Shelby County, Ala. v. Holder

United States Court of Appeals FOR THE DISTRICT OF COLUMBIA CIRCUIT Argued January 19, 2012 Decided May 18, 2012 No. 11-5256 SHELBY COUNTY, ALABAMA, APPELLANT v. ERIC H. HOLDER, JR., IN HIS OFFICIAL CAPACITY AS ATTORNEY GENERAL OF THE UNITED STATES, ET AL., APPELLEES Appeal from the United States District Court for the District of Columbia (No. 1:10-cv-00651) Bert W. Rein argued the cause for appellant. With him on the briefs were William S. Consovoy, Thomas R. McCarthy, and Brendan J. Morrissey. John C. Neiman Jr., Solicitor General, Office of the Attorney General for the State of Alabama, and Robert D. Tambling, Assistant Attorney General, were on the brief for amicus curiae State of Alabama in support of appellant. Thomas C. Horne, Attorney General, Office of the Attorney General for the State of Arizona, David R. Cole, Solicitor General, Michele L. Forney and James E. Barton II, Assistant Attorneys General, and Samuel S. Olens, Attorney 2 General, Office of the Attorney General of the State of Georgia, were on the brief for amici curiae States of Arizona and Georgia. Steven J. Lechner was on the brief as amicus curiae Mountain States Legal Foundation in support of appellant. Sarah E. Harrington, Attorney, U.S. Department of Justice, argued the cause for appellee. With her on the brief were Ronald C. Machen Jr., U.S. Attorney, and Diana K. Flynn and Linda F. Thome, Attorneys. Eric T. Schneiderman, Attorney General, Office of the Attorney General for the State of New York, Barbara D. Underwood, Solicitor General. Jim Hood, Attorney General, Office of the Attorney General for the State of Mississippi, and Kamala D. Harris, Attorney General, Office of the Attorney General for the State of California, were on the brief for amici curiae New York, et al., in support of appellees. John Payton, Debo P. Adegbile, Elise C. Boddie, Ryan P. Haygood, Dale E. Ho, Natasha M. Korgaonkar, Arthur B. Spitzer, Jon M. Greenbaum, and John M. Nonna were on the brief for intervenors-appellees Earl Cunningham, et al., in support of appellees. Deborah N. Archer and Aderson B. Francois were on the brief for amicus curiae The New York Law School Racial Justice Project in support of appellee. Elizabeth B. Wydra was on the brief for amicus curiae Constitutional Accountability Center in support of appellees. 3 Before: TATEL and GRIFFITH, Circuit Judges, and WILLIAMS, Senior Circuit Judge. Opinion for the Court filed by Circuit Judge TATEL. Dissenting opinion filed by Senior Circuit Judge WILLIAMS. TATEL, Circuit Judge: In Northwest Austin Municipal Utility District No. One v. Holder, 129 S. Ct. 2504 (2009), the Supreme Court raised serious questions about the continued constitutionality of section 5 of the Voting Rights Act of 1965. Section 5 prohibits certain “covered jurisdictions” from making any change in their voting procedures without first demonstrating to either the Attorney General or a three-judge district court in Washington that the change “neither has the purpose nor will have the effect of denying or abridging the right to vote on account of race or color.” 42 U.S.C. § 1973c(a). The Supreme Court warned that the burdens imposed by section 5 may no longer be justified by current needs and that its geographic coverage may no longer sufficiently relate to the problem it targets. Although the Court had no occasion to resolve these questions, they are now squarely before us. Shelby County, Alabama, a covered jurisdiction, contends that when Congress reauthorized section 5 in 2006, it exceeded its enumerated powers. The district court disagreed and granted summary judgment for the Attorney General. For the reasons set forth in this opinion, we affirm. I. The Framers of our Constitution sought to construct a federal government powerful enough to function effectively yet limited enough to preserve the hard-earned liberty fought 4 for in the War of Independence. They feared not state government, but centralized national government, long the hallmark of Old World monarchies. As a result, “[t]he powers delegated by the . . . Constitution to the federal government, are few and defined,” while “[t]hose which are to remain in the State governments are numerous and indefinite.” The Federalist No. 45 (James Madison). Close to the people, state governments would protect their liberties. But the experience of the nascent Republic, divided by slavery, taught that states too could threaten individual liberty. So after the Civil War, the Reconstruction Amendments were added to the Constitution to limit state power. Adopted in 1865, the Thirteenth Amendment prohibited involuntary servitude. Adopted three years later, the Fourteenth Amendment prohibited any state from “depriv[ing] any person of life, liberty, or property, without due process of law” or “deny[ing] to any person within its jurisdiction the equal protection of the laws,” and granted Congress “power to enforce” its provisions “by appropriate legislation.” U.S. Const. amend. XIV. Finally, the Fifteenth Amendment declared that “[t]he right of citizens . . . to vote shall not be denied or abridged by the United States or by any State on account of race, color, or previous condition of servitude” and vested Congress with “power to enforce this article by appropriate legislation.” U.S. Const. amend. XV. Following Reconstruction, however, “the blight of racial discrimination in voting . . . infected the electoral process in parts of our country for nearly a century.” South Carolina v. Katzenbach, 383 U.S. 301, 308 (1966). As early as 1890, “the States of Alabama, Georgia, Louisiana, Mississippi, North Carolina, South Carolina, and Virginia” began employing tests and devices “specifically designed to prevent Negroes 5 from voting.” Id. at 310. Among the most notorious devices were poll taxes, literacy tests, grandfather clauses, and property qualifications. See Shelby Cnty. v. Holder, 811 F. Supp. 2d 424, 428 (D.D.C. 2011); see also Katzenbach, 383 U.S. at 310–11. Also widely employed, both immediately following Reconstruction and again in the mid-twentieth century, were “laws designed to dilute black voting strength,” including laws that “gerrymandered election districts, instituted at-large elections, annexed or deannexed land . . . and required huge bonds of officeholders.” Shelby Cnty., 811 F. Supp. 2d at 429 (internal quotation marks omitted). The courts and Congress eventually responded. The Supreme Court struck down grandfather clauses, Guinn v. United States, 238 U.S. 347 (1915), and white primaries, Smith v. Allwright, 321 U.S. 649 (1944). Congress “enact[ed] civil rights legislation in 1957, 1960, and 1964, which sought to ‘facilitat[e] case-by-case litigation against voting discrimination.’ ” Shelby Cnty., 811 F. Supp. 2d at 430 (alteration in original) (quoting Katzenbach, 383 U.S. at 313). But Congress soon determined that such measures were inadequate: case-by-case litigation, in addition to being expensive, was slow—slow to come to a result and slow to respond once a state switched from one discriminatory device to the next—and thus had “done little to cure the problem of voting discrimination.” Katzenbach, 383 U.S. at 313. Determined to “rid the country of racial discrimination in voting,” id. at 315, Congress passed the Voting Rights Act of 1965. Unlike prior legislation, the 1965 Act combined a permanent, case-by-case enforcement mechanism with a set of more stringent, temporary remedies designed to target 6 those areas of the country where racial discrimination in voting was concentrated. Section 2, the Act’s main permanent provision, forbids any “standard, practice, or procedure” that “results in a denial or abridgment of the right of any citizen of the United States to vote on account of race or color.” 42 U.S.C. § 1973(a). Applicable nationwide, section 2 enables individuals to bring suit against any state or jurisdiction to challenge voting practices that have a discriminatory purpose or result. See Thornburg v. Gingles, 478 U.S. 30, 35 (1986). Reaching beyond case-by-case litigation and applying only in certain “covered jurisdictions,” section 5—the focus of this litigation—“prescribes remedies . . . which go into effect without any need for prior adjudication.” Katzenbach, 383 U.S. at 327–28. Section 5 suspends “all changes in state election procedure until they [are] submitted to and approved by a three-judge Federal District Court in Washington, D.C., or the Attorney General.” Nw. Austin, 129 S. Ct. at 2509. A jurisdiction seeking to change its voting laws or procedures must either submit the change to the Attorney General or seek preclearance directly from the three-judge court. If it opts for the former and if the Attorney General lodges no objection within sixty days, the proposed law can take effect. 42 U.S.C. § 1973c(a). But if the Attorney General lodges an objection, the submitting jurisdiction may either request reconsideration, 28 C.F.R. § 51.45(a), or seek a de novo determination from the three-judge district court. 42 U.S.C. § 1973c(a). Either way, preclearance may be granted only if the jurisdiction demonstrates that the proposed change to its voting law neither “has the purpose nor . . . the effect of denying or abridging the right to vote on account of race or color.” Id. Prior to section 5’s enactment, states could stay ahead of plaintiffs and courts “ ‘by passing new discriminatory voting 7 laws as soon as the old ones had been struck down.’ ” Beer v. United States, 425 U.S. 130, 140 (1976) (quoting H.R. Rep. No. 94-196, at 57–58 (1975)). But section 5 “shift[ed] the advantage of time and inertia from the perpetrators of the evil to its victim.” Katzenbach, 383 U.S. at 328. It did so by placing “the burden on covered jurisdictions to show their voting changes are nondiscriminatory before those changes can be put into effect.” Shelby Cnty., 811 F. Supp. 2d at 431. Section 5 thus “pre-empted the most powerful tools of black disenfranchisement,” Nw. Austin, 129 S. Ct. at 2509, resulting in “undeniable” improvements in the protection of minority voting rights, id. at 2511. Section 4(b) contains a formula that, as originally enacted, applied section 5’s preclearance requirements to any state or political subdivision of a state that “maintained a voting test or device as of November 1, 1964, and had less than 50% voter registration or turnout in the 1964 presidential election.” Shelby Cnty., 811 F. Supp. 2d at 432 (citing Voting Rights Act of 1965, Pub. L. No. 89-110, § 4(b), 79 Stat. 437, 438 (“1965 Act”)). Congress chose these criteria carefully. It knew precisely which states it sought to cover and crafted the criteria to capture those jurisdictions. Id. (citing testimony before Congress in 2005–2006). Unsurprisingly, then, the jurisdictions originally covered in their entirety, Alabama, Georgia, Louisiana, Mississippi, South Carolina, and Virginia, “were those southern states with the worst historical records of racial discrimination in voting.” Id. Because section 4(b)’s formula could be both over- and underinclusive, Congress incorporated two procedures for adjusting coverage over time. First, as it existed in 1965, section 4(a) allowed jurisdictions to earn exemption from coverage by obtaining from a three-judge district court a 8 declaratory judgment that in the previous five years (i.e., before they became subject to the Act) they had used no test or device “for the purpose or with the effect of denying or abridging the right to vote on account of race or color.” 1965 Act § 4(a). This “bailout” provision, as subsequently amended, addresses potential overinclusiveness, allowing jurisdictions with clean records to terminate their section 5 preclearance obligations. Second, section 3(c) authorizes federal courts to require preclearance by any non-covered state or political subdivision found to have violated the Fourteenth or Fifteenth Amendments. 42 U.S.C. § 1973a(c). Specifically, courts presiding over voting discrimination suits may “retain jurisdiction for such period as [they] may deem appropriate” and order that during that time no voting change take effect unless either approved by the court or unopposed by the Attorney General. Id. This judicial “bail-in” provision addresses the formula’s potential underinclusiveness. As originally enacted in 1965, section 5 was to remain in effect for five years. In South Carolina v. Katzenbach, the Supreme Court sustained the constitutionality of section 5, holding that its provisions “are a valid means for carrying out the commands of the Fifteenth Amendment.” 383 U.S. at 337. Congress subsequently renewed the temporary provisions, including sections 4(b) and 5, in 1970 (for five years), then in 1975 (for seven years), and again in 1982 (for twenty-five years). In each version, “[t]he coverage formula [in section 4(b)] remained the same, based on the use of voting-eligibility tests [or devices] and the rate of registration and turnout among all voters, but the pertinent dates for assessing these criteria moved from 1964 to include 1968 and eventually 1972.” Nw. Austin, 129 S. Ct. at 2510. In 1975 Congress made one significant change to section 4(b)’s scope: it amended the definition of “test or device” to include the 9 practice of providing only English-language voting materials in jurisdictions with significant non-English-speaking populations. Act of Aug. 6, 1975, Pub. L. No. 94-73, § 203, 89 Stat. 400, 401–02 (codified at 42 U.S.C. § 1973b(f)(3)). Although not altering the basic coverage formula, this change expanded section 4(b)’s scope to encompass jurisdictions with records of voting discrimination against “language minorities.” See Briscoe v. Bell, 432 U.S. 404, 405 (1977). The Supreme Court sustained the constitutionality of each extension, respectively, in Georgia v. United States, 411 U.S. 526 (1973), City of Rome v. United States, 446 U.S. 156 (1980), and Lopez v. Monterey County, 525 U.S. 266 (1999). Significantly for the issue before us, the 1982 version of the Voting Rights Act made bailout substantially more permissive. Prior to 1982, bailout was extremely limited: no jurisdiction could bail out if it had used discriminatory voting tests or practices when it first became subject to section 5, even if it had since eliminated those practices. Shelby Cnty., 811 F. Supp. 2d at 434. By contrast, after 1982 the Act allowed bailout by any jurisdiction with a “clean” voting rights record over the previous ten years. Id. The 1982 reauthorization also permitted a greater number of jurisdictions to seek bailout. Previously, “only covered states (such as Alabama) or separately-covered political subdivisions (such as individual North Carolina counties) were eligible to seek bailout.” Id. After 1982, political subdivisions within a covered state could bail out even if the state as a whole was ineligible. Id. Setting the stage for this litigation, Congress extended the Voting Rights Act for another twenty-five years in 2006. See Fannie Lou Hamer, Rosa Parks, and Coretta Scott King Voting Rights Act Reauthorization and Amendments Act of 10 2006, Pub. L. No. 109-246, 120 Stat. 577 (“2006 Act”). In doing so, it acted on the basis of a legislative record “over 15,000 pages in length, and includ[ing] statistics, findings by courts and the Justice Department, and first-hand accounts of discrimination.” Shelby Cnty., 811 F. Supp. 2d at 435 (internal quotation marks omitted). Congress also amended section 5 to overrule the Supreme Court’s decisions in Georgia v. Ashcroft, 539 U.S. 461, 479–80 (2003) (which held that “any assessment of the retrogression of a minority group’s effective exercise of the electoral franchise depends on an examination of all the relevant circumstances” and that “a court should not focus solely on the comparative ability of a minority group to elect a candidate of its choice”), and Reno v. Bossier Parish School Board, 528 U.S. 320, 328 (2000) (“Bossier II”) (which held that “the ‘purpose’ prong of § 5 covers only retrogressive dilution”). See 2006 Act § 5 (codified at 42 U.S.C. § 1973c(b)–(d)). The 2006 Act’s constitutionality was immediately challenged by “a small utility district” subject to its provisions. See Nw. Austin, 129 S. Ct. at 2508. After finding the district ineligible for bailout, the three-judge district court concluded that the reauthorized Voting Rights Act was constitutional. Nw. Austin Mun. Util. Dist. No. One v. Mukasey, 573 F. Supp. 2d 221, 283 (D.D.C. 2008). On appeal, the Supreme Court identified two “serious . . . questions” about section 5’s continued constitutionality, namely, whether the “current burdens” it imposes are “justified by current needs,” and whether its “disparate geographic coverage is sufficiently related to the problem that it targets.” Nw. Austin, 129 S. Ct. at 2512–13. But invoking the constitutional avoidance doctrine, id. at 2508, 2513, the Court interpreted the statute to allow any covered jurisdiction, including the utility district bringing suit in that case, to seek 11 bailout, thus avoiding the need to resolve the “big question,” id. at 2508: Did Congress exceed its constitutional authority when it reauthorized section 5? Now that question is squarely presented. II. Shelby County filed suit in the U.S. District Court for the District of Columbia, seeking both a declaratory judgment that sections 4(b) and 5 of the Voting Rights Act are facially unconstitutional and a permanent injunction prohibiting the Attorney General from enforcing them. Shelby Cnty., 811 F. Supp. 2d at 427. Unlike the utility district in Northwest Austin, Shelby County never sought bailout, and for good reason. Because the county had held several special elections under a law for which it failed to seek preclearance and because the Attorney General had recently objected to annexations and a redistricting plan proposed by a city within Shelby County, the County was clearly ineligible for bailout. See id. at 446 n.6. As the district court—Judge John D. Bates—recognized, the “serious constitutional questions” raised in Northwest Austin could “no longer be avoided.” Id. at 427. Addressing these questions in a thorough opinion, the district court upheld the constitutionality of the challenged provisions and granted summary judgment for the Attorney General. After reviewing the extensive legislative record and the arguments made by Shelby County, the Attorney General, and a group of defendant-intervenors, the district court concluded that “Section 5 remains a ‘congruent and proportional remedy’ to the 21st century problem of voting discrimination in covered jurisdictions.” Id. at 428. Responding to the Supreme Court’s concerns in Northwest Austin, the district court found the record evidence of 12 contemporary discrimination in covered jurisdictions “plainly adequate to justify section 5’s strong remedial and preventative measures,” id. at 492 (internal quotation marks omitted), and to support Congress’s predictive judgment that failure to reauthorize section 5 “ ‘would leave minority citizens with the inadequate remedy of a Section 2 action,’ ” id. at 498 (quoting H.R. Rep. No. 109-478, at 57 (2006)). This evidence consisted of thousands of pages of testimony, reports, and data regarding racial disparities in voter registration, voter turnout, and electoral success; the nature and number of section 5 objections; judicial preclearance suits and section 5 enforcement actions; successful section 2 litigation; the use of “more information requests” and federal election observers; racially polarized voting; and section 5’s deterrent effect. Id. at 465–66. As to section 4(b), the district court acknowledged that the legislative record “primarily focused on the persistence of voting discrimination in covered jurisdictions—rather than on the comparative levels of voting discrimination in covered and non-covered jurisdictions.” Id. at 507. Nonetheless, the district court pointed to “several significant pieces of evidence suggesting that the 21st century problem of voting discrimination remains more prevalent in those jurisdictions that have historically been subject to the preclearance requirement”—including the disproportionate number of successful section 2 suits in covered jurisdictions and the “continued prevalence of voting discrimination in covered jurisdictions notwithstanding the considerable deterrent effect of Section 5.” Id. at 506–07. Thus, although observing that Congress’s reauthorization “ensured that Section 4(b) would continue to focus on those jurisdictions with the worst historical records of voting discrimination,” id. at 506, the district court found this continued focus justified by current 13 evidence that discrimination remained concentrated in those juridictions. See id. (explaining that Congress did not renew the coverage formula to punish past sins, but rather because it found “substantial evidence of contemporary voting discrimination by the very same jurisdictions that had histories of unconstitutional conduct”). Finally, the district court emphasized that Congress had based reauthorization not on “a perfunctory review of a few isolated examples of voting discrimination by covered jurisdictions,” but had “ ‘approached its task seriously and with great care.’ ” Id. at 496 (quoting Nw. Austin, 573 F. Supp. 2d at 265). Given this, the district court concluded that Congress’s predictive judgment about the continued need for section 5 in covered jurisdictions was due “substantial deference,” id. at 498 (internal quotation marks omitted), and therefore “decline[d] to overturn Congress’s carefully considered judgment,” id. at 508. Our review is de novo. See McGrath v. Clinton, 666 F.3d 1377, 1379 (D.C. Cir. 2012) (“We review the district court’s decision to grant summary judgment de novo.”). On appeal, Shelby County reiterates its argument that, given the federalism costs section 5 imposes, the provision can be justified only by contemporary evidence of the kind of “ ‘unremitting and ingenious defiance’ ” that existed when the Voting Rights Act was originally passed in 1965. Appellant’s Br. 8 (quoting Katzenbach, 383 U.S. at 309). Insisting that the legislative record lacks “evidence of a systematic campaign of voting discrimination and gamesmanship by the covered jurisdictions,” Shelby County contends that section 5’s remedy is unconstitutional because it is no longer congruent and proportional to the problem it seeks to cure. Id. at 8–9; see also City of Boerne v. Flores, 521 U.S. 507, 520 (1997) (“There must be a congruence and proportionality between the injury to be prevented or remedied and the means adopted 14 to that end.”). In addition, Shelby County argues, section 4(b) contains an “obsolete” coverage formula that fails to identify the problem jurisdictions, and because the jurisdictions it covers are not uniquely problematic, the formula is no longer rational “ ‘in both practice and theory.’ ” Appellant’s Br. 11– 12 (quoting Katzenbach, 383 U.S. at 330). III. Northwest Austin sets the course for our analysis, directing us to conduct two principal inquiries. First, emphasizing that section 5 “authorizes federal intrusion into sensitive areas of state and local policymaking that imposes substantial federalism costs,” the Court made clear that “[p]ast success alone . . . is not adequate justification to retain the preclearance requirements.” 129 S. Ct. at 2511. Conditions in the South, the Court pointed out, “have unquestionably improved”: racial disparities in voter registration and turnout have diminished or disappeared, and “minority candidates hold office at unprecedented levels.” Id. Of course, “[i]t may be that these improvements are insufficient and that conditions continue to warrant preclearance under the Act.” Id. at 2511–12. But “the Act imposes current burdens,” and we must determine whether those burdens are “justified by current needs.” Id. at 2512. Second, the Act, through section 4(b)’s coverage formula, “differentiates between the States, despite our historic tradition that all the States enjoy equal sovereignty.” Id. (internal quotation marks omitted). And while equal sovereignty “ ‘does not bar . . . remedies for local evils,’ ” id. (omission in original) (quoting Katzenbach, 383 U.S. at 328– 29), the Court warned that section 4(b)’s coverage formula may “fail[] to account for current political conditions”—that is, “[t]he evil that § 5 is meant to address may no longer be 15 concentrated in the jurisdictions singled out for preclearance.” Id. These concerns, the Court explained, “are underscored by the argument” that section 5 may require covered jurisdictions to adopt race-conscious measures that, if adopted by non- covered jurisdictions, could violate section 2 of the Act or the Fourteenth Amendment. Id. (citing Georgia v. Ashcroft, 539 U.S. at 491 (Kennedy, J., concurring) (“[C]onsiderations of race that would doom a redistricting plan under the Fourteenth Amendment or § 2 seem to be what save it under § 5.”)). To be sure, such “[d]istinctions can be justified in some cases.” Id. But given section 5’s serious federalism costs, Northwest Austin requires that we ask whether section 4(b)’s “disparate geographic coverage is sufficiently related to the problem that it targets.” Id. Before addressing Northwest Austin’s two questions, we must determine the appropriate standard of review. As the Supreme Court noted, the standard applied to legislation enacted pursuant to Congress’s Fifteenth Amendment power remains unsettled. See id. at 2512–13 (noting, but declining to resolve the parties’ dispute over the appropriate standard of review). Reflecting this uncertainty, Shelby County argues that the “congruence and proportionality” standard for Fourteenth Amendment legislation applies, see City of Boerne, 521 U.S. at 520, whereas the Attorney General insists that Congress may use “any rational means” to enforce the Fifteenth Amendment, see Katzenbach, 383 U.S. at 324. Although the Supreme Court declined to resolve this issue in Northwest Austin, the questions the Court raised—whether section 5’s burdens are justified by current needs and whether its disparate geographic reach is sufficiently related to that problem—seem to us the very questions one would ask to determine whether section 5 is “congruen[t] and proportional[] [to] the injury to be prevented,” City of Boerne, 16 521 U.S. at 520. We thus read Northwest Austin as sending a powerful signal that congruence and proportionality is the appropriate standard of review. In any event, if section 5 survives the arguably more rigorous “congruent and proportional” standard, it would also survive Katzenbach’s “rationality” review. Of course, this does not mean that the Supreme Court’s prior decisions upholding the Voting Rights Act are no longer relevant. Quite to the contrary, Katzenbach and City of Rome tell us a great deal about “[t]he evil that § 5 is meant to address,” Nw. Austin, 129 S. Ct. at 2512, as well as the types of evidence that are probative of “current needs,” id. Moreover, City of Boerne relied quite heavily on Katzenbach for the proposition that section 5, as originally enacted and thrice extended, was a model of congruent and proportional legislation. See City of Boerne, 521 U.S. at 525–26, 530 (relying on Katzenbach to explain how the Court evaluates remedial legislation under the Fourteenth and Fifteenth Amendments); see also id. at 532–33 (describing characteristics of the Voting Rights Act, as analyzed by Katzenbach and City of Rome, that made it congruent and proportional). We can likewise seek guidance from the Court’s Fourteenth Amendment decisions applying the congruent and proportional standard to other legislation. In those cases, the Court made clear that the record compiled by Congress must contain evidence of state “conduct transgressing the Fourteenth Amendment’s substantive provisions,” Coleman v. Court of Appeals of Md., 132 S. Ct. 1327, 1333 (2012), and that invasions of state interests based on “abstract generalities,” id. at 1337, or “supposition and conjecture,” id. at 1336, cannot be sustained. Once satisfied that Congress has 17 identified a pattern of constitutional violations, however, the Court has deferred to Congress’s judgment, even in the face of a rather sparse legislative record. In Nevada Department of Human Resources v. Hibbs, for example, the Court upheld the constitutionality of the family-care provision of the Family and Medical Leave Act, which allows eligible employees to take up to twelve weeks of unpaid leave, and “creates a private right of action to seek both equitable relief and money damages against any employer (including a public agency).” 538 U.S. 721, 724 (2003) (internal quotation marks omitted). Although evidence of discriminatory leave policies by state governments was hardly extensive, see Tennessee v. Lane, 541 U.S. 509, 528–29 & n.17 (2004) (describing the limited evidence relied upon in Hibbs, “little of which concerned unconstitutional state conduct”), the Court deferred to Congress’s “reasonabl[e] conclu[sions],” Hibbs, 538 U.S. at 734, and held that the evidence was “weighty enough to justify” prophylactic legislation, id. at 735. Similarly, in Lane the Court considered whether Congress had authority under the Fourteenth Amendment to pass Title II of the Americans with Disabilities Act, which prohibits public entities, including states, from discriminating on the basis of disability in their services, programs, and activities. 541 U.S. at 513. Looking into the record and noting the long history of state discrimination against disabled individuals, the Court found it “not difficult to perceive the harm that Title II is designed to address.” See id. at 524–25. It held, again with great deference to Congress’s take on the evidence, that the record, “including judicial findings of unconstitutional state action, and statistical, legislative, and anecdotal evidence of the widespread exclusion of persons with disabilities from the enjoyment of public services,” made “clear beyond peradventure” that Title II was appropriate prophylactic legislation, id. at 529—and this despite the fact that the record 18 included only two reported decisions finding unconstitutional state action of the precise type at issue, see id. at 544 (Rehnquist, C.J., dissenting). By contrast, the Court has found that Congress exceeded its Fourteenth Amendment authority where the legislative record revealed a “virtually complete absence” of evidence of unconstitutional state conduct. Id. at 521 (majority opinion) (citing Fla. Prepaid Postsecondary Educ. Expense Bd. v. Coll. Sav. Bank, 527 U.S. 627, 647–48 (1999)); see also City of Boerne, 521 U.S. at 530 (legislative record “lack[ed] examples of modern instances” of the targeted constitutional violations); Kimel v. Fla. Bd. of Regents, 528 U.S. 62, 89 (2000) (“Congress never identified any pattern of age discrimination by the States, much less any discrimination whatsoever that rose to the level of constitutional violation.”). We read this case law with two important qualifications. First, we deal here with racial discrimination in voting, one of the gravest evils that Congress can seek to redress. See Yick Wo v. Hopkins, 118 U.S. 356, 370 (1886) (“[The right to vote] is regarded as a fundamental political right, because preservative of all rights.”); Adarand Constructors, Inc. v. Pena, 515 U.S. 200, 216 (1995) (“racial classifications [are] constitutionally suspect and subject to the most rigid scrutiny” (citation omitted) (internal quotation marks omitted)). When Congress seeks to combat racial discrimination in voting— protecting both the right to be free from discrimination based on race and the right to be free from discrimination in voting, two rights subject to heightened scrutiny—it acts at the apex of its power. See Hibbs, 538 U.S. at 736 (noting that it is “easier for Congress to show a pattern of unconstitutional violations” when it enforces rights subject to heightened scrutiny); Lane, 541 U.S. at 561–63 (Scalia, J., dissenting) (“Giving [Congress’s enforcement powers] more expansive 19 scope with regard to measures directed against racial discrimination by the States accords to practices that are distinctively violative of the principal purpose of the [Reconstruction Amendments] a priority of attention that [the Supreme] Court envisioned from the beginning, and that has repeatedly been reflected in [the Court’s] opinions.”). Expressly prohibited by the Fifteenth Amendment, racial discrimination in voting is uniquely harmful in several ways: it cannot be remedied by money damages and, as Congress found, lawsuits to enjoin discriminatory voting laws are costly, take years to resolve, and leave those elected under the challenged law with the benefit of incumbency. Second, although the federalism costs imposed by the statutes at issue in Hibbs and Lane (abrogating sovereign immunity to allow suits against states for money damages) are no doubt substantial, the federalism costs imposed by section 5 are a great deal more significant. To be sure, in most cases the preclearance process is “routine” and “efficient[],” resulting in prompt approval by the Attorney General and rarely if ever delaying elections. See Reauthorizing the Voting Rights Act’s Temporary Provisions: Policy Perspectives and Views from the Field: Hearing Before the Subcomm. on the Constitution, Civil Rights and Propery Rights of the S. Comm. on the Judiciary, 109th Cong. 312–13 (2006) (testimony of Donald M. Wright, North Carolina State Board of Elections) (stating that most preclearance submissions “take only a few minutes to prepare” and that the Justice Department cooperates with jurisdictions to ensure that “preclearance issue[s] d[o] not delay an election”). But section 5 sweeps broadly, requiring preclearance of every voting change no matter how minor. Section 5 also places the burden on covered jurisdictions to demonstrate to the Attorney General or a three-judge district court here in Washington that the 20 proposed law is not discriminatory. Given these significant burdens, in order to determine whether section 5 remains congruent and proportional we are obligated to undertake a review of the record more searching than the Supreme Court’s review in Hibbs and Lane. Although our examination of the record will be probing, we remain bound by fundamental principles of judicial restraint. Time and time again the Supreme Court has emphasized that Congress’s laws are entitled to a “presumption of validity.” City of Boerne, 521 U.S. at 535. As the Court has explained, when Congress acts pursuant to its enforcement authority under the Reconstruction Amendments, its judgments about “what legislation is needed . . . are entitled to much deference.” Id. (internal quotation marks omitted). Even when applying intermediate scrutiny, the Court has accorded Congress deference “out of respect for its authority to exercise the legislative power,” and in recognition that Congress “is far better equipped than the judiciary to amass and evaluate the vast amounts of data bearing upon legislative questions.” Turner Broad. Sys., Inc. v. FCC, 520 U.S. 180, 195, 196 (1997) (internal quotation marks omitted) (rejecting a First Amendment challenge to the “must-carry” provisions of the Cable Television Consumer Protection and Competition Act). And critically for our purposes, although Northwest Austin raises serious questions about section 5’s constitutionality, nothing in that opinion alters our duty to resolve those questions using traditional principles of deferential review. Indeed, the Court reiterated not only that “judging the constitutionality of an Act of Congress is ‘the gravest and most delicate duty that [a court] is called on to perform,’ ” Nw. Austin, 129 S. Ct. at 2513 (quoting Blodgett v. Holden, 275 U.S. 142, 147–48 (1927) (Holmes, J., concurring)), but also that “[t]he Fifteenth 21 Amendment empowers ‘Congress,’ not the Court, to determine in the first instance what legislation is needed to enforce it,” id. A. Guided by these principles, we begin with Northwest Austin’s first question: Are the current burdens imposed by section 5 “justified by current needs”? 129 S. Ct. at 2512. The Supreme Court raised this question because, as it emphasized and as Shelby County argues, the conditions which led to the passage of the Voting Rights Act “have unquestionably improved[,] . . . no doubt due in significant part to the Voting Rights Act itself.” Id. at 2511. Congress also recognized this progress when it reauthorized the Act, finding that “many of the first generation barriers to minority voter registration and voter turnout that were in place prior to the [Voting Rights Act] have been eliminated.” H.R. Rep. No. 109-478, at 12. The dissent’s charts nicely display this progress. Racial disparities in voter registration and turnout have “narrowed considerably” in covered jurisdictions and are now largely comparable to disparities nationwide. Id. at 12–17; see also Dissenting Op. at 12–13 figs.I & II. Increased minority voting, in turn, has “resulted in significant increases in the number of African-Americans serving in elected offices.” H.R. Rep. No. 109-478, at 18; see also Dissenting Op. at 15 fig.III. For example, in the six states fully covered by the 1965 Act, the number of African Americans serving in elected office increased from 345 to 3700 in the decades since 1965. H.R. Rep. No. 109-478, at 18. But Congress found that this progress did not tell the whole story. It documented “continued registration and turnout disparities” in both Virginia and South Carolina. Id. at 25. Virginia, in particular, “remain[ed] an outlier,” S. Rep. 22 No. 109-295, at 11 (2006): although 71.6 percent of white, non-Hispanic voting age residents registered to vote in 2004, only 57.4 percent of black voting age residents registered, a 14.2-point difference. U.S. Census Bureau, Reported Voting and Registration of the Total Voting-Age Population, at tbl.4a, available at http://www.census.gov/hhes/www/ socdemo/voting/publications/p20/2004/tables.html (last visited May 9, 2012). Also, although the number of African Americans holding elected office had increased significantly, they continued to face barriers to election for statewide positions. Congress found that not one African American had yet been elected to statewide office in Mississippi, Louisiana, or South Carolina. In other covered states, “ ‘often it is only after blacks have been first appointed to a vacancy that they are able to win statewide office as incumbents.’ ” H.R. Rep. No. 109-478, at 33 (quoting Nat’l Comm’n on the Voting Rights Act, Protecting Minority Voters: The Voting Rights Act at Work 1982–2005, at 38 (2006) (“Nat’l Comm’n Report”)). Congress considered other types of evidence that, in its judgment, “show[ed] that attempts to discriminate persist and evolve, such that Section 5 is still needed to protect minority voters in the future.” Id. at 21. It heard accounts of specific instances of racial discrimination in voting. It heard analysis and opinions by experts on all sides of the issue. It considered, among other things, six distinct categories of evidence: (1) Attorney General objections issued to block proposed voting changes that would, in the Attorney General’s judgment, have the purpose or effect of discriminating against minorities; (2) “more information requests” issued when the Attorney General believes that the information submitted by a covered jurisdiction is insufficient to allow a preclearance determination; (3) successful lawsuits 23 brought under section 2 of the Act; (4) federal observers dispatched to monitor elections under section 8 of the Act; (5) successful section 5 enforcement actions filed against covered jurisdictions for failing to submit voting changes for preclearance, as well as requests for preclearance denied by the United States District Court for the District of Columbia; and (6) evidence that the mere existence of section 5 deters officials from even proposing discriminatory voting changes. Finally, Congress heard evidence that case-by-case section 2 litigation was inadequate to remedy the racial discrimination in voting that persisted in covered jurisdictions. Before delving into the legislative record ourselves, we consider two arguments raised by Shelby County that, if meritorious, would significantly affect how we evaluate that record. First, Shelby County argues that section 5 can be sustained only on the basis of current evidence of “a widespread pattern of electoral gamesmanship showing systematic resistance to the Fifteenth Amendment.” Appellant’s Br. 23. According to the County, the preclearance remedy may qualify as congruent and proportional only “when it addresses a coordinated campaign of discrimination intended to circumvent the remedial effects of direct enforcement of Fifteenth Amendment voting rights.” Id. at 7. We disagree. For one thing, how could we demand evidence of gamesmanship of the sort present at the time of Katzenbach given that section 5 preclearance makes such tactics virtually impossible? Equally important, Shelby County’s argument rests on a misreading of Katzenbach. Although the Court did describe the situation in 1965 as one of “unremitting and ingenious defiance of the Constitution,” Katzenbach, 383 U.S. at 309, nothing in Katzenbach suggests 24 that such gamesmanship was necessary to the Court’s judgment that section 5 was constitutional. Rather, the critical factor was that “Congress had found that case-by-case litigation was inadequate to combat widespread and persistent discrimination in voting.” Id. at 328; see also id. at 313–15 (explaining why laws facilitating case-by-case litigation had “proved ineffective”). In City of Rome, the Court, while recognizing that “undeniable” progress had been made, sustained section 5’s constitutionality without ever mentioning gamesmanship of any kind, 446 U.S. at 181–82; it relied instead on racial disparities in registration, the low number of minority elected officials, and the number and nature of Attorney General objections, id. at 180–81. Reinforcing this interpretation of Katzenbach and City of Rome, the Supreme Court explained in City of Boerne that “[t]he [Voting Rights Act’s] new, unprecedented remedies were deemed necessary given the ineffectiveness of the existing voting rights laws, and the slow, costly character of case-by-case litigation,” 521 U.S. at 526 (citation omitted). The Court reiterated the point in Board of Trustees of the University of Alabama v. Garrett, 531 U.S. 356, 373 (2001): “In [enacting the Voting Rights] Act . . . Congress also determined that litigation had proved ineffective . . . .” This emphasis on the inadequacy of case-by-case litigation makes sense: if section 2 litigation is adequate to deal with the magnitude and extent of constitutional violations in covered jurisdictions, then Congress might have no justification for requiring states to preclear their voting changes. Put another way, what is needed to make section 5 congruent and proportional is a pattern of racial discrimination in voting so serious and widespread that case- by-case litigation is inadequate. Given this, the question before us is not whether the legislative record reflects the kind 25 of “ingenious defiance” that existed prior to 1965, but whether Congress has documented sufficiently widespread and persistent racial discrimination in voting in covered jurisdictions to justify its conclusion that section 2 litigation remains inadequate. If it has, then section 5’s “substantial federalism costs” remain justified because preclearance is still needed to remedy continuing violations of the Fifteenth Amendment. Second, Shelby County urges us to disregard much of the evidence Congress considered because it involves “vote dilution, going to the weight of the vote once cast, not access to the ballot.” Appellant’s Br. 26. Specifically, the County faults Congress for relying on selective annexations, certain redistricting techniques, at-large elections, and other practices that do not prevent minorities from voting but instead “dilute minority voting strength,” 2006 Act § 2(b)(4)(A). According to the County, because the Supreme Court has “never held that vote dilution violates the Fifteenth Amendment,” Bossier II, 528 U.S. at 334 n.3, we may not rely on such evidence to sustain section 5 as a valid exercise of Congress’s Fifteenth Amendment enforcement power. It is true that neither the Supreme Court nor this court has ever held that intentional vote dilution violates the Fifteenth Amendment. But the Fourteenth Amendment prohibits vote dilution intended “invidiously to minimize or cancel out the voting potential of racial or ethnic minorities.” City of Mobile v. Bolden, 446 U.S. 55, 66 (1980); see also, e.g., Shaw v. Reno, 509 U.S. 630, 641 (1993). Although the Court’s previous decisions upholding section 5 focused on Congress’s power to enforce the Fifteenth Amendment, the same “congruent and proportional” standard, refined by the inquiries set forth in Northwest Austin, appears to apply 26 “irrespective of whether Section 5 is considered [Fifteenth Amendment] enforcement legislation, [Fourteenth Amendment] enforcement legislation, or a kind of hybrid legislation enacted pursuant to both amendments.” Shelby Cnty., 811 F. Supp. 2d at 462 (footnote omitted); see also City of Boerne, 521 U.S. at 518 (suggesting that Congress’s “power to enforce the provisions of the Fifteenth Amendment” is “parallel” to its power to enforce the Fourteenth Amendment). Indeed, when reauthorizing the Act in 2006, Congress expressly invoked its enforcement authority under both the Fourteenth and Fifteenth Amendments. See H.R. Rep. No. 109-478, at 90 (“[T]he Committee finds the authority for this legislation under amend. XIV, § 5 and amend. XV, § 2.”); id. at 53 & n.136 (stating that Congress is acting under its Fourteenth and Fifteenth Amendment powers in reauthorizing the Voting Rights Act). Accordingly, like Congress and the district court, we think it appropriate to consider evidence of unconstitutional vote dilution in evaluating section 5’s validity. See City of Rome, 446 U.S. at 181 (citing Congress’s finding that “[a]s registration and voting of minority citizens increase[], other measures may be resorted to which would dilute increasing minority voting strength” as evidence of the continued need for section 5 (internal quotation marks omitted)). Consideration of this evidence is especially important given that so-called “second generation” tactics like intentional vote dilution are in fact decades-old forms of gamesmanship. That is, “as African Americans made progress in abolishing some of the devices whites had used to prevent them from voting,” both in the late nineteenth century and again in the 1950s and 1960s, “[o]fficials responded by adopting new measures to minimize the impact of black 27 reenfranchisement.” Voting Rights Act: Evidence of Continued Need: Hearing Before the Subcomm. on the Constitution of the H. Comm. on the Judiciary, 109th Cong. 141–43 (2006) (“Evidence of Continued Need”). These measures—“well-known” tactics such as “ ‘pack[ing]’ ” minorities into a single district, spreading minority voters thinly among several districts, annexing predominately white suburbs, and so on—were prevalent “forms of vote dilution” then, and Congress determined that these persist today. Id. Specifically, Congress found that while “first generation barriers”—flagrant attempts to deny access to the polls that were pervasive at the time of Katzenbach—have diminished, “second generation barriers” such as vote dilution have been “constructed to prevent minority voters from fully participating in the electoral process.” 2006 Act § 2(b)(2) (congressional findings). Although such methods may be “more subtle than the visible methods used in 1965,” Congress concluded that their “effect and results are the same, namely a diminishing of the minority community’s ability to fully participate in the electoral process and to elect their preferred candidates of choice.” H.R. Rep. No. 109-478, at 6. Having resolved these threshold issues, we return to the basic question: Does the legislative record contain sufficient probative evidence from which Congress could reasonably conclude that racial discrimination in voting in covered jurisdictions is so serious and pervasive that section 2 litigation remains an inadequate remedy? Reviewing the record ourselves and focusing on the evidence most probative of ongoing constitutional violations, we believe it does. To begin with, the record contains numerous “examples of modern instances” of racial discrimination in voting, City of Boerne, 521 U.S. at 530. Just a few recent examples: 28 • Kilmichael, Mississippi’s abrupt 2001 decision to cancel an election when “an unprecedented number” of African Americans ran for office, H.R. Rep. No. 109- 478, at 36–37 (internal quotation marks omitted); • Webster County, Georgia’s 1998 proposal to reduce the black population in three of the education board’s five single-member districts after the school district elected a majority black school board for the first time, Voting Rights Act: Section 5 of the Act—History, Scope, and Purpose: Hearing Before Subcomm. on the Constitution of the House Judiciary Comm., 109th Cong. 830–31 (2006) (“History, Scope, and Purpose”); • Mississippi’s 1995 attempt to evade preclearance and revive a dual registration system “initially enacted in 1892 to disenfranchise Black voters” and previously struck down by a federal court, H.R. Rep. No. 109-478, at 39; • Washington Parish, Louisiana’s 1993 attempt to reduce the impact of a majority-African American district by “immediately creat[ing] a new at-large seat to ensure that no white incumbent would lose his seat,” id. at 38; • Waller County, Texas’s 2004 attempt to reduce early voting at polling places near a historically black university and its threats to prosecute students for “illegal voting,” after two black students announced their intent to run for office, Evidence of Continued Need 185–86. The legislative record also contains examples of overt hostility to black voting power by those who control the electoral process. In Mississippi, for instance, state legislators opposed an early 1990s redistricting plan that would have 29 increased the number of black majority districts, referring to the plan publicly as the “black plan” and privately as the “nigger plan,” Modern Enforcement of the Voting Rights Act: Hearing Before the S. Comm. on the Judiciary, 109th Cong. 22 (2006) (“Modern Enforcement”) (internal quotation marks omitted); see also S. Rep. No. 109-295, at 14. In Georgia, the state House Reapportionment Committee Chairman “told his colleagues on numerous occasions, ‘I don’t want to draw nigger districts,’ ” H.R. Rep. No. 109-478, at 67 (quoting Busbee v. Smith, 549 F. Supp. 495, 501 (D.D.C. 1982)). The district court pointed to numerous additional examples of intentional discrimination in the legislative record. See Shelby Cnty., 811 F. Supp. 2d at 472–76, 477–79, 480–81, 481–85, 485–87; see also Nw. Austin, 573 F. Supp. 2d at 258–62, 289– 301. In addition to these examples of flagrant racial discrimination, several categories of evidence in the record support Congress’s conclusion that intentional racial discrimination in voting remains so serious and widespread in covered jurisdictions that section 5 preclearance is still needed. We explore each in turn. First, Congress documented hundreds of instances in which the Attorney General, acting pursuant to section 5, objected to proposed voting changes that he found would have a discriminatory purpose or effect. Significantly, Congress found that the absolute number of objections has not declined since the 1982 reauthorization: the Attorney General interposed at least 626 objections during the twenty- two years from 1982 to 2004 (an average of 28.5 each year), compared to 490 interposed during the seventeen years from 1965 to 1982 (an average of 28.8 each year). Evidence of Continued Need 172; see also S. Rep. No. 109-295, at 13–14 30 (finding 754 objections between 1982 and the first half of 2006). Formal objections were not the only way the Attorney General blocked potentially discriminatory changes under section 5. Congress found that between 1990 and 2005, “more information requests” (MIRs) prompted covered jurisdictions to withdraw or modify over 800 proposed voting changes. Evidence of Continued Need 2553, 2565; H.R. Rep. No. 109- 478, at 40–41. Although MIRs take no position on the merits of a preclearance request, Congress had evidence indicating that the Attorney General sometimes uses them to “send signals to a submitting jurisdiction about the assessment of their proposed voting change” and to “promot[e] compliance by covered jurisdictions.” Evidence of Continued Need 2541. Congress found that because “[t]he actions taken by a jurisdiction [in response to an MIR] are often illustrative of [its] motives,” the high number of withdrawals and modifications made in response to MIRs constitutes additional evidence of “[e]fforts to discriminate over the past 25 years.” H.R. Rep. No. 109-478, at 40–41. Shelby County contends that section 5 objections and MIRs, however numerous, “do[] not signal intentional voting discrimination” because they represent only the Attorney General’s opinion and need not be based on discriminatory intent. Appellant’s Br. 30–31; see also id. at 32. Underlying this argument is a fundamental principle with which we agree: to sustain section 5, the record must contain “evidence of a pattern of constitutional violations,” Hibbs, 538 U.S. at 729, and voting changes violate the constitution only if motivated by discriminatory animus, Reno v. Bossier Parish Sch. Bd., 520 U.S. 471, 481 (1997) (“Bossier I”). Although not all objections rest on an affirmative finding of intentional 31 discrimination, the record contains examples of many that do. See Nw. Austin, 573 F. Supp. 2d at 289–301 (appendix providing examples of objections based on discriminatory intent). Between 1980 and 2004, the Attorney General issued at least 423 objections based in whole or in part on discriminatory intent. Voting Rights Act: Section 5— Preclearance Standards: Hearing Before the Subcomm. on the Constitution of the H. Comm. on the Judiciary, 109th Cong. 180–81 (2005) (“Preclearance Standards”). Moreover, in the 1990s, before the Supreme Court limited the Attorney General’s ability to object based on discriminatory but non- retrogressive intent, see Bossier II, 528 U.S. 320 (limiting the scope of section 5’s purpose prong in a decision overturned by the 2006 Act), “the purpose prong of Section 5 had become the dominant legal basis for objections,” Preclearance Standards 177, with seventy-four percent of objections based in whole or in part on discriminatory intent, id. at 136. Although it is true that objections represent “only one side’s opinion,” Appellant’s Br. 30, Congress is entitled to rely upon the Attorney General’s considered judgment “when it prescribes civil remedies . . . under [section] 2 of the Fifteenth Amendment.” Katzenbach, 383 U.S. at 330 (explaining that “Congress obviously may avail itself of information from any probative source,” including evidence “adduced by the Justice Department”). In fact, in City of Rome the Supreme Court considered objections to be probative evidence of unconstitutional voting discrimination. See 446 U.S. at 181. Shelby County also points out that the percentage of proposed voting changes blocked by Attorney General objections has steadily declined—from a height of 4.06 percent (1968–1972) to 0.44 percent (1978–1982) to 0.17 percent (1993–1997) and to 0.05 percent (1998–2002). An 32 Introduction to the Expiring Provisions of the Voting Rights Act and Legal Issues Relating to Reauthorization: Hearing Before the S. Comm. on the Judiciary, 109th Cong. 219 (2006) (“Introduction to the Expiring Provisions”). But the most dramatic decline in the objection rate—which, as the district court observed, “has always been low,” Shelby Cnty., 811 F. Supp. 2d at 470—occurred in the 1970s, before the Supreme Court upheld the Act for a third time in City of Rome. See Introduction to the Expiring Provisions 219. Also, the average number of objections per year has not declined, suggesting that the level of discrimination has remained constant as the number of proposed voting changes, many likely quite minor, has increased. See H.R. Rep. No. 109-478, at 22 (showing increase in the annual number of voting changes submitted for preclearance, from 300–400 per year in the early 1970s to 4000–5000 per year in the 1990s and 2000s). As the district court pointed out, there may be “many plausible explanations for the recent decline in objection rates.” See Shelby Cnty., 811 F. Supp. 2d at 471. Even in the six years from 2000 to 2006, after objection rates had dropped to their lowest, Attorney General objections affected some 660,000 minority voters. The Continuing Need for Section 5 Pre-Clearance: Hearing Before the S. Comm. on the Judiciary, 109th Cong. 58 (2006) (“Continuing Need”). Ultimately, Congress believed that the absolute number of objections represented the better indicator of the extent of discrimination in covered jurisdictions. This judgment— whether to accord greater weight to absolute numbers or to objection rates—is precisely the kind that a legislature is “far better equipped” than a court to evaluate, Turner Broad., 520 U.S. at 195 (internal quotation marks omitted). As for MIRs, we agree with Shelby County that they are less probative of discrimination than objections. An MIR does 33 not represent a judgment on the merits, and submitting jurisdictions might have many reasons for modifying or withdrawing a proposed change in response to one. But the record contains evidence from which Congress could “reasonabl[y] infer[],” id. (internal quotation marks omitted), that at least some withdrawals or modifications reflect the submitting jurisdiction’s acknowledgement that the proposed change was discriminatory. See Evidence of Continued Need 178 (stating that a jurisdiction’s decision to withdraw a proposed changes in response to an MIR “is frequently a tacit admission of one or more proposed discriminatory changes”); id. at 809–10 (explaining that after the Attorney General requested more information on a redistricting plan containing only two majority-black districts, the jurisdiction withdrew the proposal and ultimately adopted a redistricting plan with three majority-black districts); H.R. Rep. No. 109-478, at 41 (explaining that Monterey County’s proposal to reduce the number of polling places received preclearance only after the County withdrew five polling place consolidations in response to an MIR). Given this, Congress reasonably concluded that some of the 800-plus withdrawals and modifications in response to MIRs “reflect[]” “[e]fforts to discriminate over the past 25 years.” H.R. Rep. No. 109-478, at 40. The second category of evidence relied on by Congress, successful section 2 litigation, reinforces the pattern of discrimination revealed by objections and MIRs. The record shows that between 1982 and 2005, minority plaintiffs obtained favorable outcomes in some 653 section 2 suits filed in covered jurisdictions, providing relief from discriminatory voting practices in at least 825 counties. Evidence of Continued Need 208, 251. Shelby County faults the district court for relying on evidence of successful section 2 litigation 34 “even though ‘a violation of Section 2 does not require a showing of unconstitutional discriminatory intent.’ ” Appellant’s Br. 34 (quoting Shelby Cnty., 811 F. Supp. 2d at 481). The County’s premise is correct: although the Constitution prohibits only those voting laws motivated by discriminatory intent, section 2 prohibits all voting laws for which “ ‘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 [protected] class.’ ” Bartlett v. Strickland, 556 U.S. 1, 10–11 (2009) (quoting 42 U.S.C. § 1973(b)). In practice, however, this “results test,” as applied in section 2 cases, requires consideration of factors very similar to those used to establish discriminatory intent based on circumstantial evidence. Compare Gingles, 478 U.S. at 36–37 (listing factors considered under the results test), with Rogers v. Lodge, 458 U.S. 613, 623–27 (1982) (relying on virtually identical factors to affirm a finding of intentional discrimination). Also, as the district court pointed out, “courts will avoid deciding constitutional questions” if, as is the case in virtually all successful section 2 actions, the litigation can be resolved on narrower grounds. Shelby Cnty., 811 F. Supp. 2d at 482; see also, e.g., White v. Alabama, 74 F.3d 1058, 1071 n.42 (11th Cir. 1996) (“Because we dispose of the district court’s judgment on the ground that it violates the Voting Rights Act, we need not, and indeed, should not, discuss whether the judgment violates the Equal Protection Clause.”). This explains why the legislative record contains so few published section 2 cases with judicial findings of discriminatory intent, see Dissenting Op. at 26; To Examine the Impact and Effectiveness of the Voting Rights Act: Hearing Before the Subcomm. on the Constitution of the H. Comm. on the Judiciary, 109th Cong. 986–87 (2005) (“Impact and Effectiveness”) (report by Ellen Katz et al.)— 35 courts have no need to find discriminatory intent once they find discriminatory effect. But Congress is not so limited. Considering the evidence required to prevail in a section 2 case and accounting for the obligation of Article III courts to avoid reaching constitutional questions unless necessary, we think Congress quite reasonably concluded that successful section 2 suits provide powerful evidence of unconstitutional discrimination. In addition, as with Attorney General objections, we cannot ignore the sheer number of successful section 2 cases—653 over 23 years, averaging more than 28 each year. This high volume of successful section 2 actions is particularly dramatic given that Attorney General objections block discriminatory laws before they can be implemented and that section 5 deters jurisdictions from even attempting to enact such laws, thereby reducing the need for section 2 litigation in covered jurisdictions. See Continuing Need 26 (explaining that section 5 “makes the covered jurisdiction[s] much ‘cleaner’ than they would have been without Section 5 coverage”). Third, Congress relied on evidence of “the tens of thousands of Federal observers that have been dispatched to observe elections in covered jurisdictions.” 2006 Act § 2(b)(5). Specifically, 300 to 600 observers were dispatched annually between 1984 and 2000, H.R. Rep. No. 109-478, at 44, amounting to 622 separate dispatches (most or all involving multiple observers) to covered jurisdictions, Evidence of Continued Need 180–82; see also 42 U.S.C. § 1973f(a)(2) (authorizing dispatch of federal observers to covered jurisdictions based upon either “written meritorious complaints from residents, elected officials, or civic participation organizations,” or the Attorney General’s judgment that observers are necessary to enforce the Fourteenth or Fifteenth Amendment). Of these, sixty-six 36 percent were concentrated in five of the six states originally covered by section 5—Alabama, Georgia, Louisiana, Mississippi, and South Carolina. H.R. Rep. No. 109-478, at 44. In some instances, monitoring by federal observers “bec[ame] the foundation of Department of Justice enforcement efforts,” as in Conecuh County, Alabama, and Johnson County, Georgia, where reports by federal observers enabled the federal government to bring suit against county officials for discriminatory conduct in polling locations, ultimately resulting in consent decrees. Id.; see also Voting Rights Act: Sections 6 and 8—The Federal Examiner and Observer Program: Hearing Before the Subcomm. on the Constitution of the H. Comm. on the Judiciary, 109th Cong. 42–43 (2006) (“Sections 6 and 8”). As Congress saw it, this continued need for federal observers in covered jurisdictions is indicative of discrimination and “demonstrates that the discriminatory conduct experienced by minority voters is not solely limited to tactics to dilute the voting strength of minorities but continues to include tactics to disenfranchise, such as harassment and intimidation inside polling locations.” H.R. Rep. No. 109-478, at 44. Shelby County insists that the Attorney General’s decision to dispatch federal observers “indicates only that . . . there might be conduct with the effect of disenfranchising minority citizens, which might or might not be purposeful discrimination.” Appellant’s Br. 35–36. As the district court explained, however, “observers are not assigned to a particular polling location based on sheer speculation; they are only dispatched if ‘there is a reasonable belief that minority citizens are at risk of being disenfranchised.’ ” Shelby Cnty., 811 F. Supp. 2d at 486 (quoting H.R. Rep. No. 109-478, at 44). Indeed, the Justice Department conducts pre- election investigations in order to identify jurisdictions where 37 federal observers are likely to be necessary. See Sections 6 and 8, at 37–39 (explaining that the Justice Department conducts pre-election surveys and field investigations to identify jurisdictions where federal observers will be needed). The record shows that federal observers in fact witnessed discrimination at the polls, sometimes in the form of intentional harassment, intimidation, or disparate treatment of minority voters. See id. at 30–31 (describing discriminatory treatment and harassment of minorities by poll officials in Alabama); id. at 34 (describing discriminatory treatment of minority voters in Texas and Arizona); id. at 43 (describing the exclusion of African Americans from service as poll workers in Johnson County, Georgia). Thus, although the deployment of federal observers is hardly conclusive evidence of unconstitutional discrimination, we think Congress could reasonably rely upon it as modest, additional evidence of current needs. Fourth, Congress found evidence of continued discrimination in two types of preclearance-related lawsuits. Examining the first of these—actions brought to enforce section 5’s preclearance requirement—Congress noted that “many defiant covered jurisdictions and State and local officials continue to enact and enforce changes to voting procedures without the Federal Government’s knowledge.” H.R. Rep. No. 109-478, at 41. Between 1982 and 2004, at least 105 successful section 5 enforcement actions were brought against such jurisdictions. Evidence of Continued Need 250. Shelby County believes that successful section 5 enforcement actions are “not reliable evidence of intentional voting discrimination” because “[t]he most that a section 5 enforcement action can establish . . . is that a voting change— and quite possibly a nondiscriminatory voting change—was not properly submitted for preclearance.” Appellant’s Br. 34. 38 But the legislative record does contain evidence that at least some of the 105 successful section 5 enforcement suits were initiated in response to attempts by covered jurisdictions to implement purposefully discriminatory laws without federal oversight. See Shelby Cnty., 811 F. Supp. 2d at 480 (describing section 5 actions against Mississippi and Waller County, Texas, “in which the unprecleared voting changes appeared to have been motivated by discriminatory animus”); Evidence of Continued Need 176 (explaining that after a section 5 enforcement suit forced Mississippi to submit its dual registration law for preclearance, the Attorney General objected based on the law’s racially discriminatory purpose and effect). Therefore, Congress could reasonably have concluded that such cases, even if few in number, provide at least some evidence of continued willingness to evade the Fifteenth Amendment’s protections, for they reveal continued efforts by recalcitrant jurisdictions not only to enact discriminatory voting changes, but to do so in defiance of section 5’s preclearance requirement. In addition to section 5 enforcement suits, Congress found evidence of continued discrimination in “the number of requests for declaratory judgments [for preclearance] denied by the United States District Court for the District of Columbia.” 2006 Act § 2(b)(4)(B). The number of unsuccessful judicial preclearance actions appears to have remained roughly constant since 1966: twenty-five requests were denied or withdrawn between 1982 and 2004, compared to seventeen between 1966 and 1982. Evidence of Continued Need 177–78, 275. Shelby County does not contest the relevance of this evidence. Finally, and bolstering its conclusion that section 5 remains necessary, Congress “f[ound] that the existence of 39 Section 5 deterred covered jurisdictions from even attempting to enact discriminatory voting changes.” H.R. Rep. No. 109- 478, at 24. In Congress’s view, “Section 5’s strong deterrent effect” and “the number of voting changes that have never gone forward as a result of [that effect]” are “[a]s important as the number of objections that have been interposed to protect minority voters against discriminatory changes” that had actually been proposed. Id. As Congress explained, “ ‘[o]nce officials in covered jurisdictions become aware of the logic of preclearance, they tend to understand that submitting discriminatory changes is a waste of taxpayer time and money and interferes with their own timetables, because the chances are good that an objection will result.’ ” Id. (quoting Nat’l Comm’n Report 57). For this reason, the mere existence of section 5 “ ‘encourage[s] the legislature to ensure that any voting changes would not have a discriminatory effect on minority voters, and that it would not become embroiled in the preclearance process.’ ” Id. (quoting Laughlin McDonald, The Case for Extending and Amending the Voting Rights Act: Voting Rights Litigation, 1982–2006: A Report of the Voting Rights Project of the American Civil Liberties Union 15 (2006)). Congress considered testimony that section 5 has had just this effect on state and local redistricting processes. See H.R. Rep. No. 109-478, at 24 (describing section 5’s “critical” influence on the Georgia legislature’s redistricting process, which culminated in a plan that was precleared with no objection by the Attorney General (internal quotation marks omitted)); Evidence of Continued Need 362–63 (explaining how concerns about obtaining preclearance prevented Fredericksburg, Virginia, from eliminating an African American majority district). In other words, Congress had “some reason to believe that without [section 5’s] deterrent effect on potential misconduct,” the 40 evidence of continued discrimination in covered jurisdictions “might be considerably worse.” S. Rep. No. 109-295, at 11. Shelby County argues that Congress’s finding of deterrence reflects “ ‘outdated assumptions about racial attitudes in the covered jurisdictions’ ” that we should not “indulge[].” Appellant’s Br. 38 (quoting Nw. Austin, 129 S. Ct. at 2525 (Thomas, J., concurring in judgment in part and dissenting in part)). We agree that evaluating section 5’s deterrent effect raises sensitive and difficult issues. As the dissent rightly points out, the claimed effect is hard to measure empirically and even harder to consider judicially. Dissenting Op. at 24. We also agree with the dissent that section 5 could not stand based on claims of deterrence alone, nor could deterrence be used in some hypothetical case to justify renewal “to the crack of doom,” id. But the difficulty of quantifying the statute’s deterrent effect is no reason to summarily reject Congress’s finding that the evidence of racial discrimination in voting would look worse without section 5—a finding that flows from record evidence unchallenged by the dissent. As explained above, Congress’s deterrent effect finding rests on evidence of current and widespread voting discrimination, as well as on testimony indicating that section 5’s mere existence prompts state and local legislators to conform their conduct to the law. And Congress’s finding—that is, a finding about how the world would have looked absent section 5—rests on precisely the type of fact-based, predictive judgment that courts are ill- equipped to second guess. See Turner Broad., 520 U.S. at 195 (“In reviewing the constitutionality of a statute, courts must accord substantial deference to the predictive judgments of Congress.” (internal quotation marks omitted)). 41 This brings us, then, to Congress’s ultimate conclusion. After considering the entire record, including • 626 Attorney General objections that blocked discriminatory voting changes; • 653 successful section 2 cases; • over 800 proposed voting changes withdrawn or modified in response to MIRs; • tens of thousands of observers sent to covered jurisdictions; • 105 successful section 5 enforcement actions; • 25 unsuccessful judicial preclearance actions; • and section 5’s strong deterrent effect, i.e., “the number of voting changes that have never gone forward as a result of Section 5,” H.R. Rep. No. 109- 478, at 24; Congress found that serious and widespread intentional discrimination persisted in covered jurisdictions and that “case-by-case enforcement alone . . . would leave minority citizens with [an] inadequate remedy.” Id. at 57. In reaching this conclusion, Congress considered evidence that section 2 claims involve “intensely complex litigation that is both costly and time-consuming.” Modern Enforcement 96; see also Introduction to the Expiring Provisions 141 (describing a Federal Judicial Center study finding that voting rights cases require nearly four times more work than an average district court case and rank as the fifth most work-intensive of the sixty-three types of cases analyzed); City of Boerne, 521 U.S at 526 (noting the “slow costly character of case-by-case litigation” under section 2). It heard from witnesses who explained that “it is incredibly difficult for minority voters to pull together the resources needed” to pursue a section 2 lawsuit, particularly at the local level and in rural 42 communities. Modern Enforcement 96; see also History, Scope, and Purpose 84 (explaining that voters “in local communities and particularly in rural areas . . . do not have access to the means to bring litigation under Section 2”). Such testimony is particularly significant given that the vast majority of section 5 objections (92.5 percent from 2000 to 2005) pertained to local voting changes. See Michael J. Pitts, Let’s Not Call the Whole Thing Off Just Yet: A Response to Samuel Issacharoff’s Suggestion to Scuttle Section 5 of the Voting Rights Act, 84 Neb. L. Rev. 605, 612–13 (2005); see also id. at 616 (“[S]ection 2 cases are much less likely to be filed when it comes to redistricting in smaller jurisdictions[.]”). Congress also heard testimony that during the time it takes to litigate a section 2 action—often several years—proponents of a discriminatory law may enjoy its benefits, potentially winning elections and gaining the advantage of incumbency before the law is overturned. Impact and Effectiveness 43–44. Given all of this, and given the magnitude and persistence of discrimination in covered jurisdictions, Congress concluded that case-by-case litigation—slow, costly, and lacking section 5’s prophylactic effect—“would be ineffective to protect the rights of minority voters.” H.R. Rep. No. 109-478, at 57. According to Shelby County, “[e]valuation of the probative evidence shows there is no longer systematic resistance to the Fifteenth Amendment in the covered jurisdictions that cannot be solved through case-by-case litigation.” Appellant’s Br. 38. Congress, however, reached a different conclusion, and as explained above, the County has offered no basis for thinking that Congress’s judgment is either unreasonable or unsupported by probative evidence. The dissent accuses us of “overstat[ing] the inadequacies of § 2, such as cost and the consequences of delay.” Dissenting 43 Op. at 8. But the conclusion that section 2 is inadequate is Congress’s, not ours. The dissent believes that the costs of section 2 actions can “be assumed by the Department of Justice,” id., but it cites nothing in the record to support such speculation. The dissent also believes that “courts may as always use the standard remedy of a preliminary injunction to prevent irreparable harm caused by adjudicative delay.” Id. at 8–9. But Congress knows that plaintiffs can seek preliminary injunctions and reasonably determined that this possibility— that plaintiffs with few resources litigating a fact-intensive section 2 case will be able to satisfy the heavy burden required for preliminary injunctive relief—was insufficient to alleviate its concerns about the inadequacy of section 2 actions. The point at which section 5’s strong medicine becomes unnecessary and therefore no longer congruent and proportional turns on several critical considerations, including the pervasiveness of serious racial discrimination in voting in covered jurisdictions; the continued need for section 5’s deterrent and blocking effect; and the adequacy of section 2 litigation. These are quintessentially legislative judgments, and Congress, after assembling and analyzing an extensive record, made its decision: section 5’s work is not yet done. Insofar as Congress’s conclusions rest on predictive judgments, we must, contrary to the dissent’s approach, apply a standard of review even “more deferential than we accord to judgments of an administrative agency.” Turner Broad., 520 U.S. at 195. Given that we may not “displace [an agency’s] choice between two fairly conflicting views, even though the court would justifiably have made a different choice had the matter been before it de novo,” Universal Camera Corp. v. NLRB, 340 U.S. 474, 488 (1951), we certainly cannot do so here. Of course, given the heavy federalism costs that section 44 5 imposes, our job is to ensure that Congress’s judgment is reasonable and rests on substantial probative evidence. See Turner Broad., 520 U.S. at 195 (“In reviewing the constitutionality of a statute . . . [o]ur sole obligation is to assure that, in formulating its judgments, Congress has drawn reasonable inferences based on substantial evidence.” (internal quotation marks omitted)). After thoroughly scrutinizing the record and given that overt racial discrimination persists in covered jurisdictions notwithstanding decades of section 5 preclearance, we, like the district court, are satisfied that Congress’s judgment deserves judicial deference. B. Having concluded that section 5’s “current burdens” are indeed justified by “current needs,” we proceed to the second Northwest Austin inquiry: whether the record supports the requisite “showing that a statute’s disparate geographic coverage is sufficiently related to the problem that it targets.” 129 S. Ct. at 2512. Recall that this requirement stems from the Court’s concern that “[t]he Act . . . differentiates between the States, despite our historic tradition that all the States enjoy ‘equal sovereignty.’ ” Id. “The evil that § 5 is meant to address,” the Court observed, “may no longer be concentrated in the jurisdictions singled out [by section 4(b)] for preclearance.” Id. Before examining the record ourselves, we emphasize that the Act’s disparate geographic coverage—and its relation to the problem of voting discrimination—depends not only on section 4(b)’s formula, but on the statute as a whole, including its mechanisms for bail-in and bailout. Bailout functions as an integral feature of section 4’s coverage scheme: jurisdictions are subject to section 5 only if (1) they 45 are captured by section 4(b), and (2) they have not bailed out, meaning that they have failed to demonstrate a clean voting record as defined in section 4(a). See 42 U.S.C. §§ 1973b(a), 1973c(a). In addition, jurisdictions not captured by section 4(b) but which nonetheless have serious, recent records of voting discrimination, may be “bailed in”—i.e., subjected to section 5 preclearance—pursuant to section 3(c). See 42 U.S.C. § 1973a(c). Therefore, the question before us is whether the statute as a whole, not just the section 4(b) formula, ensures that jurisdictions subject to section 5 are those in which unconstitutional voting discrimination is concentrated. The most concrete evidence comparing covered and non- covered jurisdictions in the legislative record comes from a study of section 2 cases published on Westlaw or Lexis between 1982 and 2004. Impact and Effectiveness 964–1124 (report by Ellen Katz et al.). Known as the Katz study, it reached two key findings suggesting that racial discrimination in voting remains “concentrated in the jurisdictions singled out for preclearance,” Nw. Austin, 129 S. Ct. at 2512. First, the study found that of the 114 published decisions resulting in outcomes favorable to minority plaintiffs, 64 originated in covered jurisdictions, while only 50 originated in non-covered jurisdictions. Thus, although covered jurisdictions account for less than 25 percent of the country’s population, they accounted for 56 percent of successful section 2 litigation since 1982. Impact and Effectiveness 974; see also H.R. Rep. No. 109-478, at 53. When the Katz data is adjusted to reflect these population differences (based on the Census Bureau’s 2004 population estimates, the most recent data then available to Congress), the rate of successful section 2 cases in covered jurisdictions (.94 per million residents) is nearly four times the rate in non-covered jurisdictions (.25 per million 46 residents), as illustrated in the chart below. See Ellen Katz & The Voting Rights Initiative, VRI Database Master List (2006), http://sitemaker.umich.edu/votingrights/files/master list.xls; U.S. Dep’t of Justice, Section 5 Covered Jurisdictions, http://www.justice.gov/crt/about/vot/sec_5/ covered.php (last visited May 9, 2012); U.S. Census Bureau, Annual Estimates of the Population for the United States and States, and for Puerto Rico: April 1, 2000 to July 1, 2004, available at http://www.census.gov/popest/data/historical/ 2000s/vintage_2004/state.html (last visited May 9, 2012); U.S. Census Bureau, Annual Estimates of the Resident Population for Counties: April 1, 2000 to July 1, 2004, available at http://www.census.gov/popest/data/counties/ totals/2004/CO-EST2004-01.html (last visited May 9, 2012); U.S. Census Bureau, Population Estimates: Minor Civil Divisions: 2000 to 2004, available at http://www.census.gov/ popest/data/cities/totals/2004/SUB-EST2004-5.html (last visited May 9, 2012). Successful Published Section 2 Cases per Million Residents 1.00 Cases per Million Residents 0.90 0.80 0.70 0.60 0.50 0.40 0.30 0.20 0.10 0.00 Covered Jurisdictions Non-Covered Jurisdictions 47 Second, the study found higher success rates in covered jurisdictions than in non-covered jurisdictions. Specifically, 40.5 percent of published section 2 decisions in covered jurisdictions resulted in favorable outcomes for plaintiffs, compared to only 30 percent in non-covered jurisdictions. Impact and Effectiveness 974. The difference between covered and non-covered jurisdictions becomes even more pronounced when unpublished section 2 decisions—primarily court-approved settlements—are taken into account. As the Katz study noted, published section 2 lawsuits “represent only a portion of the section 2 claims filed or decided since 1982” since many claims were settled or otherwise resolved without a published opinion. Id. at 974. According to data compiled by the National Commission on the Voting Rights Act and Justice Department historian Peyton McCrary, there have been at least 686 unpublished successful section 2 cases since 1982, amounting to a total of some 800 published and unpublished cases with favorable outcomes for minority voters. See Decl. of Dr. Peyton McCrary 13 (“McCrary Decl.”). Of these, approximately 81 percent were filed in covered jurisdictions. Id. When this data is broken down state-by-state, separately identifying covered and non-covered portions of partially covered states, the concentration of successful section 2 cases in the covered jurisdictions is striking. Of the eight states with the highest number of successful published and unpublished section 2 cases per million residents—Alabama, Mississippi, Arkansas, Texas, South Carolina, Georgia, and the covered portions of South Dakota and North Carolina—all but one are covered. See Supp. Decl. of Dr. Peyton McCrary 3–7; U.S. Dep’t of Justice, Section 5 Covered Jurisdictions, http://www.justice.gov/crt/about/vot/sec_5/covered.php (last visited May 9, 2012); U.S. Census Bureau, Annual Estimates 48 of the Population for the United States and States, and for Puerto Rico: April 1, 2000 to July 1, 2004, available at http://www.census.gov/popest/data/historical/2000s/vintage _2004/state.html (last visited May 9, 2012); U.S. Census Bureau, Annual Estimates of the Resident Population for Counties: April 1, 2000 to July 1, 2004, available at http://www.census.gov/popest/data/counties/totals/2004/CO- EST2004-01.html (last visited May 9, 2012); U.S. Census Bureau, Population Estimates: Minor Civil Divisions: 2000 to 2004, available at http://www.census.gov/popest/data/cities/ totals/2004/SUB-EST2004-5.html (last visited May 9, 2012). The only exception is Arkansas, which, though not captured by section 4(b), was subjected to partial preclearance pursuant to a 1990 federal court order, i.e., “bailed in.” See Jeffers v. Clinton, 740 F. Supp. 585, 601–02 (E.D. Ark. 1990). Similarly, of the fourteen states with the highest number of successful published and unpublished section 2 cases per million residents—the eight listed above, plus Montana, Louisiana, New Mexico, Virginia, and the non-covered portions of South Dakota and North Carolina—eleven are either covered, including the seven states originally covered by the 1965 Act, or were bailed in for some period (Arkansas and New Mexico). See Travis Crum, Note, The Voting Rights Act’s Secret Weapon: Pocket Trigger Litigation and Dynamic Preclearance, 119 Yale L.J. 1992, 2010 & nn.100–01 (2010) (discussing bail-in of Arkansas and New Mexico). This data is displayed in the chart on the following page. 49 Cases Per Million Residents 10 15 20 25 30 35 40 45 0 5 South Dakota - C Alabama Mississippi North Carolina - C Arkansas Texas Successful Published and Unpublished Section 2 Cases South Carolina Georgia Montana South Dakota - NC Louisiana New Mexico North Carolina - NC Virginia Rhode Island per Million Residents North Dakota Delaware California - C Florida - NC Tennessee Maryland Illinois Hawaii New York - C Colorado Florida - C Indiana Nebraska Connecticut New York - NC (NC) Jurisdictions Non-Covered Jurisdictions Bailed-In Jurisdictions (C) Covered Massachusetts California - NC Utah Arizona Missouri Pennsylvania New Jersey Wisconsin Ohio Michigan - NC 50 Shelby County objects to the use of unpublished section 2 data, pointing out that although Congress considered the National Commission’s analysis of unpublished cases in covered jurisdictions, the legislative record does not contain McCrary’s analysis of unpublished cases in non-covered jurisdictions. We agree that there are reasons to approach this data with caution: McCrary prepared his analysis after the 2006 reauthorization, and because his data regarding unpublished cases in non-covered jurisdictions was collected separately from the data on unpublished cases in covered jurisdictions, we cannot be certain that the data collection methods were identical. That said, the Supreme Court has considered post-enactment evidence to find at least one law congruent and proportional, see Lane, 541 U.S. at 524–25 nn.6–9 & 13 (citing articles and cases published ten or more years after the Americans with Disabilities Act was enacted, as well as recent versions of statutes and regulations), and here a majority of the unpublished cases from non-covered jurisdictions (as well as all from covered jurisdictions) appears in the legislative record, see McCrary Decl. 10. Also, while the Katz data on published cases is necessarily underinclusive, see Impact and Effectiveness 974 (explaining that the published cases analyzed by the Katz study “represent only a portion” of all section 2 actions), Shelby County has identified no errors or inconsistencies in the data analyzed by McCrary. Indeed, McCrary points out that even if his methodology identified only half of the unpublished cases in non-covered jurisdictions, “there would still be 393 more settlements resolved favorably for minority voters in” covered jurisdictions. McCrary Decl. 11. For these reasons, although we would not rely solely on the combined published and unpublished data, we think it provides helpful additional evidence that corroborates the disparities in the level of 51 discrimination between covered and non-covered jurisdictions revealed by the published data. The section 2 data, moreover, does not tell the whole story. As explained above, Congress found that section 5, which operates only in covered jurisdictions, deters or blocks many discriminatory voting laws before they can ever take effect and become the target of section 2 litigation. “Section 5’s reach in preventing discrimination is broad. Its strength lies not only in the number of discriminatory voting changes it has thwarted, but can also be measured by the submissions that have been withdrawn from consideration, the submissions that have been altered by jurisdictions in order to comply with the [Voting Rights Act], or in the discriminatory voting changes that have never materialized.” H.R. Rep. No. 109- 478, at 36. Accordingly, if discrimination were evenly distributed throughout the nation, we would expect to see fewer successful section 2 cases in covered jurisdictions than in non-covered jurisdictions. See Continuing Need 26 (explaining that section 5 “makes the covered jurisdiction[s] much ‘cleaner’ than they would have been without Section 5 coverage”). Yet we see substantially more. Shelby County makes two main arguments in response to this evidence. First, citing Katzenbach’s finding that the coverage formula was “rational in both practice and theory,” 383 U.S. at 330, it contends that section 4(b) is irrational because it relies on “decades-old data.” Appellant’s Br. 59. “It cannot be constitutional,” Shelby County insists, “to rely on decades-old voting data to establish current voting discrimination.” Id. In addition, the County claims that in 1965 Congress was concerned with “first-generation” barriers—tests and devices that denied access to the ballot— and crafted the coverage formula to capture states that erected 52 such barriers and had low registration rates. But in 2006, although Congress was more concerned with “second- generation” barriers—vote dilution techniques that weaken “minority voting effectiveness”—it retained a coverage formula aimed at first-generation problems. Thus, Shelby County concludes, “[t]here is a serious mismatch between the conduct targeted by Congress and the factors that trigger coverage under Section 4(b).” Id. at 60. This argument rests on a misunderstanding of the coverage formula. As the district court explained, the election years that serve as coverage “triggers” under section 4(b) “were never selected because of something special that occurred in those years.” Shelby Cnty., 811 F. Supp. 2d at 505. Instead, Congress identified the jurisdictions it sought to cover—those for which it had “evidence of actual voting discrimination,” Katzenbach, 383 U.S. at 329—and then worked backward, reverse-engineering a formula to cover those jurisdictions. See id. (explaining that “Congress began work with reliable evidence of actual voting discrimination in a great majority of the States and political subdivisions affected by the new remedies of the Act” and that it “eventually evolved” a formula “to describe these areas”). The coverage formula relied on tests and devices “because of their long history as a tool for perpetrating the evil,” and voting rates because “widespread disenfranchisement must inevitably affect the number of actual voters.” Id. at 330. In other words, Congress chose the section 4(b) criteria not because tests, devices, and low participation rates were all it sought to target, but because they served as accurate proxies for pernicious racial discrimination in voting. The question, then, is not whether the formula relies on old data or techniques, but instead whether it, together with bail-in and bailout, continues to identify the jurisdictions with the worst 53 problems. If it does, then even though the formula rests on decades-old factors, the statute is rational in theory because its “disparate geographic coverage” remains “sufficiently related to the problem that it targets.” Nw. Austin, 129 S. Ct. at 2512. Of course, Shelby County’s real argument is that the statute fails this test, i.e., that it no longer actually identifies the jurisdictions “uniquely interfering with the right Congress is seeking to protect through preclearance.” Appellant’s Br. 62. The County points out that Congress never made a finding that racial discrimination in voting was “concentrated in the jurisdictions singled out for preclearance.” Nw. Austin, 129 S. Ct. at 2512. The County also argues that the Katz study is at best inconclusive, for some non-covered states, such as Illinois and the non-covered portions of New York, had more successful published section 2 lawsuits than did several covered states. In any event, it claims, “aggregated statistics showing slightly more Section 2 litigation with ‘favorable outcomes’ in covered jurisdictions as a group is not a rational basis for subjecting individually-targeted States to another 25 years of preclearance.” Appellant’s Br. 70. Shelby County’s first point—that Congress failed to make a finding—is easily answered. Congress did not have to. United States v. Lopez, 514 U.S. 549, 562 (1995) (Congress “normally is not required to make formal findings” in order to legislate). The proper question is whether the record contains sufficient evidence to demonstrate that the formula continues to target jurisdictions with the most serious problems. See Nw. Austin, 129 S. Ct. at 2512. This presents a close question. The record on this issue is less robust than the evidence of continued discrimination, see supra Part III.A, although this is in part due to the difficulty of comparing jurisdictions that have been subject to two very different enforcement regimes, 54 i.e., covered jurisdictions are subject to both sections 2 and 5 while non-covered jurisdictions are subject only to section 2. And although the Katz data in the aggregate does suggest that discrimination is concentrated in covered jurisdictions, just three covered states—Alabama, Louisiana, and Mississippi— account for much of the disparity. The covered states in the middle of the pack—North Carolina, South Carolina, Virginia, Texas, and Georgia—are about on par with the worst non-covered jurisdictions. And some covered states— Alaska and Arizona—had no successful published section 2 cases at all. As explained above, however, this data presents an incomplete picture of covered jurisdictions. When we consider the Katz data in conjunction with other record evidence, the picture looks quite different. For instance, although Georgia had only three successful published section 2 cases between 1982 and 2004, during that time the state had 66 successful unpublished section 2 cases, 83 section 5 objections, and 17 successful section 5 enforcement actions. Evidence of Continued Need 250–51, 272. In addition, between 1990 and 2005, jurisdictions in Georgia withdrew 90 proposed voting changes in response to MIRs. Id. at 2566. South Carolina is similar. Although the state had only 3 successful published section 2 cases, it had 30 successful unpublished section 2 cases, 74 section 5 objections, and 10 successful section 5 enforcement actions, as well as 26 voting changes withdrawn in response to MIRs and 51 changes that could not lawfully be implemented for failure to respond to MIRs. Id. at 250–51, 272, 2566. South Carolina, moreover, is one of the covered states that not only has continued racial disparities in voter registration and turnout, but that has never elected an African American to statewide office. See supra p. 22. Accordingly, even if only a relatively small portion of 55 objections, withdrawn voting changes, and successful section 5 enforcement actions correspond to unconstitutional conduct, and even if there are substantially more successful unpublished section 2 cases in non-covered jurisdictions than the McCrary data reveals, these middle-range covered jurisdictions appear to be engaged in much more unconstitutional discrimination compared to non-covered jurisdictions than the Katz data alone suggests. In fact, the discrepancy between covered and non-covered jurisdictions is likely even greater given that, as Congress found, the mere existence of section 5 deters unconstitutional behavior in the covered jurisdictions. That is, the middle-range covered states appear comparable to some non-covered jurisdictions only because section 5’s deterrent and blocking effect screens out discriminatory laws before section 2 litigation becomes necessary. Had section 5 not been in effect, one would expect significantly more discrimination in North Carolina, South Carolina, Virginia, Texas, and Georgia, all covered by section 5, than in the non-covered states with the worst records. See S. Rep. No. 109-295, at 11 (suggesting that “without the Voting Rights Act’s deterrent effect,” the evidence of discrimination in the covered jurisdictions “might be considerably worse”). To be sure, the coverage formula’s fit is not perfect. But the fit was hardly perfect in 1965. Accordingly, Katzenbach’s discussion of this issue offers a helpful guide for our current inquiry, particularly when we consider all probative record evidence of recent discrimination—and not just the small subset of section 2 cases relied upon by the dissent, see Dissenting Op. at 25–26. In 1965, the formula covered three states in “which federal courts ha[d] repeatedly found substantial voting discrimination”—Alabama, Louisiana, and Mississippi, Katzenbach, 383 U.S. at 329, the same three 56 states that, notwithstanding more than forty years of section 5 enforcement, still account for the highest rates of published successful section 2 litigation, as well as large numbers of unpublished successful section 2 cases, section 5 objections, federal observer coverages, and voting changes withdrawn or modified in response to MIRs. But the 1965 formula also “embrace[d] two other States—Georgia and South Carolina— plus large portions of a third State—North Carolina—for which there was more fragmentary evidence of recent voting discrimination mainly adduced by the Justice Department and the Civil Rights Commission.” Id. at 329–30. Today, the middle-range covered jurisdictions—North Carolina, South Carolina, Virginia, Texas, and Georgia—look similar: although the legislative record contains fewer judicial findings of racial discrimination in these states, it contains at least fragmentary evidence, in part based on Attorney General objections, that these states continue to engage in unconstitutional racial discrimination in voting. Finally, the 1965 formula swept in several other jurisdictions—including Alaska, Virginia, and counties in Arizona, Hawaii, and Idaho—for which Congress apparently had no evidence of actual voting discrimination. See id. at 318, 329–30. Today, the Act likewise encompasses jurisdictions for which there is some evidence of continued discrimination—Arizona and the covered counties of California, Florida, and New York, see Evidence of Continued Need 250–51, 272—as well as jurisdictions for which there appears little or no evidence of current problems—Alaska and a few towns in Michigan and New Hampshire. Critically, moreover, and as noted above, in determining whether section 5 is “sufficiently related to the problem that it targets,” we look not just at the section 4(b) formula, but at the statute as a whole, including its provisions for bail-in and 57 bailout. Bail-in allows jurisdictions not captured by section 4’s coverage formula, but which nonetheless discriminate in voting, to be subjected to section 5 preclearance. Thus, two non-covered states with high numbers of successful published and unpublished section 2 cases—Arkansas and New Mexico—were subjected to partial preclearance under the bail-in provision. See Jeffers, 740 F. Supp. at 601–02; Crum, 119 Yale L.J. at 2010 & n.101 (citing Sanchez v. Anaya, No. 82-0067M, slip op. at 8 (D.N.M. Dec. 17, 1984)). Federal courts have also bailed in jurisdictions in several states, including Los Angeles County, California; Escambia County, Florida; Thurston County, Nebraska; Bernalillo County, New Mexico; Buffalo County, South Dakota; Charles Mix County, South Dakota; and the city of Chattanooga, Tennessee. See Crum, 119 Yale L.J. at 2010 & nn.102–08. Bailout plays an even more important role in ensuring that section 5 covers only those jurisdictions with the worst records of racial discrimination in voting. As the Supreme Court explained in City of Boerne, the availability of bailout “reduce[s] the possibility of overbreadth” and helps “ensure Congress’ means are proportionate to [its] ends.” 521 U.S. at 533; see also Katzenbach, 383 U.S. at 329 (“Acknowledging the possibility of overbreadth, the Act provides for termination of special statutory coverage at the behest of States and political subdivisions in which the danger of substantial voting discrimination has not materialized during the preceding five years.”). As of May 9, 2012, having demonstrated that they no longer discriminate in voting, 136 jurisdictions and sub-jurisdictions had bailed out, including 30 counties, 79 towns and cities, 21 school boards, and 6 utility or sanitary districts. U.S. Dep’t of Justice, Section 4 of the Voting Rights Act, http://www.justice.gov/crt/about/vot/ misc/sec_4.php#bailout_list (last visited May 9, 2012) (“DOJ 58 Bailout List”). In fact, by ruling in Northwest Austin that any jurisdiction covered by section 5 could seek bailout—a development unmentioned by the dissent—the Supreme Court increased significantly the extent to which bailout helps “ensure Congress’ means are proportionate to [its] ends,” Boerne, 521 U.S. at 533. See Nw. Austin, 129 S. Ct. at 2516 (holding that “all political subdivisions—not only those described in § 14(c)(2)—are eligible to file a bailout suit”). Not surprisingly, then, the pace of bailout increased after Northwest Austin: of the successful bailout actions since 1965, 30 percent occurred in the three years after the Supreme Court issued its decision in 2009. See DOJ Bailout List, http://www.justice.gov/crt/about/vot/misc/sec_4.php#bailout_ list. Also, the Attorney General “has a number of active bailout investigations, encompassing more than 100 jurisdictions and subjurisdictions from a range of States.” Br. for Att’y Gen. as Appellee at 47–48, LaRoque v. Holder, No. 11-5349 (D.C. Cir. May 18, 2012). The importance of this significantly liberalized bailout mechanism cannot be overstated. Underlying the debate over the continued need for section 5 is a judgment about when covered jurisdictions—many with very bad historic records of racial discrimination in voting—have changed enough so that case-by-case section 2 litigation is adequate to protect the right to vote. Bailout embodies Congress’s judgment on this question: jurisdictions originally covered because of their histories of discrimination can escape section 5 preclearance by demonstrating a clean record on voting rights for ten years in a row. See 42 U.S.C. § 1973b(a)(1) (bailout criteria). As the House Report states, “covered status has been and continues to be within the control of the jurisdiction such that those jurisdictions that have a genuinely clean record and want to terminate coverage have the ability to do so.” H.R. Rep. No. 59 109-478, at 25. Bailout thus helps to ensure that section 5 is “sufficiently related to the problem that it targets,” Nw. Austin, 129 S. Ct. at 2512. Shelby County complains that bailout helps only “at the margins,” Appellant’s Br. 53; see also Dissenting Op. at 29, and the dissent emphasizes that only about 1 percent of covered jurisdictions and subjurisdictions have applied for bailout, Dissenting Op. at 29. But absent evidence that there are “clean” jurisdictions that would like to bail out but cannot meet the standards, the low bailout rate tells us nothing about the effectiveness of the bailout provision. See Shelby Cnty., 811 F. Supp. 2d at 500–01 (describing “several plausible explanations for th[e] failure to seek bailout,” including “the minimal administrative cost associated with preclearance, and the fact that covered jurisdictions see no need to avoid the preclearance requirement”). As the dissent concedes, since 1982 no bailout application has been denied, Dissenting Op. at 29, and Congress considered evidence that the bailout criteria “are easily proven for jurisdictions that do not discriminate in their voting practices.” Voting Rights Act: An Examination of the Scope and Criteria for Coverage Under the Special Provisions of the Act: Hearing Before the Subcomm. on the Constitution of the H. Comm. on the Judiciary, 109th Cong. 90 (2005). The dissent speculates that “opaque standards” may prevent bailouts, Dissenting Op. at 29, but neither it nor Shelby County specifically challenges Congress’s definition of what constitutes a clean jurisdiction or how the Attorney General is applying the bailout criteria. In fact, as noted above, Shelby County never even tried to bail out and has brought only a facial challenge. If something about the bailout criteria themselves or how the Attorney General is applying them is preventing jurisdictions with clean records from escaping section 5 preclearance, those 60 criteria can be challenged in a separate action brought by any adversely affected jurisdiction. See United States v. Salerno, 481 U.S. 739, 745 (1987) (explaining that in a facial challenge, “[t]he fact that [a law] might operate unconstitutionally under some conceivable set of circumstances is insufficient to render it wholly invalid”). This, then, brings us to the critical question: Is the statute’s “disparate geographic coverage . . . sufficiently related to the problem that it targets”? Nw. Austin, 129 S. Ct. at 2512. Of course, if the statute produced “a remarkably bad fit,” Dissenting Op. at 25, then we would agree that it is no longer congruent and proportional. But as explained above, although the section 4(b) formula relies on old data, the legislative record shows that it, together with the statute’s provisions for bail-in and bailout—hardly “tack[ed] on,” id. at 30 (internal quotation marks omitted), but rather an integral part of the coverage mechanism—continues to single out the jurisdictions in which discrimination is concentrated. Given this, and given the fundamental principle that we may not “strik[e] down an Act of Congress except upon a clear showing of unconstitutionality,” Salazar v. Buono, 130 S. Ct. 1803, 1820 (2010) (plurality opinion), we see no principled basis for setting aside the district court’s conclusion that section 5 is “sufficiently related to the problem that it targets,” Nw. Austin, 129 S. Ct. at 2512. C. We turn, finally, to the dissent’s argument that section 5 “requires a jurisdiction not only to engage in some level of race-conscious decisionmaking, but also on occasion to sacrifice principles aimed at depoliticizing redistricting.” Dissenting Op. at 4; see also Nw. Austin, 129 S. Ct. at 2512 (explaining that “federalism concerns are underscored by the 61 argument that . . . ‘considerations of race that would doom a redistricting plan under the Fourteenth Amendment or § 2 seem to be what save it under § 5’ ” and that “[a]dditional constitutional concerns are raised in saying that this tension between §§ 2 and 5 must persist in covered jurisdictions and not elsewhere” (quoting Georgia v. Ashcroft, 539 U.S. at 491 (Kennedy, J., concurring))). According to the dissent, this concern and the burden imposed by section 5 are aggravated by the amendments to section 5 Congress added in conjunction with the 2006 reauthorization. Dissenting Op. at 5–7; see also 2006 Act § 5. The dissent’s thoughtful arguments face a serious obstacle. Shelby County neither challenges the constitutionality of the 2006 amendments or even argues that they increase section 5’s burdens, nor does it argue that section 5 requires covered jurisdictions to undertake impermissible considerations of race. These issues, in other words, are entirely unbriefed, and as we have repeatedly made clear, “appellate courts do not sit as self-directed boards of legal inquiry and research, but essentially as arbiters of legal questions presented and argued by the parties before them.” Carducci v. Regan, 714 F.2d 171, 177 (D.C. Cir. 1983). Where, as here, “counsel has made no attempt to address the issue, we will not remedy the defect, especially where, as here, important questions of far-reaching significance are involved.” Id. (internal quotation marks omitted). Even were they not forfeited, the dissent’s concerns would not have satisfied the standards for mounting a facial constitutional challenge. Such a challenge, the Supreme Court has made clear, is “the most difficult . . . to mount successfully, since the challenger must establish that no set of circumstances exists under which the Act would be valid.” 62 Salerno, 481 U.S. at 745. Yet the amendments, as well as the Supreme Court’s concern that section 5 may sometimes require otherwise impermissible race-conscious decisionmaking, are implicated only in a subset of cases. Specifically, the amendment overturning Bossier II is implicated only in cases involving a discriminatory but non- retrogressive purpose, see 42 U.S.C. § 1973c(c); the amendments overturning Georgia v. Ashcroft, like the Supreme Court’s concern about race-conscious decisionmaking, are implicated primarily in redistricting cases where section 5 seems to require consideration of race as a “ ‘predominant factor.’ ” See Nw. Austin, 129 S. Ct. at 2512 (quoting Georgia v. Ashcroft, 539 U.S. at 491 (Kennedy, J., concurring)); 42 U.S.C. § 1973c(b), (d). In other words, even assuming the dissent is correct, it would not have established that “no set of circumstances exists under which the Act would be valid,” Salerno, 481 U.S. at 745. Indeed, addressing the dissent’s arguments would lead us into the very kind of “speculation” and “anticipat[ion]” of constitutional questions that require courts to “disfavor[]” facial challenges. Wash. State Grange v. Wash. State Republican Party, 552 U.S. 442, 450 (2008) (internal quotation marks omitted). IV. In Northwest Austin, the Supreme Court signaled that the extraordinary federalism costs imposed by section 5 raise substantial constitutional concerns. As a lower federal court urged to strike this duly enacted law of Congress, we must proceed with great caution, bound as we are by Supreme Court precedent and confined as we must be to resolve only the precise legal question before us: Does the severe remedy of preclearance remain “congruent and proportional”? The legislative record is by no means unambiguous. But Congress drew reasonable conclusions from the extensive evidence it 63 gathered and acted pursuant to the Fourteenth and Fifteenth Amendments, which entrust Congress with ensuring that the right to vote—surely among the most important guarantees of political liberty in the Constitution—is not abridged on account of race. In this context, we owe much deference to the considered judgment of the People’s elected representatives. We affirm. So ordered. WILLIAMS, Senior Circuit Judge, dissenting: Section 5 of the Voting Rights Act imposes rather extraordinary burdens on “covered” jurisdictions—nine states (and every jurisdiction therein), plus a host of jurisdictions scattered through several other states. See Voting Section, U.S. Dep’t of Justice, Section 5 Covered Jurisdictions, http://www.justice.gov/crt/ab out/vot/sec_5/covered.php (last visited May 9, 2012) (listing the covered jurisdictions). Unless and until released from coverage (a process discussed below), each of these jurisdictions must seek the Justice Department’s approval for every contemplated change in election procedures, however trivial. See 42 U.S.C. § 1973c. Alternatively, it can seek approval from a three-judge district court in the District of Columbia. See id. Below I’ll address the criteria by which the Department and courts assess these proposals; for now, suffice it to say that the act not only switches the burden of proof to the supplicant jurisdiction, but also applies substantive standards quite different from those governing the rest of the nation. Section 4(b) of the act states two criteria by which jurisdictions are chosen for this special treatment: whether a jurisdiction had (1) a “test or device” restricting the opportunity to register or vote and (2) a voter registration or turnout rate below 50%. See 42 U.S.C. § 1973b(b). But § 4(b) specifies that the elections for which these two criteria are measured must be ones that took place several decades ago. The freshest, most recent data relate to conditions in November 1972—34 years before Congress extended the act for another 25 years (and thus 59 years before the extension’s scheduled expiration). See id. The oldest data—and a jurisdiction included because of the oldest data is every bit as covered as one condemned under the newest—are another eight years older. See id. 2 Of course sometimes a skilled dart-thrower can hit the bull’s eye throwing a dart backwards over his shoulder. As I will try to show below, Congress hasn’t proven so adept. Whether the criteria are viewed in absolute terms (are they adequate in themselves to justify the extraordinary burdens of § 5?) or in relative ones (do they draw a rational line between covered and uncovered jurisdictions?), they seem to me defective. They are not, in my view, “congruent and proportional,” as required by controlling Supreme Court precedent. My colleagues find they are. I dissent. * * * Although it is only the irrational coverage formula of § 4(b) that I find unconstitutional, it is impossible to assess that formula without first looking at the burdens § 5 imposes on covered jurisdictions. Any answer to the question whether § 4(b) is “sufficiently related to the problem it targets,” Northwest Austin Municipal Utility Dist. No. One v. Holder, 129 S. Ct. 2504, 2512 (2009), that is, whether it is “congruent and proportional,” must be informed by the consequences triggered by § 4(b). (I agree with the majority that Northwest Austin “send[s] a powerful signal that congruence and proportionality is the appropriate standard of review.” Maj. Op. at 16.)1 The greater the burdens imposed by § 5, the more 1 Given such a standard, I cannot understand how we could apply Salerno’s “no set of circumstances” test, see Maj. Op. at 61- 62, quite apart from the test’s questionable continued vitality, see, e.g., Washington State Grange v. Washington State Republican Party, 552 U.S. 442, 449 (2008). Suppose Congress had actually designed the coverage formula by having the chair of the Senate Judiciary Committee throw darts at a map and had included every jurisdiction where a dart landed. Would we be expected to reject a facial challenge simply on a showing that the behavior of one covered jurisdiction was so blatantly unconstitutional as to cry out for application of § 5? 3 accurate the coverage scheme must be. If, for example, § 5 merely required covered jurisdictions to notify the Justice Department of an impending change in voting procedures, without giving the Department power to delay or thwart implementation, even a rather loose coverage formula would likely appear proportional. But § 5 requires much more than notice. For covered jurisdictions, it mandates anticipatory review of state legislative or administrative acts, requiring state and local officials to go hat in hand to Justice Department officialdom to seek approval of any and all proposed voting changes. See 42 U.S.C. § 1973c(a). Since its inception, even supporters of the Voting Rights Act have recognized that the preclearance regime was particularly “strong medicine” for a particularly extreme problem. Voting Rights Act: Hearings on H.R. 6400 Before Subcomm. No. 5 of the House Comm. on the Judiciary, 89th Cong. 110 (1965) (statement of Rep. Chelf). When it first upheld the VRA, the Supreme Court recognized it as a “complex scheme of stringent remedies” and § 5 in particular as an “uncommon exercise of congressional power.” South Carolina v. Katzenbach, 383 U.S. 301, 315, 334 (1966). And only a few years ago the Supreme Court reminded us that the federalism costs of § 5 are “substantial.” Northwest Austin, 129 S. Ct. at 2511. A critical aspect of those costs is the shifted burden of proof (a matter I’ll discuss below in the realm of its most significant application). So too is the section’s broad sweep: § 5 applies to any voting change proposed by a covered jurisdiction, without regard to kind or magnitude, and thus governs many laws that likely could never “deny or abridge” a “minority group’s opportunity to vote.” See 42 U.S.C. § 1973c(a); Allen v. State Bd. of Elections, 393 U.S. 544, 566 (1969) (“The legislative history on the whole supports the view that Congress intended to reach any state enactment, 4 which altered the election law of a covered State in even a minor way.”). This obvious point is underscored by the miniscule and declining share of covered jurisdictions’ applications that draw Justice Department objections—with only five objections for every ten thousand submissions between 1998 and 2002. See Richard L. Hasen, Congressional Power to Renew the Preclearance Provisions of the Voting Rights Act After Tennessee v. Lane, 192 OHIO ST. L.J. 177, 192 & fig.3 (2005) (noting that the Department’s objection rate has “been falling steadily” ever since the early years of the VRA and equaled 0.05% between 1998 and 2002). In the vast majority of cases, then, the overall effect of § 5 is merely to delay implementation of a perfectly proper law. Of course the most critical features of § 5 are the substantive standards it applies to the covered jurisdictions. Whether a proposed voting change can be precleared turns on whether it would have a retrogressive effect on minority voters. See Beer v. United States, 425 U.S. 130, 141 (1976). In practice this standard requires a jurisdiction not only to engage in some level of race-conscious decisionmaking, but also on occasion to sacrifice principles aimed at depoliticizing redistricting. Suppose a covered jurisdiction sought to implement what we may loosely call “good government” principles. It might, for example, delegate the task of redistricting to a computer programmed to apply criteria such as compactness, contiguity, conformity to existing political boundaries, and satisfaction of one person, one vote requirements. Despite these worthy goals, the resulting plan, if it happened to reduce the number of majority-minority districts, would fail preclearance, as the government acknowledged at oral argument. See Tr. of Oral Arg. at 37- 38. As Justice Kennedy cautioned in Georgia v. Ashcroft, 539 U.S. 461 (2003), “[C]onsiderations of race that would doom a redistricting plan under the Fourteenth Amendment . . . seem 5 to be what save it under § 5.” Id. at 491 (Kennedy, J., concurring); see also Miller v. Johnson, 515 U.S. 900, 927 (1995) (noting that Justice Department’s “implicit command that States engage in presumptively unconstitutional race- based districting brings the Act . . . into tension with the Fourteenth Amendment”). Unfortunately, when Congress passed the 2006 version of the VRA, it not only disregarded but flouted Justice Kennedy’s concern. New subsections (b) and (d) were added to § 5 to overturn Georgia v. Ashcroft, thereby restricting the flexibility of states to experiment with different methods of maintaining (and perhaps even expanding) minority influence. The Georgia Court had prescribed a holistic approach to § 5, instructing courts confronting a proposed voting change “not [to] focus solely on the comparative ability of a minority group to elect a candidate of its choice,”2 539 U.S. at 480 (majority opinion), but also to consider the “extent to which a new plan changes the minority group’s opportunity to participate in the political process” writ large, id. at 482. Georgia thus gave covered jurisdictions an opportunity to make trade-offs between concentrating minority voters in increasingly safe districts and spreading some of those voters out into additional districts; the latter choice, the Court pointed out, might increase the “substantive representation” they enjoy and lessen the risks of “isolating minority voters from the rest of the State” and of “narrowing [their] political influence to only a fraction of political districts.” Id. at 481; see also Samuel Issacharoff, Is Section 5 of the Voting Rights 2 The discourse revolving around § 5 invariably assumes that members of a minority have virtually identical interests and preferences. I follow that pattern here, reserving for the end of the opinion consideration of how such an assumption relates to the real world and to the 15th Amendment. 6 Act a Victim of Its Own Success?, 104 COLUM. L. REV. 1710, 1729 (2004) (expressing concern that § 5’s “narrow focus on securing the electability of minority candidates could compromise the range of political accords available to minority voters and thereby, under conditions of mature political engagement, actually thwart minority political gains”); David Epstein & Sharyn O’Halloran, Measuring the Electoral and Policy Impact of Majority-Minority Voting Districts, 43 AM. J. POL. SCI. 367, 390-92 (1999) (noting that overreliance on majority-minority districts means that “moderate senators will likely be replaced by extremists,” undermining the ability to create “biracial coalitions [which] are a key to passing racially progressive policies”). In so doing, the Court recognized that a minority group might in fact “achieve greater overall representation . . . by increasing the number of representatives sympathetic to the interests of minority voters,” rather than merely by electing the maximum possible number of representatives dependent on securing a majority of minority votes. 539 U.S. at 483. As amended, the act forecloses this choice. Preclearance now has an exclusive focus—whether the plan diminishes the ability of minorities (always assumed to be a monolith) to “elect their preferred candidates of choice,” irrespective of whether policymakers (including minority ones) decide that a group’s long-term interests might be better served by less concentration—and thus less of the political isolation that concentration spawns. See 42 U.S.C. § 1973c(b); id. § 1973c(d); see also Texas v. United States, -- F. Supp. 2d --, 2011 WL 6440006, at *4 (D.D.C. Dec. 22, 2011) (interpreting the amended law to overturn Georgia). The amended § 5 thus not only mandates race-conscious decisionmaking, but a particular brand of it. In doing so, the new § 5 aggravates both the federal-state tension with which Northwest Austin was concerned and the tension between § 5 and the 7 Reconstruction Amendments’ commitment to nondiscrimination. Another 2006 amendment makes the § 5 burden even heavier. Section 5 prohibits preclearance of laws that have the “purpose” of “denying or abridging the right to vote on account of race or color.” 42 U.S.C. § 1973c(a). The Court had interpreted “purpose” to be consistent with § 5’s effects prong, so that the term justified denying preclearance only to changes with a “retrogressive” purpose, rather than changes with either that or a discriminatory purpose. See Reno v. Bossier Parish School Bd., 528 U.S. 320, 341 (2000) (“Bossier II”). The 2006 amendments reversed that decision, specifying that “purpose” encompassed “any discriminatory purpose.” 42 U.S.C. § 1973c(c) (emphasis added). This broadening of the § 5 criteria may seem unexceptionable, but the Court had previously found that assigning covered jurisdictions the burden of proving the absence of discriminatory purpose was precisely the device that the Department had employed in its pursuit of maximizing majority-minority districts at any cost: “The key to the Government’s position, which is plain from its objection letters if not from its briefs to this court . . . , is and always has been that Georgia failed to proffer a nondiscriminatory purpose for its refusal in the first two submissions to take the steps necessary to create [an additional] majority-minority district.” Miller, 515 U.S. at 924. By inserting discriminatory purpose into § 5, and requiring covered jurisdictions affirmatively to prove its absence, Congress appears to have, at worst, restored “the Justice Department’s implicit command that States engage in presumptively unconstitutional race- based districting,” id. at 927, and at best, “exacerbate[d] the substantial federalism costs that the preclearance procedure already exacts,” Bossier II, 528 U.S. at 336. 8 The majority correctly notes that Shelby did not argue that either of these amendments is unconstitutional. See Maj. Op. at 61. Neither do I. Appellant does argue however that § 4(b) is unconstitutional, that is, that § 4(b) is not a congruent and proportional response to the problem currently posed by voting discrimination. To answer that question one must necessarily first assess the severity of the consequences of coverage under § 4(b) (i.e., subjection to § 5 as it exists today). See supra at p. 2. Whether Congress is free to impose § 5 on a select set of jurisdictions also depends in part, of course, on possible shortcomings in the remedy that § 2 provides for the country as a whole. That section creates a right to sue any jurisdiction to stop voting practices that “result[] in a denial or abridgement” of the right to vote “on account of race or color.” 42 U.S.C. § 1973(a). Doubtless the section is less drastic a remedy than § 5 (and thus by some criteria less effective). But it is easy to overstate the inadequacies of § 2, such as cost and the consequences of delay. Compare Maj. Op. at 41-42. Unlike in most litigation, plaintiffs’ costs for § 2 suits can in effect be assumed by the Department of Justice by its either exercising its authority to bring suit itself, see, e.g., United States v. Blaine County, 363 F.3d 897 (9th Cir. 2004), or by intervening in support of the plaintiff, as it often does. See, e.g., Brown v. Bd. of School Comm’rs, 706 F.2d 1103, 1107 (11th Cir. 1983). So far as Departmental resource constraints are concerned, narrowing § 5’s reach would, as a matter of simple arithmetic, enable it to increase § 2 enforcement with whatever resources it stopped spending on § 5. For those cases where the Justice Department still fails to intervene, § 2 provides for reimbursement of attorney and expert fees for prevailing parties. See 42 U.S.C. § 1973l(e). Finally, as to the risk that discriminatory practices may take hold before traditional litigation has run its course, courts may as always use the standard remedy of a 9 preliminary injunction to prevent irreparable harm caused by adjudicative delay. See Perry v. Perez, 132 S. Ct. 934, 942 (2012). Indeed, the ubiquitous availability of § 2 is of course a reminder that § 5 was created for the specific purpose of overcoming state and local resistance to federal anti- discrimination policy. When the Supreme Court first upheld the act in 1966, it found that § 5 was necessary because “case- by-case litigation,” now governed by § 2, was “inadequate to combat the widespread and persistent discrimination in voting.” Katzenbach, 383 U.S. at 328. While § 2 was tailored to redress actual instances of discrimination, § 5 was crafted to overcome a “century of systematic resistance to the Fifteenth Amendment” and ongoing “obstructionist tactics.” Id. But life in the covered jurisdictions has not congealed in the 48 years since the first triggering election (or the 40 years since the most recent). “[C]urrent burdens . . . must be justified by current needs,” Northwest Austin, 129 S. Ct. at 2512, and the burden imposed by § 5 has only grown heavier in those same years. In order for § 4(b) to be congruent and proportional then, the disparity in current evidence of discrimination between the covered and uncovered jurisdictions must be proportionate to the severe differential in treatment imposed by § 5. Put another way, a distinct gap must exist between the current levels of discrimination in the covered and uncovered jurisdictions in order to justify subjecting the former group to § 5’s harsh remedy, even if one might find § 5 appropriate for a subset of that group. * * * 10 I now turn to assessing the evidence used to justify the § 4(b) coverage formula. The parties have offered no sophisticated statistical analysis of voting discrimination in the covered and uncovered jurisdictions, and what follows does not purport to fill the sophistication gap. The data considered are drawn from the evidence the parties have cited, as well as the more general set compiled by Congress, especially data the Supreme Court has previously found important. For instance, when it upheld the preclearance regime in 1980, the Supreme Court noted both the “significant disparity” that still existed between African- American and white voter registration rates, and the fact that the number of black elected officials in covered jurisdictions “fell far short of being representative” of the number of African-Americans residing in covered jurisdictions. City of Rome v. United States, 446 U.S. 156, 180-81 (1980). Beyond voter registration and black elected officials, the parties point us to comparative, state-by-state data detailing the number of federal observers sent into states to oversee elections, plus the number of successful § 2 lawsuits. I take each of these in turn. Voter Registration and Turnout Section 4(b)’s coverage formula is keyed to two indicators of voter access: voter turnout and the use of tests and devices in voter registration. See 42 U.S.C. § 1973b(b). In 1966 the Supreme Court characterized the VRA as “specifically designed” to remedy the “misuse of tests and devices” that characterized the “widespread and persistent discrimination” at the time. Katzenbach, 383 U.S. at 331. Section 5 was thus meant, at the very least, to ensure that members of minority groups had equal access to the voting booth. 11 Figures I and II3 focus on this central problem. The two charts compare white and black registration and turnout rates in the 2004 election, using state-by-state estimates from the U.S. Census Bureau. See U.S. Census Bureau, Reported Voting and Registration of the Total Voting-Age Population, at tbl.4a, available at http://www.census.gov/hhes/www/socde mo/voting/publications/p20/2004/tables.html. Each chart takes the number of non-Hispanic whites who registered or turned out as a proportion of the total citizen voting-age population (“CVAP”) and compares that ratio to the same ratio for the black population, i.e., it displays the ratio of these two ratios for each state. Thus the greater the ratio (and the further to the left on the chart), the greater the racial disparity. The chart excludes states where the Census Bureau was unable to make reliable estimates of black registration and turnout rates (presumably because the black population was too small to get a sufficient sample).4 3 All the charts exclude Michigan and New Hampshire, both partially covered states, because the few small townships covered constitute only a minute portion of those states and, as far as I can tell, have never been the subject of a § 5 action. 4 The only covered jurisdictions excluded are Alaska, New Hampshire, and South Dakota. Of those, only Alaska is a fully covered state. The other states excluded for want of data are Hawaii, Idaho, Iowa, Kansas, Maine, Montana, Nebraska, New Mexico, North Dakota, Oregon, Rhode Island, Utah, Vermont, West Virginia, and Wyoming. 12 13 14 There appears to be no positive correlation between inclusion in § 4(b)’s coverage formula and low black registration or turnout. Quite the opposite. To the extent that any correlation exists, it appears to be negative— condemnation under § 4(b) is a marker of higher black registration and turnout. Most of the worst offenders—states where in 2004 whites turned out or were registered in significantly higher proportion than African-Americans—are not covered. These include, for example, the three worst— Massachusetts, Washington, and Colorado. And in Alabama and Mississippi, often thought of as two of the worst offenders, African-Americans turned out in greater proportion than whites. Black Elected Officials The other metric that the Rome Court considered was the number of black elected officials. Figure III uses U.S. Census Bureau data from 2000 and a state-by-state breakdown of such officials from that same year and displays the number of African-Americans who had been elected to office as a proportion of their share of the total CVAP in a given state. See David A Bostis, Joint Ctr. for Pol. & Econ. Studies, Black Elected Officials: A Statistical Summary 2000, available at http://www.jointcenter.org/research/black-elected-officials-a- statistical-summary-2000; U.S. Census Bureau, Voting-Age Population and Voting-Age Citizens, at tbls.1-1 & 1-3, available at http://www.census.gov/population/www/cen2000 /briefs/phc-t31/index.html. Thus, the higher the percentage (and accordingly the further to the right on the chart), the closer African-Americans’ share of elected positions is to equaling their share of the CVAP. States where the African- American share of CVAP was less than 3% are excluded. 15 16 Again the results are the inverse of § 4(b)’s presuppositions. Covered jurisdictions have far more black officeholders as a proportion of the black population than do uncovered ones. Of the ten states with the highest proportion of black elected officials relative to population, eight are covered states, with the top five all being fully covered states (Virginia, Louisiana, South Carolina, Mississippi, and Alabama). Nor can the poor scores achieved by some uncovered states be chalked up to small black populations. Illinois, Missouri, Delaware and Michigan, where African- Americans comprise at least 10% of the CVAP, all fall to the left (i.e., on the worse side) of every one of the states fully covered by § 4(b). While the relatively high number of black officeholders in covered states might be taken as a testament to § 5’s past success, no one could credibly argue that the numbers are proof of the coverage scheme’s continued rationality. In upholding § 5, the district court acknowledged that the number of black elected officials had increased but found the nature of the positions insufficient, pointing particularly to the nationwide disparity between the black proportion of the population (11.9%) and the number of black officials elected to statewide office (5%). Shelby County v. Holder, 811 F. Supp. 2d 424, 468-69 (D.D.C. 2011). It is unclear how this supports singling out the covered jurisdictions. Of the 35 black officials holding statewide elective office in the whole country in 2000 (including 2 from the U.S. Virgin Islands), nearly a third (11) came from fully covered states, Bostis, supra, at 24 tbl.7A, a proportion roughly equivalent to these jurisdictions’ share of the nation’s African-American citizen voting-age population (about 33%), see U.S. Census Bureau, Voting-Age Population and Voting-Age Citizens, supra, at tbl.1-3. Of course one might expect that the higher average African-American share of the population in the covered states would lead to a higher share of statewide elected 17 officials. But if on that account one thinks there has been a shortfall in the covered states, it might be caused in part by the Justice Department’s policy of maximizing majority-minority districts, with the concomitant risks of “isolating minority voters from the rest of the State” and “narrowing [their] political influence to only a fraction of political districts.” Georgia v. Ashcroft, 539 U.S. 461, 481 (2003). If African- American candidates primarily face solidly African-American constituencies, and thus develop political personas pitched overwhelmingly to the Democratic side of the aisle, it would hardly be surprising that they might face special obstacles seeking statewide office (assuming, of course, racially- polarized voting, as § 5 does). See Epstein, supra, at 390-92. Federal Observers Section 8 of the VRA authorizes the Department to send federal observers to covered jurisdictions in order to enter polling places and monitor elections if “necessary to enforce the guarantees of the 14th or 15th amendment.” 42 U.S.C. § 1973f(a)(2)(B). Additionally, § 3(a) permits a court to authorize the appointment of federal observers in any political subdivision, whether covered or uncovered, if the court finds it “appropriate to enforce the voting guarantees of the fourteenth or fifteenth amendment.” Id. § 1973a(a); see also id. § 1973f(a)(1). In an extensive report, the National Commission on the Voting Rights Act mapped the number of occasions these observers had been assigned to states in the 22-year period between the prior VRA authorization (1982) and the 2004 election. See Nat’l Comm’n on the Voting Rights Act, Protecting Minority Voters: The Voting Rights Act at Work 1982-2005, at 61 & Map 10B (Feb. 2006) (“Nat’l Comm’n Report”). Figure IV shows the state-by-state distribution of observer coverages per million minority residents, where the minority population is calculated by subtracting the non-Hispanic white population from the total 18 2004 population, as estimated by the U.S. Census Bureau. See U.S. Census Bureau, Annual Estimates of the Population for Race Alone and Hispanic or Latino Origin for the United States and States: July 1, 2004, available at http://www.censu s.gov/popest/data/historical/2000s/vintage_2004/state.html. Superficially, Figure IV supports § 4(b), indicating that observers are being sent to covered states more often than to uncovered ones. Six of the “worst” eight states are covered ones. But a number of factors undermine any serious inference. First, the National Commission report explains that it has captured “each occasion when federal observers are detailed to a jurisdiction covered by Section 5 or Section 203.” Nat’l Comm’n Report at 60 (emphasis added). The apparent implication is that the Commission didn’t purport to collect data for jurisdictions not covered by either of those sections; if so, the data are useless for comparative purposes. Indeed, testimony before Congress suggests that the Civil Rights Division simply doesn’t use “observers” for uncovered states, preferring instead to send its own staff lawyers to monitor elections “[i]n areas of the country where Federal observers cannot be sent” (presumably meaning, “cannot be sent without the necessity and deterrent of getting court approval”). Voting Rights Act: Sections 6 and 8—The Federal Examiner and Observer Program: Hearing Before the Subcomm. on the Constitution of the Comm. on the Judiciary, 109th Cong. 196 (2005) (statement of Bernard Schlozman). In fact, when calling this to Congress’s attention, a Department official noted that the “the great bulk of . . . recent enforcement cases since, say 1993, have involved jurisdictions (e.g., Massachusetts, California, New York, New Jersey, Florida, Washington, and Pennsylvania) where there is no statutory authority to send Federal observers.” Id. 19 20 Even if we were to assume the National Commission’s figures to be complete, and thus that every federal observer between 1982 and 2004 was sent to a jurisdiction already covered under some part of the VRA (either § 5 or § 203), this suggests another limitation on the data’s relevance: The same Department that administers § 5 preclearance also decides where to send observers, so it is unsurprising that the covered states, which are already in the Department’s sights, would also receive the most observers. Finally, § 3 forces the Justice Department to go to court for authorization to assign observers to uncovered areas, while § 8 imposes no such hurdle for the covered ones, undermining further the data’s already questionable value. Successful Section 2 Lawsuits The final metric for which comparative data exist is reported, successful § 2 lawsuits. Appellees point us to a comprehensive list of reported, post-1982 § 2 cases compiled by Professor Ellen Katz and the Voting Rights Initiative at the University of Michigan Law School. See Ellen Katz & The Voting Rights Initiative, VRI Database Master List (2006) (“Katz Master List”), available at http://sitemaker.umich.edu/ votingrights/files/masterlist.xls. Relying on these data, the district court noted that more than 56% of successful § 2 suits from 1982 to 2006 have been filed in covered jurisdictions, although those jurisdictions comprise only a quarter of the nation’s population. See Shelby County, 811 F. Supp. 2d at 506. But the persuasive power of this statistic dissolves when we disaggregate the data by state. Figure V looks at each state’s number of successful § 2 lawsuits between 1982 and 2005, per million residents, using the same 2004 U.S. Census Bureau population estimates used above. Because Professor Katz’s database helpfully informs us whether each lawsuit 21 was located in a covered or uncovered jurisdiction, it is possible to break out the covered portions of partially covered states from the uncovered portions:5 A “(C)” below the state’s abbreviation indicates that the data pertain only to the covered portion of that state, and an “(NC)” indicates the opposite. Because one successful case in a covered portion of South Dakota in 24 years produced a ratio of 43 cases for every hypothetical million residents, the covered portions of South Dakota are excluded in order to avoid distorting the chart’s scale. 5 In order to separately calculate the populations of the covered portions of partially covered states (namely, New York, California, North Carolina, and Florida), Chart V uses the county-specific population estimates from the U.S. Census Bureau. See U.S. Census Bureau, Annual Estimates of the Resident Population for Counties: April 1, 2000 to July 1 2004, http://www.census.gov/pope st/data/counties/totals/2004/CO-EST2004-01.html (linking to county-specific data for these states and others); Voting Section, U.S. Dep’t of Justice, Section 5 Covered Jurisdictions, http://www.justice.gov/crt/about/vot/sec_5/covered.php (last visited May 9, 2012). 22 23 Like the federal observer data discussed above, Figure V suggests that a more narrowly tailored coverage formula— capturing only Mississippi, Alabama, and Louisiana, and possibly the covered portions of South Dakota and North Carolina—might be defensible. But beyond these, the covered jurisdictions appear indistinguishable from their uncovered peers. The five worst uncovered jurisdictions, including at least two quite populous states (Illinois and Arkansas), have worse records than eight of the covered jurisdictions: the six covered states appearing to the right, plus two fully covered states—Arizona and Alaska—which do not appear on the chart at all because there has been not one successful § 2 suit in those states in the whole 24-year period. Of the ten jurisdictions with the greatest number of successful § 2 lawsuits, only four are covered (five if we add back in the covered portion of South Dakota). A formula with an error rate of 50% or more does not seem “congruent and proportional.” To bolster these numbers, the majority relies on an account of purportedly successful, but unreported § 2 cases, numbers that it rightly notes one should “approach . . . with caution.” Maj. Op. at 50. Indeed, beyond the serious concerns about these data already elucidated by the majority (e.g., completely different groups gathered the data regarding covered and uncovered jurisdictions), we also have almost no information for how Mr. McCrary and his staff identified particular cases as “successful” or not. All we know is that he required “some evidence” that the case was “resolved” under § 2 and “some reference” to settlement. Joint Appendix 95. And the inference of “success” from evidence of possible settlements seems exceptionally weak, for both the unreported cases in the covered jurisdictions compiled by the National Commission and those from the uncovered jurisdictions compiled by Mr. McCrary. It overlooks not only the range of outcomes embraced in the concept of settlement but also the 24 strategic factors, including legal fees and reputational risk, that go into a jurisdiction’s decision to settle. Additionally, defenders of the coverage scheme point to two circumstances that might also artificially reduce § 2 figures for the covered states, namely the “blocking” effect of actual § 5 vetoes, and the deterrent effect of jurisdictions’ having to seek preclearance. As to blocking, there seems little basis to infer that many of the 626 objections spread over 24 years were substitutes for successful § 2 suits. Any such inference is undermined by the Department’s ability to almost costlessly “Just Say No,” the allocation of the burden of proof to the jurisdiction, the legal fees that fighting the Department will entail, and the difference in the substantive standards governing § 2 and § 5 proceedings. As to the imputed deterrence, it is plainly unquantifiable. If we assume that it has played a role, how much should we inflate the covered states’ figures to account for it, and which covered states? Given much weight, the supposed deterrent effect would justify continued VRA renewals out to the crack of doom. Indeed, Northwest Austin’s insistence that “current burdens . . . must be justified by current needs,” 129 S. Ct. at 2512, would mean little if § 5’s supposed deterrent effect were enough to justify the current scheme. See Tr. of Oral Arg. at 28, Northwest Austin Municipal Utility Dist. No. One v. Holder, 129 S. Ct. 2504 (2009) (No. 08-322) (statement of Chief Justice Roberts) (“Well, that’s like the old—you know, it’s the elephant whistle. You know, I have this whistle to keep away the elephants. . . . Well, there are no elephants, so it must work.”). * * * To recap, of the four metrics for which comparative data exist, one (voter registration and turnout) suggests that the 25 coverage formula completely lacks any rational connection to current levels of voter discrimination, another (black elected officials), at best does nothing to combat that suspicion, and, at worst, confirms it, and two final metrics (federal observers and § 2 suits) indicate that the formula, though not completely perverse, is a remarkably bad fit with Congress’s concerns. Given the drastic remedy imposed on covered jurisdictions by § 5, as described above, I do not believe that such equivocal evidence can sustain the scheme. The Supreme Court’s initial review of the formula in 1966 provides a model for evaluating such an imperfect correlation. It assessed the evidence of discrimination before it and divided the covered jurisdictions into three categories: (1) a group for which “federal courts have repeatedly found substantial voting discrimination”; (2) another group “for which there was more fragmentary evidence of recent voting discrimination”; and (3) a third set consisting of the “few remaining States and political subdivisions covered by the formula,” for which there was little or no such evidence of discrimination, but whose use of voting tests and low voter turnout warranted inclusion, “at least in the absence of proof that they have been free of substantial voting discrimination in recent years.” Katzenbach, 383 U.S. at 329-30. In that original review, the Supreme Court placed three states (Alabama, Mississippi, and Louisiana) in category one, another three (Georgia, South Carolina, and the covered portions of North Carolina) in category two, and finally two fully covered states (Virginia and Alaska) plus a few counties in Hawaii, Idaho, and Arizona, in category three. The evidence adduced above yields a far worse fit than the data reviewed in Katzenbach. Indeed, one would be hard- pressed to put any of the covered jurisdictions into Katzenbach’s first category. Based on any of the comparative data available to us, and particularly those metrics relied on in 26 Rome, it can hardly be argued that there is evidence of a “substantial” amount of voting discrimination in any of the covered states, and certainly not at levels anywhere comparable to those the Court faced in Katzenbach. In terms of successful § 2 law suits, only three covered states— Mississippi, Louisiana, and Alabama—plus uncovered Montana—have more than two successful suits per million residents over the past quarter-century (excluding of course the covered portion of South Dakota, which scores high only because with such a small population the one suit there produces a high ratio per hypothetical million); in fact, these three states are the only ones with more than 10 successful suits in the 24 years between 1982 and 2006.6 See Katz Master List. And of course, even this number may be artificially large since a successful § 2 suit does not necessarily entail a finding of unconstitutional behavior (i.e., intentionally discriminatory acts); indeed, the Katz Study itself reports only 12 findings of intentional discrimination in the covered jurisdictions over the same two-and-a-half decades, and on my reading of the cases Professor Katz lists, there are even fewer. See, e.g., Brown v. Bd. of School Comm’rs, 706 F.2d 1103, 1107 (11th Cir. 1983) (listed in both the Senate and Katz reports as a case finding discriminatory intent, but the case finds such intent only as to an electoral system enacted in 1876). Even assuming that these small numbers would qualify as “fragmentary evidence” adequate to place those three in Katzenbach’s second category, that leaves six fully covered states (plus several jurisdictions in partially covered states) in category three, many more than in 1966, when only two fully 6 I exclude North Carolina here because four of its ten successful suits were located in uncovered portions of the state. See Katz Master List. 27 covered states (Virginia and Alaska) were not included in either category one or two. See Katzenbach, 383 U.S. at 318, 329-30. A coverage scheme that allows two or three of the worst offenders to drag down other covered jurisdictions, whose continued inclusion is merely a combination of historical artifact and Congress’s disinclination to update the formula, can hardly be thought “congruent and proportional.” See Nathaniel Persily, The Promise and Pitfalls of the New Voting Rights Act, 117 YALE L.J. 174, 208-09 (2007) (concluding that any “debate over the coverage formula” would “likely have led to the complete unraveling” of the VRA’s 2006 reauthorization campaign); id. at 208 (“The most one can say in defense of the formula is that it is the best of the politically feasible alternatives . . . .”). Congress’s inability to agree on a currently coherent formula is not a good reason for upholding its extension of an anachronism. Moreover, the Court in 1966 relied on rather a natural inference from the data available. The tight relationship between the two trigger criteria (i.e., voter turnout and the use of voting “tests and devices”) and evidence of discrimination in the states in categories one and two, made it logical to suppose that Congress reasonably inferred a comparable fit for the remaining covered jurisdictions for which direct evidence of discrimination was missing (i.e., those in category three). But today the trigger criteria have lost any inherent link to the key concern. The newest triggering data hark back to 1972, 34 years before the current formula was enacted, and nearly 60 years before the current act expires. Indeed, if the formula were to be updated to use more recent election data, it would cover only Hawaii. See 152 CONG. REC. H5131, H5181 (daily ed. July 13, 2006). More critically, the Court’s acceptance of the § 4(b) formula in 1966 was explicitly based on certain reasonable understandings of § 5’s focus. Explaining why it saw no 28 serious problem in the challengers’ claim of underinclusiveness—§ 4(b)’s exclusion of localities not employing “tests or devices” but showing evidence of voting discrimination by other means—the Court observed that Congress had learned that persistent discrimination “has typically entailed the misuse of tests and devices, and this was the evil for which the new remedies were specifically designed.” Katzenbach, 383 U.S. at 331 (emphasis added). Despite § 5’s language imposing preclearance on all manner of voting rules not within the act’s definition of “tests or devices,” the Court understandably saw the act as focused on, or in its words “specifically designed” for, rooting out “the misuse of tests and devices.” But § 5 litigation no longer centers at all on “tests and devices.” Instead, the majority of § 5 objections today concern redistricting. See Peyton McCrary et al., The Law of Preclearance: Enforcing Section 5, in THE FUTURE OF THE VOTING RIGHTS ACT 20, 25 tbl.2.1 (David Epstein et al. eds., 2006) (redistricting objections comprised only 17% of Justice Department objections in the 1970s; in the ‘90s, they constituted 52% of all objections). Accordingly, quite apart from the trigger criteria’s hopeless fossilization, the intrinsic link between them and their consequences has ceased to exist. Nor is the coverage formula materially helped by the VRA’s bailout provision. Although Katzenbach did note that § 4(a)’s bailout provision might alleviate concerns about overinclusiveness, see 383 U.S. at 331, its ability to act as a reliable escape hatch is questionable. In its original form, § 4(a) essentially permitted bailout for any jurisdiction that had not used a voting “test or device” in the previous five years. See Voting Rights Act of 1965, Pub. L. 89-110, § 4(a), 79 Stat. 437, 438. This in effect excluded any covered jurisdiction whose record was not clean as of the date of initial enactment, and until 1982 the later reenactments’ language continued that effect (i.e., allowed access to bailout only for 29 those jurisdictions with clean records as of the VRA’s initial adoption). While the majority correctly notes that the 1982 amendments relaxed that constraint, see Maj. Op. at 9, those same amendments tightened the remaining substantive standards. A covered jurisdiction can now obtain bailout if, and only if, it can demonstrate that, during the preceding ten years, it has (simplifying slightly): (1) effectively engaged in no voting discrimination (proven by the absence of any judicial finding of discrimination or even a Justice Department “objection” (unless judicially overturned)); (2) faithfully complied with § 5 preclearance; (3) “eliminated voting procedures and methods of election which inhibit or dilute equal access to the electoral process”; and (4) engaged in “constructive efforts to eliminate intimidation and harassment of persons exercising rights protected” under the act and “in other constructive efforts, such as the expanded opportunity for convenient registration.” 42 U.S.C. § 1973b(a)(1). Perhaps because of these opaque standards, actual bailouts have been rare; only 136 of the more than 12,000 covered political subdivisions (i.e., about 1%) have applied for bailout (all successfully). Appellant’s Reply Br. 37; Voting Section, U.S. Dep’t of Justice, Terminating Coverage Under the Act’s Special Provisions, http://www.justice.gov/crt/about/vot/misc/sec_4.php#bailout (last visited May 9, 2012) (listing successful bailouts). Moreover, a successful action under § 4(a) does not actually end federal oversight of bailed-out jurisdictions; for a decade after bailout, the court “retain[s] jurisdiction” just in case the Justice Department or “any aggrieved person” wishes to file a motion “alleging that conduct has occurred which . . . would have precluded” bailout in the first place. 42 U.S.C. § 1973b(a)(5). All of this suggests that bailout may be only the most modest palliative to § 5’s burdens. One scholar hypothesizes that bailout may “exist[] more as a fictitious way out of 30 coverage than [as] an authentic way of shoring up the constitutionality of the coverage formula.” Persily, supra, at 213. In fairness, the same scholar also entertains various other explanations, including the possibility that the eligible jurisdictions are just the ones for whom § 5 poses only a very light burden, see id. at 213-14, and ultimately concludes that no one knows which theory “best explains the relative absence of bailouts,” id. at 214. Regardless of the reason for the trivial number of bailouts, irrational rules—here made so by their encompassing six states and numerous additional jurisdictions not seriously different from the uncovered states—cannot be saved “by tacking on a waiver procedure” such as bailout. ALLTEL Corp. v. FCC, 838 F.2d 551, 561 (D.C. Cir. 1988); cf. U.S. Telecomm. Ass’n v. FCC, 359 F.3d 554, 571 (D.C. Cir. 2004). Finally the government argues that because the VRA is meant to protect the fundamental right of racial minorities (i.e., a suspect classification), a heightened level of deference to Congress is in order. Appellees’ Br. 22-23. Purportedly supporting this proposition is Chief Justice Rehnquist’s statement in Nevada Dep’t of Human Resources v. Hibbs, 538 U.S. 721 (2003), that when a statute is designed to protect a fundamental right or to prevent discrimination based on a suspect classification, “it [is] easier for Congress to show a pattern of state constitutional violations.” Id. at 736. But the passage simply makes the point that where a classification is presumptively invalid (e.g., race), an inference of unlawful discrimination follows almost automatically from rules or acts that differentiate on the presumptively forbidden basis, whereas for classifications judged under the “rational basis” test, such as disability or age, “Congress must identify, not just the existence of age- or disability-based state decisions, but a widespread pattern of irrational reliance on such criteria.” Id. at 735 (emphasis added). This special element of race or other presumptively unconstitutional classifications 31 has no bearing on review of whether Congress’s remedy “fits” the proven pattern of discrimination. To hold otherwise would ignore completely the “vital principles necessary to maintain separation of powers and the federal balance” that the Court held paramount in Boerne (which of course also involved a fundamental right, namely the right to practice one’s religion). City of Boerne v. Flores, 521 U.S. 507, 536 (1997). * * * A current political dispute—state adoptions of voter identification requirements—highlights the oddity of § 4(b). In 2005, the state of Indiana enacted a law requiring its citizens to present a government-issued photo identification before voting. Against a variety of legal challenges, the Supreme Court upheld the law. See Crawford v. Marion County Election Bd., 553 U.S. 181 (2008). In 2011, Texas and South Carolina both passed similar laws. See Gina Smith, Haley Signs Voter ID Bill into Law, THE STATE, May 18, 2011; Sommer Ingram, Gov. Rick Perry Signs Voter ID Bill into Law, ASSOC. PRESS, May 27, 2011, available at http://www.yumasun.com/articles/perry-51036-monitortx- rick-austin.html. But because of those states’ inclusion under § 4(b), they had to look to Justice Department attorneys in Washington to seek further approval. In the end, the Department blocked both laws. See Jerry Markon, S.C.’s Voter ID Law Rejected, WASH. POST, Dec. 24, 2011, at A4; Daniel Gilbert, Election 2012: Texas Law Requiring Voter IDs Is Blocked, WALL ST. J., Mar. 13, 2012, at A4. Why should voter ID laws from South Carolina and Texas be judged by different criteria (at a minimum, a different burden of persuasion, which is often critical in cases involving competing predictions of effect) from those governing Indiana? A glimpse at the charts shows that 32 Indiana ranks “worse” than South Carolina and Texas in registration and voting rates, as well as in black elected officials (Figures I, II and III). As to federal observers, Indiana appears clearly “better”—it received none (Figure IV). As to successful § 2 suits South Carolina and Texas are “worse” than Indiana, but all three are below the top ten offenders, which include five uncovered states (Figure V). This distinction in evaluating the different states’ policies is rational? Despite a congressional record of over 15,000 pages and 22 hearings, Shelby County, 811 F. Supp. 2d at 496, there is little to suggest that § 4(b)’s coverage formula continues to capture jurisdictions with especially high levels of voter discrimination. To the extent that the answer is, as the district court suggested, that Congress wished to “continue to focus on those jurisdictions with the worst historical records of voting discrimination,” id. at 506, such an overwhelming focus on historical practices appears foreclosed by Northwest Austin’s requirement that current burdens be justified by current needs. It goes without saying that racism persists, as evidenced by the odious examples offered by the majority, see Maj. Op. at 27-29. But without more evidence distinguishing current conditions in the covered jurisdictions from those in the uncovered ones, § 4(b)’s coverage formula appears to be as obsolete in practice as one would expect, in a dynamic society, for markers 34-to-59 years old. Accordingly, I dissent. * * * The analysis above is my sole basis for finding § 4(b) of the VRA unconstitutional and thus for dissenting from the court’s opinion. I need not and do not reach the 33 constitutionality of § 5 itself. But before concluding, I want to address a critical aspect of § 5, and of some of the cases interpreting earlier versions of that section. I address it first simply as a matter of language—specifically the use of language to obscure reality—and then in relation to the words and political philosophy of the 15th Amendment. Though unnecessary to my dissent’s outcome, the troubling tension between the act’s encouragement of racial gerrymandering and the ideals embodied in the 15th Amendment seems worthy of attention. Section 5(b) makes unlawful any voting practice or procedure with respect to voting “that has the purpose of or will have the effect of diminishing the ability of any citizens of the United States on account of race or color . . . to elect their preferred candidates of choice.” 42 U.S.C. 1973c(b) (emphasis added). And of course similar phrasing has been included in § 2 since 1982. See Voting Rights Act Amendments of 1982, Pub. L. No. 97-205, § 3, 96 Stat. 131, 134 (codified at 42 U.S.C. § 1973(b)) (prohibiting policies that prevent minority groups’ equal opportunity “to elect representatives of their choice.”). The language (or a close equivalent) seems to have originated in one of the Court’s earliest opinions on § 5, though only as an offhand phrase in its explanation of how a shift from district to at-large voting might dilute minority impact: “Voters who are members of a racial minority might well be in the majority in one district, but a decided minority in the county as a whole. This type of change could therefore nullify their ability to elect the candidate of their choice.” Allen v. State Bd. of Elections, 393 U.S. 544, 569 (1969). But the use of such language became troubling in Georgia v. Ashcroft, where the Court said that in the application of § 5 “a court should not focus solely on the comparative ability of a minority group to elect a candidate of its choice.” 539 U.S. 34 461, 480 (2003) (emphasis added). The “solely” of course indicates approval of such a consideration as one among several criteria for compliance with § 5. Implied from the statutory “their” is necessarily a “they.” In the context of a statute speaking of impingements on citizens’ voting “on account of race or color,” and indeed in the universally accepted understanding of the provision, the “they” are necessarily members of minority groups. But in what sense do minority groups as such have a “preferred candidate”? Individuals, of course, have preferred candidates, but groups (unless literally monolithic) can do so only in the limited sense that a majority of the group may have a preferred candidate. Thus, when the provision is translated into operational English, it calls for assuring “the ability of a minority group’s majority to elect their preferred candidates.” This raises the question of what happened to the minority group’s own minority—those who dissent from the preferences of the minority’s majority? Of course in any polity that features majority rule, some people are bound to be outvoted on an issue or a candidate and thus to “lose”—on that round of the ongoing political game. Such losses are a necessary function of any system requiring less than unanimity (which would be hopelessly impractical). And in an open society that allows people freely to form associations, and to design those associations, some people obviously will be members of associations whose representatives from time to time express, in their name, opinions they do not share. But that again is a necessary function of having associations free to adopt a structure that empowers their leadership to speak with less than unanimous backing. 35 But the implied “they” of § 5 is not a polity in itself; nor is it an association freely created by free citizens. Quite the reverse: It is a group constructed artificially by the mandate of Congress, entirely on the lines of race or ethnicity. On what authority has Congress constructed such groups? Purportedly the 15th Amendment to the Constitution. But that says that the “right of citizens of the United States to vote shall not be denied or abridged by the United States or by any State on account of race, color, or previous condition of servitude.” It is hard to imagine language that could more clearly invoke universal individual rights. It is “citizens” who are protected, and they are protected from any denial of their rights that might be based on the specified group characteristics—race, color, or previous condition of servitude. The members of Congress who launched the amendment, said Senator Willard Warner, “profess to give to each individual an equal share of political power.” CONG. GLOBE, 40th Cong., 3d Sess. 861 (1869). The 15th Amendment was a pivot point in the struggle for universal human rights. The roots of the struggle are deep and obscure. Many trace the concept to the three great monotheistic religions, Judaism, Christianity, and Islam. See, e.g., MICHELINE R. ISHAY, THE HISTORY OF HUMAN RIGHTS (2004) (noting the contributions of these three traditions, among others). No matter how spotty the actual performance of those religions’ adherents may have been over the centuries, the idea of a single God, claiming the allegiance of all mankind, surely implies a recognition of the dignity and worth of all humans, undistorted by local group loyalties historically linked to local gods. Perhaps the Enlightenment, though in tension with organized religion, has a better title; it is clearly the immediate root of the French Declaration of the 36 Rights of Man and of the Citizen. But at all events the 15th Amendment states a clear national commitment to universal, individual political rights regardless of race or color. Of course conventional political discourse often uses such terms as “the black vote,” “the youth vote,” “the senior vote,” etc. But those who use these terms—politicians, their consultants, pundits, journalists—know perfectly well that they are oversimplifications, used to capture general political tendencies, not a justification for creating or assuming a political entity that functions through a demographic group’s “majority.” The Supreme Court has recognized that these generalizations are no such justification. In Shaw v. Reno, 509 U.S. 629 (1993), it confronted racial gerrymandering that took the form of including in one district persons separated by geographic and political boundaries and who “may have little in common with one another but the color of their skin.” Id. at 647. Such a plan: bears an uncomfortable resemblance to political apartheid. It reinforces the perception that members of the same racial group—regardless of their age, education, economic status, or the community in which they live— think alike, share the same political interests, and will prefer the same candidates at the polls. We have rejected such perceptions elsewhere as impermissible stereotypes. Id. The pre-Enlightenment history of continental Europe included just such entities—“estates,” whose members voted separately from those of the other estates. Most famously, separately elected representatives of the nobility, the clergy, and the “common” people gathered in 1789 in the French Estates-General. For the last time. By the middle of that year, the Estates-General had ceased to exist. By transforming 37 itself into a National Assembly, it precipitated the French Revolution and the permanent abolition of voting by estates, ultimately throughout Europe. The 15th Amendment can be traced back to that basic development. Section 5’s mandate to advance “the ability of any citizens of the United States on account of race or color . . . to elect their preferred candidates of choice” is a partial retreat to pre-Revolutionary times, an era perhaps now so long past that its implications are forgotten. None of this is to suggest that the country need for a minute countenance deliberate voting rule manipulations aimed at reducing the voting impact of any racial group, whether in the form of restrictions on ballot access or of boundary-drawing. And in judicial proceedings to stamp out such manipulations, it would of course be no defense for the perpetrators to say that they sought only to downweight a minority’s majority. But a congressional mandate to assure the electoral impact of any minority’s majority seems to me more of a distortion than an enforcement of the 15th Amendment’s ban on abridging the “right of citizens of the United States to vote . . . on account of race, color, or previous condition of servitude.” Preventing intentional discrimination against a minority is radically different from actively encouraging racial gerrymandering in favor of the minority (really, the majority of the minority), as § 5 does. Assuming there are places in which a colorblind constitution does not suffice as a “universal constitutional principle,” Parents Involved in Community Schools v. Seattle School Dist. No. 1, 551 U.S. 701, 788 (2007) (opinion of Kennedy, J.), the voting booth should not be one of them.