination, not just the panel’s requirement of a simple disparate impact. States with a long history of discrimination in voting might be more likely to use felon disenfranchisement laws to deprive minorities of the vote.
*1124The Supreme Court has emphasized that enforcement legislation should be geographically targeted when the threat of violations varies from place to place. In United States v. Morrison, 529 U.S. 598, 120 S.Ct. 1740, 146 L.Ed.2d 658 (2000), the Court struck down the Violence Against Women Act, noting that it “applies uniformly throughout the Nation. Congress’s findings indicate that the problem of discrimination against the victims of gender-motivated crimes does not exist in all States, or even most States. By contrast, the § 5 remedy upheld in Katzenbach v. Morgan[, 384 U.S. 641, 86 S.Ct. 1717, 16 L.Ed.2d 828 (1966)] ... was directed only to the States where the evil found by Congress existed, and in South Carolina v. Katzenbach[, 383 U.S. 301, 86 S.Ct. 803, 15 L.Ed.2d 769 (1966)] ... the remedy was directed only to those States in which Congress found that there had been discrimination.” Morrison, 529 U.S. at 626-27, 120 S.Ct. 1740. Boerne also cited the special-coverage provisions of section 5 of the VRA as an example of a constitutional enforcement measure required by the threat of purposeful discrimination and emphasized its “confine[ment] to those regions of the country where voting discrimination had been most flagrant.” 521 U.S. at 532-33, 117 S.Ct. 2157. In applying section 2 to felon disenfranchisement laws, the panel has handed us a sweeping remedy of its own creation that, like the statute invalidated in Boerne, “is so out of proportion to a supposed remedial or preventive object that it cannot be understood as responsive to, or designed to prevent, unconstitutional behavior. It appears, instead, to attempt a substantive change in constitutional protections.” Boerne, 521 U.S. at 532, 117 S.Ct. 2157.
It is unlikely that Congress could have reached felon disenfranchisement even if it wanted to, at least not without a substantial evidentiary record and a more tailored remedy. In interpreting the VRA to reach felon disenfranchisement in a state without a history of race discrimination like Washington, the panel has created a constitutional problem that Congress itself avoided.
The panel’s decision seriously compromises the constitutionality of section 2. Despite a 1984 case summarily affirming a district court decision upholding its constitutionality, see Mississippi Republican Executive Comm. v. Brooks, 469 U.S. 1002, 105 S.Ct. 416, 83 L.Ed.2d 343 (1984), section 2’s constitutionality remains an open question. See, e.g., Bush v. Vera, 517 U.S. 952, 990, 116 S.Ct. 1941, 135 L.Ed.2d 248 (1996) (O’Connor, J., concurring); Johnson v. De Grandy, 512 U.S. 997, 1028-29, 114 S.Ct. 2647, 129 L.Ed.2d 775 (1994) (Kennedy, J., concurring); Chisom v. Roemer, 501 U.S. 380, 418, 111 S.Ct. 2354, 115 L.Ed.2d 348 (1991) (Kennedy, J., dissenting). See generally Holder v. Hall, 512 U.S. 874, 891, 114 S.Ct. 2581, 129 L.Ed.2d 687 (1994) (Thomas, J., concurring in judgment, joined by Scalia, J.).
In recent years, the Court has repeatedly rejected broad interpretations of the VRA, obviously troubled by the constitutional implications. Reno v. Bossier Parish School Board, 528 U.S. 320, 120 S.Ct. 866, 145 L.Ed.2d 845 (2000), refused to adopt a possible interpretation of section 5 of the VRA because “[s]uch a reading would also exacerbate the ‘substantial’ federalism costs that the preclearance procedure already exacts” and raise concerns about the constitutionality of that provision. Id. at 336, 120 S.Ct. 866. In Vera, the Court held that the VRA does not require excessive reliance on race in dis-tricting because that would offend equal protection. 517 U.S. at 979-81, 116 S.Ct. 1941. Likewise, Miller v. Johnson, 515 U.S. 900, 115 S.Ct. 2475, 132 L.Ed.2d 762 (1995), rejected the Justice Department’s interpretation of section 5 of the VRA and *1125held that “[t]here is no indication Congress intended such a far-reaching application of § 5, so we reject the [proposed] interpretation of the statute and avoid the constitutional problems that interpretation raises.” Id. at 927, 115 S.Ct. 2475. See generally Holder, 512 U.S. at 885, 114 S.Ct. 2581 (O’Connor, J., concurring in part and concurring in the judgment); Presley v. Etowah County Comm’n, 502 U.S. 491, 112 S.Ct. 820, 117 L.Ed.2d 51 (1992). Jurists and commentators alike recognize the constitutional pitfalls of interpreting section 2 too broadly. See, e.g., Johnson, 353 F.3d at 1314 (Kravitch, J., dissenting); Goosby v. Town of Hempstead, 180 F.3d 476, 502 n. 4 (2d Cir.1999) (Leval, J. concurring) (“Requiring discriminatory intent to prove vote' dilution [claims under section 2] reduces the otherwise serious tension between section 2 and constitutional principles.”); Theane Evan-gelis, The Constitutionality of Compensating for Low Minority Voter Turnout in Districting, 77 N.Y.U. L. Rev. 796, 798 (2002); Heather K. Gerken, Understanding the Right to an Undiluted Vote, 114 Harv. L. Rev. 1663, 1737 (2001); Karlan, supra, at 725-26; Douglas Laycock, Conceptual Gulfs in City of Boerne v. Flores, 39 Wm. & Mary L. Rev. 743, 749-52 (1998). See also Samuel Issacharoff, Richard H. Pildes & Pamela S. Karlan, The Law of Democracy: Legal Structure of the Political Process 741-42, 859-66 (2001). It is our job to side-step those pitfalls whenever possible.
We have a duty to recognize limitations on congressional power and avoid interpreting statutes in a way that would extend these powers beyond constitutional limits. “[W]here a statute is susceptible of two constructions, by one of which grave and doubtful constitutional questions arise and by the other of which such questions are avoided, our duty is to adopt the latter.” United States ex rel. Attorney General v. Del. & Hudson Co., 213 U.S. 366, 408, 29 S.Ct. 527, 53 L.Ed. 836 (1909); see also Feltner v. Columbia Pictures Television, Inc., 523 U.S. 340, 356, 118 S.Ct. 1279, 140 L.Ed.2d 438 (1998) (Scalia, J., concurring in judgment). My colleagues fail in their duty not to adopt a constitutionally deficient interpretation of the VRA.
6. The panel’s decision has far-reaching consequences. As long as there are statistical disparities by race in the criminal justice system and consequently in the rate of felony conviction, the panel’s interpretation of the VRA will require states to erect voting booths in prisons. This result is inevitable, as there is no stopping point to the .panel’s rationale. If states can’t exclude felons formerly incarcerated from the franchise, then they surely can’t exclude felons currently behind bars. Once felons have a right to vote, someone will bring suit to require the states to bring the polls to them, since they can’t go to the polls themselves.
Yet every state in our circuit — indeed, every- state in the country save Maine and Vermont — does not allow imprisoned felons to vote. Human Rights Watch, The Sentencing Project, Felony Disenfranchisement Laws in the United States, available at http://www.sentencing project.org/pdfs/1046.pdf. Arizona, Nevada and Washington do not allow ex-felons to vote. Alaska and California also prohibit paroled felons from voting, and Alaska extends that prohibition to those on probation. Moreover, race-based statistical disparities in the rate of felon disenfranchisement exist in every state in our circuit. Human Rights Watch, The Sentencing Project, Losing the Vote, at 9, available at http://sentencingproject. org/ pdfs/9080.pdf.
The panel’s decision also has widespread implications for other legitimate state electoral practices. All sorts of state and local *1126decisions about the time, place and manner of elections will be subject to attack by anyone who can show a disparate impact in an area external to voting that translates into a disparate impact on voting.
For example, according to the 2000 census, minorities have lower incomes than whites. Carmen DeNavas-Walt et al., Money Income in the United States: 2000, at 2 tbl.A, available at http://ict. cas.psu.edu/resources/Census/PDF/C2K_ Income Jm_USA.pdf. People with lower incomes are less likely to participate in the political process. See R. Michael Alvarez & Jonathan Nagler, The Likely Consequences of Internet Voting for Political Representation, 34 Loy. L.A. L. Rev. 1115, 1121 (2001). Evidence of socioeconomic disparities could be the source of countless lawsuits. Plaintiffs will doubtless claim that those disparities lead to disparities in turnout, which lead to a disparate impact on minority voters, if states choose to purge their voter registration lists of the names of voters who have not recently participated in order to combat voting fraud. This is precisely the claim rejected by the Third Circuit in Ortiz, but the panel’s decision would make it possible in our circuit.
After the panel’s decision, plaintiffs could bring a section 2 challenge based on statistical disparities if states adopt Internet voting, which Arizona already tested in the 2000 Democratic presidential primary. See Stephen B. Pershing, The Voting Rights Act in the Internet Age: An Equal Access Theory for Interesting Times, 34 Loy. L.A. L. Rev. 1171, 1172 (2001); see also Eben Moglen & Pamela S. Karlan, The Soul of a New Political Machine: The Online, the Color Line and Electronic Democracy, 34 Loy. L.A. L. Rev. 1089, 1089 (2001) (“The digital divide means that minority citizens have less access to web-based sources of political information and may be less able to use voting techniques, such as online voting, that require a computer.”); Alvarez & Nagler, supra, at 1135. Plaintiffs could show disparities in wealth, leading to disparities in computer ownership and Internet access, leading to disparities in participation on election day. The authors of a study of Arizona’s test-drive of Internet voting concluded that “non-white voters did not vote on the Internet as often as whites, so the Internet voting option seems unlikely to improve the voting rights of minorities. Instead, Internet voting seems likely to weaken the voting rights of minorities, as in this particular case minority turnout dropped substantially more than did white turnout.” Alvarez & Nagler, supra, at 1147.
Holding elections on a Tuesday could be a thing of the past if a plaintiff somewhere can show that minority voters are disproportionately more likely to be hourly wage earners, who are disproportionately less likely to vote because they can’t take time off from work. See Nat’l Comm’n on Fed. Election Reform, To Assure Pride and Confidence in the Electoral Process 40-42 (2001), available at http://www.reformelec-tions.org/data/reports/99_full_report.pdf. Under the panel’s rationale, elections would have to be held on weekends or holidays.
The permutations are endless. The bottom line is that virtually every decision by a state as to voting practices will be vulnerable, no matter how unrelated to race. The fallout from the panel’s decision will be felt for a long time to come.
* * *
Every state in our circuit bars felons from the voting booth. The panel’s decision will change all that. It contradicts our case law and the law of at least four other circuits, making us an outlier in voting rights jurisprudence. It does so without so much as acknowledging congres*1127sional approval of felon disenfranchisement and without any consideration of the grave constitutional consequences of its actions. I am troubled not only by my colleagues’ insistence on an indefensible interpretation of the Voting Rights Act, but also by their utter disregard for our precedent. I dissent.