State of Texas v. Holder

888 F. Supp. 2d 113, 2012 U.S. Dist. LEXIS 127119

• Court: District Court, District of Columbia
• Decided: August 30, 2012
• Docket: Civil Action No. 2012-0128
• Status: Published

Opinion

TATEL, Circuit Judge.

Pursuant to section 5 of the Voting Rights Act of 1965, Texas seeks a declaratory judgment that Senate Bill 14 (SB 14), a newly-enacted law requiring in-person voters to present a photo ID, “neither has the purpose nor will have the effect of denying or abridging the right to vote on account of race[,] color,” or “member[sMp] [in] a language minority group.” 42 U.S.C. §§ 1973c(a), 1973b(f)(2). To satisfy section 5’s effect requirement, Texas must demonstrate that SB 14 will not “lead to a retrogression in the position of racial minorities with respect to their effective exercise of the electoral franchise.” Beer v. United States, 425 U.S. 130, 141, 96 S.Ct. 1357, 47 L.Ed.2d 629 (1976). For the reasons set forth in this opinion, we find that Texas has failed to make this showing — in fact, record evidence demonstrates that, if implemented, SB 14 will likely have a retrogressive effect. Given this, we have no need to consider whether Texas has satisfied section 5’s purpose element. Accordingly, we deny the state’s request for a declaratory judgment.

I.

Under Texas’s current election code, i.e., pre-SB 14, any Texan who wishes to vote must file a registration application with the county elections registrar. That application must include the voter’s name, date of birth, and a sworn affirmation of U.S. citizenship. Tex. Elec. Code § 13.002. If the application is approved, the registrar delivers a “voter registration certificate” to the applicant, either in person or via U.S. mail. Id. §§ 13.142, 13.144. This “certificate”— actually a paper postcard — has no photograph, but does include a voter’s name, gender, year of birth, and a unique voter ID number. When presented at the polls, a voter registration certificate entitles the registrant to cast an in-person ballot.

Registered voters who fail to present a voter registration certificate may nonetheless cast an in-person ballot if they (1) execute an affidavit stating that they do not have their certificate, and (2) present an alternate “acceptable” form of identification. Id. §§ 63.008, 63.0101. In addition to a voter registration certificate, Texas’s current election code recognizes eight broad categories of documents as “acceptable” voter ID. These include birth certificates, expired and non-expired driver’s licenses, U.S. passports, U.S. citizenship papers, utility bills, “official mail addressed to the person ... from a governmental entity,” any “form of identification containing the person’s photograph that establishes the person’s identity,” and “any other form of identification prescribed by the secretary of state.” Id. § 63.0101. All in-person voters are subject to these ID requirements regardless of age or physical condition. But certain voters — including those who are 65 or older, disabled, or expect to be absent or in jail on Election Day — may choose to vote by mail without presenting identification. Id. §§ 82.001-004.

Senate Bill 14, enacted in 2011, is more stringent than existing Texas law. If implemented, SB 14 will require in-person voters to identify themselves at the polls using one of five forms of government-issued photo identification, two state and three federal: (1) a driver’s license or personal ID card issued by the Texas Department of Public Safety (DPS); (2) a license to carry a concealed handgun, also issued by DPS; (3) a U.S. military ID card; (4) a U.S. citizenship certificate with photograph; or (5) a U.S. passport. Tex. Elec. Code § 63.0101 (January 1, 2012). Unlike Texas’s current code, which allows voters to present either photographic or non-photographic ID, SB 14 requires every form of acceptable ID to include a photograph of the voter. Also unlike the current code, SB 14 prohibits the use of IDs that have expired more “than 60 days before the date of presentation” at the polls. Id. Finally, SB 14 will prohibit voters from identifying themselves using only the pictureless “voter registration certificate” issued by a county registrar.

Prospective voters lacking one of the forms of photo ID listed in SB 14 will be able to obtain a photographic “election identification certifícate” (EIC) for use at the polls. A pocket-sized card “similar in form to ... a driver’s license,” Tex. Transp. Code § 521A.001(e), an EIC, like a driver’s license, will be distributed through the DPS, and prospective voters will have to visit a DPS office to get one.

Although SB 14 prohibits DPS from “collect[ing] a fee for an [EIC],” id. § 521A.001(b), EICs will not be costless. Not only will prospective voters have to expend time and resources traveling to a DPS office, but once there they will have to verify their identity by providing “satisfactory” documentation to DPS officials. Specifically, prospective voters will need to provide (1) one piece of “primary identification,” (2) two pieces of “secondary identification,” or (3) one piece of “secondary identification” plus two pieces of “supporting identification” in order to receive an EIC. 37 Tex. Admin. Code § 15.182. A “primary” identification is an expired Texas driver’s license or personal identification card that has been expired for at least 60 days but not more than two years. Id. § 15.182(2). A “secondary” identification is one of the following:

• an original or certified copy of a birth certificate;

• an original or certified copy of a court order indicating an official change of name and/or gender; or

U.S. citizenship or naturalization papers without an identifiable photo.

Id. § 15.182(3). A wide array of documents qualify as “supporting identification,” including school records, Social Security cards, pilot’s licenses, and out-of-state driver’s licenses. Id. § 15.182(4).

In sum, SB 14 will require every EIC applicant to present DPS officials with at least one of the following underlying forms of identification:

• an expired Texas driver’s license or personal ID card;

• an original or certified copy of a birth certificate;

• U.S. citizenship or naturalization papers; or

• a court order indicating a change of name and/or gender.

Importantly, it costs money to obtain any of these documents. This means that EIC applicants — i.e., would-be voters— who possess none of these underlying forms of identification will have to bear out-of-pocket costs. For Texas-born voters who have changed neither their name nor gender, the cheapest way to obtain the required documentation will be to order a certified copy of their birth certificate from the Texas Bureau of Vital Statistics at a cost of $22. See Advisory Regarding Election Identification Certificates, ECF No. 308, at 2. (A copy of a court order indicating a change of name and/or gender costs $5 for the records search, plus $1 per page for the court order. Actually obtaining a legal change of name and/or gender costs far more — at least $152. See Attorney General’s Response to the State’s Advisory Regarding Election Identification Certificates, ECF No. 330, at 2-3.) More expensive options exist as well, ranging from $30 for an “expedited” birth certificate order all the way up to $354 for a copy of U.S. citizenship or naturalization papers. See, e.g., Advisory Regarding Election Identification Certificates, ECF No. 308, at 2.

SB 14 largely retains Texas’s existing rules for elderly and disabled voters. Voters over age 65 will still be able to vote by mail, although they will have to present an SB 14-qualifying photo ID if they choose to vote at the polls. Disabled voters, too, will be able to continue voting by mail, and those who choose to vote at the polls will still be able to identify themselves using the photoless postcard “voter registration certificate” issued by county elections registrars. To obtain this latter exemption, however, disabled Texans will need to provide written documentation of disability from either the Social Security Administration or Department of Veterans Affairs. Tex. Elec. Code § 13.002®.

Texas Governor Rick Perry signed SB 14 into law on May 27, 2011. The law, however, has yet to take effect because,'as a jurisdiction covered by section 5 of the Voting Rights Act of 1965, 28 C.F.R. pt. 51 App., Texas may not implement any change in its voting procedures without first obtaining “preclearance” from either the United States Attorney General or a three-judge panel of this court. 42 U.S.C. § 1973c(a). To obtain preclearance, Texas must demonstrate that SB 14 “neither has the purpose nor will have the effect of denying or abridging the right to vote on account of race[,] color,” or “member[ship] [in] a language minority group.” Id. §§ 1973c(a), 1973b(f)(2).

Texas filed a preclearance application with the Attorney General on July 25, 2011. Under the Voting Rights Act, the Attorney General has sixty days to “interpose[] an objection” to a changed voting procedure. Id. § 1973c(a). But here that process was delayed by the Attorney General’s requests for additional information as to (1) the number of voters who lack a DPS-issued driver’s license or personal ID card, and (2) the percentage of those voters who are minorities. See 28 C.F.R. §§ 51.37(b), 51.39(a)(1) (stating that when supplemental submissions are provided “the 60-day period for the pending submission will be recalculated from the Attorney General’s receipt of the supplementary information.”). Nearly six months after filing its initial preclearance request, on January 12, 2012, Texas submitted to the Attorney General a computer-generated list of 795,955 registered voters it was unable to match with corresponding entries in DPS’s driver’s license and personal ID database. This “no-match” list consisted of “304,389 voters (38.2%) who are Hispanic and 491,566 (61.8%) who are non-Hispanic.” Am. Compl., ECF No. 25 Ex. 7 at 2. But Texas warned that it had “reservations about the reliability of [its] data.” Compl., ECF No. 1 Ex. 5 at 1. Specifically, Texas explained that its DPS database and its voter registration list “were not designed to be merged,” and that “name changes [and] inconsistent use of nicknames or initials” between the two lists could cause “numerous incorrect 'no-match’ results.” Id. at 2. Moreover, pointing out that it had used Spanish surnames as a proxy for Hispanic voters— “an imprecise substitute for accurate racial data” — Texas explained that its no-match list constituted an unreliable estimate of ID possession rates among Hispanic voters. Id.

On March 12, 2012, the Attorney General denied preclearance, concluding that Texas had failed to show that SB 14 will not have “the effect of denying or abridging the right to vote on account of race”— i.e., that it will not have a retrogressive effect. Am. Compl., ECF No. 25 Ex. 7 at 1-2 (citing Georgia v. United States, 411 U.S. 526, 93 S.Ct. 1702, 36 L.Ed.2d 472 (1973); 28 C.F.R. § 51.52). The Attorney General gave two reasons for the denial. First, without responding to Texas’s concerns about the reliability of its no-match list, the Attorney General concluded that Texas’s data showed that “Hispanic registered voters are more than twice as likely as non-Hispanic registered voters to lack” a DPS-issued driver’s license or ID card. Id. at 2. Second, Texas had failed to show that the availability of a purportedly “free” EIC would mitigate the “impact of S.B. 14 on Hispanic registered voters.” Id. at 3. The Attorney General pointed out that if a prospective voter lacks the documents needed to obtain an EIC, “the least expensive option will be to spend $22 on a copy of the voter’s birth certificate.” Id. Fur thermore, “an applicant for an [EIC] will have to travel to a driver’s license office,” yet “in 81 of the state’s 254 counties, there are no operational driver’s license offices,” and many of those offices have limited hours of operation. Id. at 4. These constraints, the Attorney General concluded, could impose additional burdens on prospective voters who need an EIC — particularly on those without a car. Id. Moreover, Texas had “failed to propose, much less adopt, any program for individuals who have to travel a significant distance to a DPS office, who have limited access to transportation, or who are unable to get to a DPS office during their hours of operation.” Id. at 5. Given all this, the Attorney General concluded that Texas “has not met its burden of proving that ... the proposed requirement will not have a retrogressive effect, or that any specific features of the proposed law will prevent or mitigate that retrogression.” Id. Although the Attorney General’s denial rested on the potential retrogressive effect of SB 14 on Hispanic voters, he noted that Texas had “provided no data on whether African American or Asian registered voters are also disproportionately affected by S.B. 14.” Id. at 3.

Finally, the Attorney General declined to determine whether SB 14 had been enacted with a discriminatory purpose — an independent reason for denying preclearance. Because Texas “failed to meet its burden of demonstrating that the proposed law will not have a retrogressive effect,” the Attorney General explained, he had no need to “make any determination as to whether the state has established that the proposed changes were adopted with no discriminatory purpose.” Id. at 5.

In the meantime, while the Attorney General was considering SB 14, he denied preclearance to South Carolina’s new voter ID law. Shortly thereafter, on January 24, Texas, noting the South Carolina denial and the “seeming probability of an eventual rejection of Senate Bill 14 by DOJ,” filed this request for judicial preclearance. See Compl., ECF. No. 1 at 8. Although Texas’s initial complaint sought only a declaratory judgment of preclearance, the state later added a claim that section 5 of the Voting Rights Act, as reauthorized in 2006, “exceeds the enumerated powers of Congress and conflicts with Article TV of the Constitution and the Tenth Amendment.” Am. Compl., ECF No. 25 at 1-2.

The Attorney General of the United States was listed as the named defendant. (For the sake of clarity, we hereinafter refer to the party-defendant in this case as the “United States,” and refer to the “Attorney General” only when discussing administrative preclearance decisions). In addition, we later granted motions to intervene filed by several voting rights groups, as well as a number of organizations representing racial minorities in Texas. See Minute Order, 04/13/2012. These included the Texas Legislative Black Caucus, the League of Women Voters of Texas, the Southwest Voter Registration Education Project, and the Mexican American Caucus of the Texas House of Representatives. We also granted motions to intervene filed by several individual Texas voters. In order to reduce the litigation burden on Texas, we directed all intervenors to consolidate their briefing and argument. See id.

Following the Attorney General’s March 12 denial of preclearance, this litigation took on obvious urgency, as it represented Texas’s only chance of implementing SB 14 before the November 2012 elections. Although the D.C. Circuit recently affirmed the facial constitutionality of section 5, Shelby County v. Holder, 679 F.3d 848 (D.C.Cir.2012), we remain cognizant of the Supreme Court’s holding in Northwest Austin Municipal Utility District No. One v. Holder that section 5 imposes “substan tial federalism costs,” 557 U.S. 193, 202, 129 S.Ct. 2504, 174 L.Ed.2d 140 (2009) (internal quotation marks omitted). As a result, on March 27, before the United States had even filed an answer to Texas’s amended complaint, we granted Texas’s request for an expedited litigation schedule. In doing so, we rejected the United States’s contention that a trial was infeasible before the end of the summer, scheduled a one-week trial on the judicial preclearance issue to begin on July 9, and promised to issue our decision by August 31 — the date on which Texas needed a decision in order to implement SB 14 in time for the November election. See Order, ECF No. 107 at 1. As we explained, it would “raise serious constitutional questions” if Texas were prevented from implementing SB 14 merely because the United States was too busy to prepare for trial. See Northwest Austin, 557 U.S. at 204, 129 S.Ct. 2504. These federalism concerns are particularly acute in the voter ID context. After all, states not covered by section 5 have successfully implemented voter ID laws to “deter[ ] and detect[ ] voter fraud .... improve and modernize election procedures .... [and] safeguard[ ] voter confidence.” Crawford v. Marion Cnty. Election Bd., 553 U.S. 181, 191, 128 S.Ct. 1610, 170 L.Ed.2d 574 (2008). Thus, given our “historic tradition that all the States enjoy ‘equal sovereignty,”’ Northwest Austin, 557 U.S. at 203, 129 S.Ct. 2504 (citation omitted), we thought it essential to ensure that Texas had every possible opportunity to show that its own voter ID law could be implemented in time for the November elections. With the consent of the parties, we deferred consideration of Texas’s constitutional challenge, explaining that this claim “shall not be addressed unless the Court denies judicial preclearance of [SB 14].” Initial Scheduling Order, ECF No. 43 at 1. We then set an accelerated discovery and briefing schedule. Id.

Similar federalism concerns influenced our resolution of several discovery disputes. For example, seeking to show that SB 14 was motivated by discriminatory purpose, the United States moved to compel the production of testimony and documents from Texas state legislators. See Order, ECF No. 167. Texas sought to withhold this evidence, arguing that its production would violate legislative privilege. Cognizant that “federal intrusion into sensitive areas of state and local policymaking” imposes “substantial federalism costs,” Northwest Austin, 557 U.S. at 202, 129 S.Ct. 2504, and guided by Arlington Heights v. Metropolitan Housing Development Corp., we largely sided with Texas. See 429 U.S. 252, 268, 97 S.Ct. 555, 50 L.Ed.2d 450 (1977) (recognizing a testimonial and evidentiary privilege for “members of [a] decisionmaking body”). We shielded all evidence relating to “legislative acts” or “a legislator’s motivations with respect to a bill.” Order, ECF No. 167 at 11. We also allowed Texas to withhold certain communications between legislators and executive agencies. Id. at 9. Finally, we shielded most, though not all, evidence in the possession of Texas Lieutenant Governor David Dewhurst. See Order, ECF No. 154. This latter issue was complicated by the fact that the Texas lieutenant governor serves both as a member of the executive branch and as President of the Senate, so the degree to which he qualifies as a “legislator” is unclear. Id. at 4. Describing this as a “very close call,” and believing ourselves “obliged to apply the Voting Rights Act in a manner that minimizes federal intrusion,” we erred on the side of shielding evidence in the lieutenant governor’s possession. Id. at 6 (quotation omitted).

Our efforts to accelerate this litigation, however, were often undermined by Texas’s failure to act with diligence or a proper sense of urgency. As memorialized in our May 7, 2012 order, Texas “repeatedly ignored or violated directives and orders of this Court that were designed to expedite discovery.” Order, ECF No. 107 at 2. Most significantly, Texas failed to produce its voter registry, DPS ID, and license-to-carry databases to the United States until 35 days after the established discovery deadline. Id. Production of these databases to DefendanWIntervenors took place even later — 40 days after the initial deadline — and was further complicated by data-accessibility errors. See Notice Concerning Database Discovery, ECF No. 119 at 2. These errors seriously hindered DefendanWIntervenors’ ability to prepare and proffer expert testimony based on this data. See Order, ECF No. 137 at 2-4. Citing these delays, the United States again moved to postpone trial. We denied this motion, explaining that “[djespite the fact that patience and equity do not count in Texas’[s] favor when considering discovery and scheduling issues, the statute requires our best efforts for an early trial date.” Id. at 2-3. After all, we emphasized, “[t]he questions under the Voting Rights Act presented here are too important to let even Texas’[s] missed discovery ... force a change to the July 9 trial date.” Id. at 2 (citation omitted).

Although Texas was able to maintain the July 9 trial date, its dilatory approach to discovery prevented it from obtaining one potentially crucial piece of evidence. Nearly a month after discovery commenced, Texas served the Attorney General with a discovery request seeking data regarding the three types of federal ID permitted by SB 14: U.S. passports, military ID cards, and citizenship certificates. Memo, in Support of Mot. to Compel, ECF No. 130-1 at 1-2. Texas sought access to this federal data to determine the number of Texas voters who lack any form of SB 14-qualifying ID. In response, the Attorney General informed Texas that because federal identification databases are outside his “possession, custody or control,” he was unable to produce them. See Order, ECF No. 179 at 2. He advised Texas to serve subpoenas on the three U.S. agencies who physically control the databases— the Departments of State, Homeland Security, and Defense. See id. at 2-3. Inexplicably, however, Texas never served these subpoenas. Indeed, for thirty days Texas failed to take any action at all vis-á-vis the federal databases. Texas finally filed a motion to compel the Attorney General’s production of the federal databases on May 21 — the last possible day to file such a motion. Order, ECF No. 137 at 4. We ultimately denied this motion, explaining that Texas had failed to establish that the Attorney General maintains control over the databases. Order, ECF No. 179 at 4. And because Texas had mysteriously failed to serve subpoenas on the agencies in physical possession of the databases, we concluded that “[a]ny prejudice to Texas from the failure to obtain this information is assignable solely to Texas.” Id. at 5.

Nevertheless, mindful that the federal databases could prove crucial to Texas’s case, we asked the state to decide: would it rather (1) commence trial on July 9, 2012 without federal data, or (2) delay trial, potentially obtain access to the databases, but risk an inability to implement SB 14 for the November 2012 elections? Texas responded clearly and unequivocally: it preferred to go ahead with the July 9 trial date, even without access to the federal databases. Id. at 6-7. Texas’s counsel even downplayed the importance of federal data, stating: “I don’t want to give the impression that if we can’t get [information on federal IDs], we don’t think we can prove our case.” Id.

As Texas requested, trial commenced on July 9. Over the course of the week-long trial, we heard live testimony from 20 witnesses, including election lawyers; Texas state legislators; civil rights leaders; and experts in history, political science, and statistics. The parties also submitted thousands of pages of deposition testimony, expert reports, scholarly articles, and other paper evidence. The trial concluded with three-and-a-half hours of closing arguments.

Based on this extensive record, Texas argues that SB 14 was enacted to prevent voter fraud, and denies that race was a motivating factor. Texas also argues that record evidence affirmatively proves that SB 14 will have no discriminatory effect. For their part, the United States and Defendant-Intervenors argue that the specter of in-person voter fraud is a chimera meant to mask the discriminatory purpose behind SB 14. According to these parties, the record contains virtually no evidence of in-person voter fraud in Texas and this, combined with certain procedural irregularities that occurred during the passage of SB 14, the state’s history of racial discrimination, and other evidence, proves that the bill’s purpose was to disenfranchise minorities. Moreover, the United States and D efendant-Intervenors argue that SB 14 will have a discriminatory effect — that is, it will “lead to a retrogression in the position of racial minorities with respect to their effective exercise of the electoral franchise.” Beer, 425 U.S. at 141, 96 S.Ct. 1357.

In resolving these legal issues, we do not review the Attorney General’s denial of preclearance, but determine for ourselves whether SB 14 has the purpose or effect “of denying or abridging the right to vote on account of race[,] color,” or “member[ship] [in] a language minority group.” 42 U.S.C. §§ 1973c(a), 1973b(f)(2); see 28 C.F.R. § 51.49 (“The decision of the Attorney General not to object to a submitted change or to withdraw an objection [under section 5] is not reviewable.”). We do so in the following opinion, which “shall constitute the Court’s Findings of Fact and Conclusions of Law as required by the Federal Rules.” City of Rome, Ga. v. United States, 472 F.Supp. 221, 223 (D.D.C.1979) (three-judge court); see also Fed.R.Civ.P. 52 Advisory Notes 1946 (stating that findings of fact “should be a part of the judge’s opinion and decision, either stated therein or stated separately”).

II.

Before examining the evidence, we set forth the legal framework that governs this case.

A.

As the Supreme Court has “often reiterated[,] ... voting is of the most fundamental significance under our constitutional structure.” Illinois State Bd. of Elections v. Socialist Workers Party, 440 U.S. 173, 184, 99 S.Ct. 983, 59 L.Ed.2d 230 (1979). Indeed, the right to vote free from racial discrimination is expressly protected by the Constitution. The Fifteenth Amendment provides that “[t]he right of citizens of the United States to vote shall not be denied or abridged ... on account of race, color, or previous condition of servitude.” U.S. Const, amend. XV. Moreover, the Supreme Court has held that the Fourteenth Amendment, which prohibits states from “denying] to any person ... the equal protection of the laws,” U.S. Const, amend. XIV, applies to voting. As the Court has explained, “once the franchise is granted to the electorate, lines may not be drawn which are inconsistent with the Equal Protection Clause of the Fourteenth Amendment.” Harper v. Virginia State Bd. of Elections, 383 U.S. 663, 665, 86 S.Ct. 1079, 16 L.Ed.2d 169 (1966). Adopted in the immediate aftermath of the Civil War, these two amendments were aimed at protecting the rights and liberties of freed slaves in the former Confederacy.

Despite these Constitutional safeguards, “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, 86 S.Ct. 803, 15 L.Ed.2d 769 (1966). Following Reconstruction, many Southern states began enacting ballot access measures which were “specifically designed to prevent Negroes from voting.” Id. at 310, 86 S.Ct. 803. “Among the most notorious devices were poll taxes, literacy tests, grandfather clauses, and property qualifications.” Shelby Cnty., 679 F.3d at 853. Though race-neutral on their face, such measures were deliberately calculated to reduce the number of African Americans able to vote. See Katzenbach, 383 U.S. at 311, 86 S.Ct. 803 (noting that when literacy tests were enacted, “more than two-thirds of the adult Negroes were illiterate while less than one-quarter of the adult whites were unable to read or write”); id. at 311 n. 9, 86 S.Ct. 803 (quoting South Carolina Senator Ben Tillman as stating, “The only thing we can do as patriots and as statesmen is to take from the ‘ignorant blacks’ every ballot that we can under the laws of our national government.”) (alterations omitted). The Supreme Court ultimately invalidated many of these laws on the grounds that they violated the Fifteenth Amendment. Id. at 311-12, 86 S.Ct. 803 (collecting cases). Nevertheless, states were able to stay one step ahead of the courts “ ‘by passing new discriminatory voting laws as soon as the old ones had been struck down.’ ” Beer, 425 U.S. at 140, 96 S.Ct. 1357 (quoting H.R. Rep. No. 94-196, at 57-58 (1975)).

It was against this backdrop of “unremitting and ingenious defiance of the Constitution” that Congress passed the Voting Rights Act of 1965. Katzenbach, 383 U.S. at 309, 86 S.Ct. 803. Enacted pursuant to Congress’s authority to enforce the Fifteenth Amendment “by appropriate legislation,” U.S. Const, amend. XV, the Act was intended to eliminate the “insidious and pervasive evil” of racial discrimination in voting. Katzenbach, 383 U.S. at 309, 86 S.Ct. 803. As relevant here, section 5 of the 1965 Act required certain “covered jurisdictions” to “preclear” every proposed change in their voting procedures with either the Attorney General or a three-judge panel of this court. 42 U.S.C. § 1973c(a). Only if a covered jurisdiction can demonstrate that a proposed change “neither has the purpose nor will have the effect of denying or abridging the right to vote on account of race or color” will that change take effect. Id. Thus, by requiring a covered jurisdiction to preclear a change before implementing it, section 5 “shift[ed] the advantage of time and inertia from the perpetrators of the evil to its victims.” Katzenbach, 383 U.S. at 328, 86 S.Ct. 803.

Under the 1965 Act, a jurisdiction was “covered” by section 5 if it “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.” Voting Rights Act of 1965, Pub. L. No. 89-110, § 4(b), 79 Stat. 437, 438. In crafting this formula, “Congress chose [its] criteria carefully.” Shelby Cnty., 679 F.3d at 855. “It knew precisely which states it sought to cover” — those with the worst legacy of racial discrimination in voting — “and crafted the criteria to capture those jurisdictions.” Id. “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 record of racial discrimination in voting.” Id. (internal quotations and citations omitted).

Although section 5 was enacted as a temporary provision, Congress has consistently renewed it: in 1970 (for five years), in 1975 (for seven years), in 1982 (for twenty-five years), and in 2006 (for twenty- five years). Since its enactment, the relevant portions of the Voting Rights Act have largely remained the same, with one exception of particular significance to this case. In 1975, Congress expanded the coverage formula to include jurisdictions that had substantial non-English-speaking populations but provided English-only voting materials at the polls. 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)). It was this change that brought Texas within the scope of Section 5’s coverage. 28 C.F.R. pt. 51 App.

One final point bears particular emphasis: under section 5, the covered jurisdiction bears the burden of proof. This means that a covered jurisdiction must show by a preponderance of the evidence that a proposed voting change lacks both (1) discriminatory purpose and (2) retro„gressive effect. As the Supreme Court has recognized, this is a “difficult burden,” for “[a]s a practical matter it is never easy to prove a negative.” Reno v. Bossier Parish Sch. Bd. (“Bossier Parish V), 520 U.S. 471, 480, 117 S.Ct. 1491, 137 L.Ed.2d 730 (1997) (quoting, in part, Elkins v. United States, 364 U.S. 206, 218, 80 S.Ct. 1437, 4 L.Ed.2d 1669 (1960)). Nevertheless, the burden of proof in section 5 cases is both “well established,” Georgia, 411 U.S. at 538, 93 S.Ct. 1702, and uncontested by Texas.

B.

At the outset, Texas makes two arguments that, if correct, would allow it to prevail as a matter of law. We consider each in turn.

First, Texas argues that application of section 5’s effect element to voter ID laws is inappropriate because such laws can never “deny[ ] or abridg[e] the right to vote.” 42 U.S.C. § 1973c(a). According to Texas, voter ID requirements are, at worst, a “minor inconvenience[ ],” analogous to “laws requiring citizens to register to vote.” Proposed Findings of Fact by State of Texas (“Texas Proposed Findings”), ECF No. 202 at 42. Of course, “many citizens decide that the benefits of voting are not worth the burdens associated with registering to vote.” Id. But this, Texas contends, is precisely the point: would-be voters who refuse to countenance “minor inconveniences,” like registration requirements, have chosen not to vote. Similarly, Texas contends that voters who opt to go without photo ID and decline to obtain one prior to the election have eschewed their right to vote. In either case, Texas concludes, the choice lies with prospective voters, so voting rights can hardly be considered to have been “denied” or “abridged” by the state. Id. at 43.

This argument completely misses the point of section 5. As explained above, covered jurisdictions must prove that any change in voting procedures would not “deny[ ] or abridge the right to vote.” 42 U.S.C. § 1973c(a). This is true “no matter how small” the change. Allen v. State Bd. of Elections, 393 U.S. 544, 568, 89 S.Ct. 817, 22 L.Ed.2d 1 (1969). But in an attempt to advance its own definition of “deny” and “abridge” — one that would essentially exempt voter ID laws from section 5 preclearance — Texas ignores what the Supreme Court has said these terms mean. We thus repeat it here: in order to meet their burden, covered jurisdictions must show that none of their “voting-procedure changes ... would lead to a retrogression in the position of racial minorities with respect to their effective exercise of the electoral franchise.” Beer, 425 U.S. at 141, 96 S.Ct. 1357. In other words, covered jurisdictions must show that any change in voting procedures will not “worsen the position of minority voters” compared to the general populace. Reno v. Bossier Parish Sch. Bd. (“Bossier Parish II”), 528 U.S. 320, 324, 120 S.Ct. 866, 145 L.Ed.2d 845 (2000). And while it is true that some voter ID laws impose only “minor inconvenience” and present little threat to the “effective exercise of the electoral franchise” — and would thus be easily precleared under section 5 — this cannot be the case for all potential voter ID laws. If, for example, a state charged $500 for acceptable forms of voter ID, obtaining that ID would impose more than a “minor inconvenience.” The same would be true if voters were forced to travel to a distant and inaccessible state capital to obtain an ID. Again, we emphasize that Texas bears the burden of proof. Accordingly, if, as Texas argues, SB 14 imposes only a “minor inconvenience” on voters, the consequence of that argument is not that SB 14 would be exempt from section 5, but rather that it could easily be precleared because it would not undermine minorities’ “effective exercise of the electoral franchise.” Beer, 425 U.S. at 141, 96 S.Ct. 1357.

Our rejection of Texas’s unqualified assertion that laws are immune from section 5 so long as they can be tied to “voter choice” should come as little surprise, for another three-judge district court recently rejected a similar argument advanced by none other than the State of Texas. In Texas v. United States, 831 F.Supp.2d 244 (D.D.C.2011) (three-judge court), the court denied Texas’s motion for summary judgment requesting preclearance of its redistricting plan. Along the way, the court rejected Texas’s contention that if Hispanic voters would only choose to vote at the same rate as whites, a legislative district with a 50.1% Hispanic citizen voting-age population would provide Hispanics the ability to elect their preferred candidates. Id. at 262-66. The court noted that “educational and economic conditions [are] such that mere attainment of citizen voting-age status might have no real effect on [Hispanics’] ability to elect representatives of choice.” Id. at 264. The court thus concluded that it was required to perform “a more complicated retrogression analysis than Texas wants this court to approve.” Id.

Comparable logic applies here. Just as educational and economic conditions might affect whether minorities “choose” to vote, those conditions could also affect whether minorities “choose” to obtain photo ID. Poorer people, for example, may be disproportionately unable to pay the costs associated with obtaining SB 14-qualifying ID. Thus, cognizant of the decision of our sister court and fully persuaded by its reasoning, we decline Texas’s recycled invitation to collapse the entire retrogression analysis into a question of voter “choice.”

Texas’s second argument rests on the Supreme Court’s decision in Crawford v. Marion County Election Board. There, the Court upheld Indiana’s voter ID law, holding that it “imposes only a limited burden on voters’ rights” under the First and Fourteenth Amendments. 553 U.S. at 202-03, 128 S.Ct. 1610. In some respects, the Indiana law is similar to SB 14, requiring in-person voters to present photo ID at the polls, while also requiring Indiana driver’s license offices to provide free photo ID. Id. at 185-86, 128 S.Ct. 1610. Moreover, like Texas, Indiana’s chief justification for its ID law was the prevention of in-person voter fraud. Id. at 191, 194-96, 128 S.Ct. 1610. Given these similarities, Texas contends that Crawford controls this case, especially in light of the Supreme Court’s pronouncement that infringement on the “equal sovereignty” of states raises “serious constitutional questions.” Northwest Austin, 557 U.S. at 203-04, 129 S.Ct. 2504. After all, Texas argues, if Indiana can implement a photo ID law to protect against voter fraud, why can’t Texas do the same?

By contrast, the United States argues that Crawford is largely irrelevant to this litigation. It points out that Crawford involved a First and Fourteenth Amendment facial challenge to a voter ID law enacted by a state not covered by section 5. As such, the Crawford plaintiffs, who sought to have the law invalidated, bore a “heavy burden of persuasion,” requiring them to show that the law was invalid “in all its applications.” Crawford, 553 U.S. at 200, 128 S.Ct. 1610 (emphasis added). Here, however, Texas bears the burden of proving that SB 14 lacks discriminatory purpose and retrogressive effect. Georgia, 411 U.S. at 538, 93 S.Ct. 1702. Thus, the United States concludes, Crawford is essentially inapplicable to the issues before us.

In our view, the correct answer lies somewhere between these two positions. Contrary to Texas’s argument, Crawford does not control this case. In Crawford itself, the Court noted that it was “considering] only the statute’s broad application to dll Indiana voters.” Crawford, 553 U.S. at 202-03, 128 S.Ct. 1610 (emphases added). Here, not only do we face different questions — does SB 14 have discriminatory purpose or retrogressive effect? — but we focus on the limited subset of voters who are racial and language minorities. And unlike Indiana in Crawford, Texas bears the burden of proof.

Contrary to the position taken by the United States, however, Crawford informs our analysis of SB 14 in two important ways. The first goes to purpose. It is crucial, we think, that the Court held in Craivford that Indiana could act to prevent in-person voter fraud despite the fact that “[t]he record contains no evidence of any such fraud actually occurring in Indiana at any time in its history.” Id. at 194, 128 S.Ct. 1610 (emphasis added). Indeed, the Court emphatically held that “[ijhere is no question about the legitimacy or importance of’ this interest. Id. at 202-03, 128 S.Ct. 1610 (emphasis added). After all, “the ‘electoral system cannot inspire public confidence if no safeguards exist to deter or detect fraud or to confirm the identity of voters.’ ” Id. at 197, 128 S.Ct. 1610 (quoting Jimmy Carter and James A. Baker III, Building Confidence in U.S. Elections § 2.5 (Sept. 2005)). Given this, we reject the argument, urged by the United States at trial, that the absence of documented voter fraud in Texas somehow suggests that Texas’s interests in protecting its ballot box and safeguarding voter confidence were “pretext.” A state interest that is unquestionably legitimate for Indiana — without any concrete evidence of a problem — is unquestionably legitimate for Texas as well. As Texas points out, holding otherwise would, notwithstanding section 5’s facial validity, seriously threaten the “equal sovereignty” of states. Northwest Austin, 557 U.S. at 203, 129 S.Ct. 2504. The inquiry into whether SB 14 was enacted with discriminatory purpose thus cannot hinge on whether Texas can cite documented instances of in-person voter fraud — although, of course, other evidence, such as the types of circumstantial evidence discussed in Arlington Heights, could nonetheless suggest that Texas invoked the specter of voter fraud as pretext for racial discrimination. See Bossier Parish I, 520 U.S. at 489, 117 S.Ct. 1491 (“Other considerations relevant to the purpose inquiry include, among other things, ‘the historical background of the [jurisdiction’s] decision’; ‘[t]he specific sequence of events leading up to the challenged decision’; ‘[departures from the normal procedural sequence’; and ‘[t]he legislative or administrative history, especially ... [any] contemporary statements by members of the decisionmaking body.’ ” (quoting Arlington Heights, 429 U.S. at 268, 97 S.Ct. 555) (alterations in original)); see also Arlington Heights, 429 U.S. at 265, 97 S.Ct. 555 (noting that, in order to prove discriminatory intent, a plaintiff need not “prove that the challenged action rested solely on racially discriminatory purposes” (emphasis added)).

Our second point relates to section 5’s effect element. In Crawford, the Court thought it critical that “the photo identification cards issued by Indiana’s [Bureau of Motor Vehicles] are ... free.” 553 U.S. at 198, 128 S.Ct. 1610. Rejecting an argument that obtaining free photo ID cards imposed an undue burden on would-be voters, the Court explained:

For most voters who need them, the inconvenience of making a trip to the BMV, gathering the required documents, and posing for a photograph surely does not qualify as a substantial burden on the right to vote, or even represent a significant increase over the usual burdens of voting.

Id. This holding, though made in the context of a constitutional challenge, has obvious ramifications for this section 5 case. If in most instances the “inconvenience of making a trip to the BMV ... does not qualify as a substantial burden on the right to vote,” id., we fail to see how that same inconvenience could, absent more, undermine the “effective exercise of the electoral franchise” for minority voters. Beer, 425 U.S. at 141, 96 S.Ct. 1357. In other words, according to Crawford, there are certain responsibilities and inconveniences that citizens must bear in order to exercise their right to vote, and a one-time trip to the driver’s license office is, in most situations, simply one of those responsibilities.

It is important to bear in mind, however, that Crawford involved a facial challenge to the ID law’s effects on “all Indiana voters.” Crawford, 553 U.S. at 203, 128 S.Ct. 1610. The Court was therefore discussing the burden of “making a trip to the BMV” generally. Indeed, the Court expressly held only that the burdens associated with obtaining a photo ID were insubstantial “[f]or most voters.” Id. at 198, 128 S.Ct. 1610 (emphasis added). Obviously, “most” is different from “all.” Crawford thus cannot be read as holding that a trip to the BMV can never “qualify as a substantial burden on the right to vote.” Id. And logically so. After all, would-be voters who must take a day off work to travel to a distant driver’s license office have most certainly been exposed to burdens beyond those usually associated with voting. The same is likely true if prospective voters must pay a substantial amount of money to obtain a photo ID or wait in line for hours to get one. In some circumstances these heavy burdens could well discourage citizens from voting at all. And if such burdens fall disproportionately on racial or language minorities, they would have retrogressive effect “with respect to their effective exercise of the electoral franchise.” Beer, 425 U.S. at 141, 96 S.Ct. 1357.

The upshot of all of this is that Texas can prove that SB 14 lacks retrogressive effect even if a disproportionate number of minority voters in the state currently lack photo ID. But to do so, Texas must prove that these would-be voters could easily obtain SB 14-qualifying ID without cost or major inconvenience.

III.

With these principles in mind, we turn to the record. Because “courts have no need to find discriminatory intent once they find [retrogressive] effect,” Shelby Cnty., 679 F.3d at 869, and because evidence that a law which “bears more heavily on one race than another” — i.e., has disproportionate effect — is itself “the important starting point for assessing discriminatory intent,” Bossier Parish I, 520 U.S. at 489, 117 S.Ct. 1491 (internal quotations and citations omitted), we begin with section 5’s effect element.

This discussion proceeds as follows. We begin with Texas’s argument that, as a general proposition, voter ID laws have little effect on turnout — an argument that relies on social science literature and the experiences of Georgia and Indiana following enactment of their photo ID laws. Next, we consider evidence submitted by Texas, the United States, and DefendantIntervenors analyzing whether minorities disproportionately lack the forms of ID permitted by SB 14. For the reasons given below, we reject all of this evidence and, because Texas has submitted nothing more, conclude that the state has failed to meet its burden of demonstrating that SB 14 lacks retrogressive effect. As we shall explain, however, this case does not hinge solely on the burden of proof. Undisputed record evidence demonstrates that racial minorities in Texas are disproportionally likely to live in poverty and, because SB 14 will weigh more heavily on the poor, the law will likely have retrogressive effect.

Texas begins with a broad argument: that social science evidence demonstrates voter turnout is generally unaffected by the stringency of a state’s voter ID laws. In other words, Texas contends that voters vote regardless of the identification requirements imposed on them at the polls and that SB 14 will thus have “no significant effect at all.” Texas Proposed Findings, ECF No. 202 at 9. And because ID requirements have no bearing on whether voters — minorities or otherwise — turn out on Election Day, Texas concludes that SB 14 will have no retrogressive effect.

We are unable to credit this line of argument because the effect of voter ID laws on turnout remains a matter of dispute among social scientists. Texas relies heavily on a 2009 paper by Dr. Stephen Ansolabehere, a Harvard political scientist who (as discussed infra) happens to be one of the United States’s expert witnesses in this case. In his paper, Dr. Ansolabehere concludes, based on a telephone survey of eligible voters nationwide, that “almost no one ... stay[s] away from the polls for want of appropriate identification.” TA 1475. But the United States introduced into evidence a 2011 paper by Dr. Michael Alvarez of the California Institute of Technology which reaches precisely the opposite conclusion. Applying a statistical regression model to voting data from all 50 states, Dr. Alvarez concludes that photo ID requirements impose “significant negative burdens on voters.” U.S. Ex. 551 at 29. The Alvarez study predicts that imposition of a photo ID requirement in any given state will depress overall voter turnout by approximately 10%. Id. Texas— which bears the burden of proof — has failed to produce any evidence undermining the validity of the Alvarez study. Instead, it focuses entirely on Dr. Ansolabehere’s 2009 paper. Yet Dr. Ansolabehere himself testified that “other published research disagrees with me,” Trial Tr. 7/12/2012 (PM) 33:6, specifically pointing out that Dr. Alvarez’s study found that some photo ID laws have “quite a big effect” on turnout. Id. 102:17-18. We thus have no basis for finding that Dr. Ansolabehere’s 2009 paper represents any sort of academic consensus about the impact of voter ID laws.

Turning from national studies to state-specific data, Texas next focuses on the experiences of Indiana and Georgia — two states that recently implemented photo ID laws. Relying on expert testimony from University of Texas political scientist Daron Shaw, Texas argues that its population is demographieally “similar to” Georgia’s and Indiana’s, and that these states’ experiences with photo ID requirements suggest that SB 14 will have “no significant effect at all” on turnout in Texas. Texas Proposed Findings, ECF No. 202 at 9. At trial, Dr. Shaw testified that survey data from the 2008 Presidential primaries showed that virtually no Georgia or Indiana voters reported being turned away from the polls because of a lack of photo ID. Trial Tr. 7/11/2012 (AM) 24:6-19. Moreover, this finding remained constant across racial lines: in Indiana “0 percent of whites, 0 percent of blacks, 0 percent of Hispanics report that they were not allowed to vote”; in Georgia, “0 percent of whites, 1 percent of blacks, 0 percent of Hispanics said they were not allowed to vote” because they lacked photo ID. Id. 25:2-7. These figures were particularly notable, Dr. Shaw emphasized, because social scientists had previously concluded that “there were [disparate ID] possession rates by race” in both Georgia and Indiana. Id. 25:17-18. From this, Texas urges us to draw three conclusions: (1) photo ID laws ultimately prevent very few people from voting; (2) photo ID laws have no disproportionate effect on racial minorities; and (3) disparate ID possession rates have little effect on turnout. We reject these proposed findings because the circumstances in Georgia and Indiana are significantly different from those in Texas.

First, and most important, SB 14 is far stricter than either Indiana’s or Georgia’s voter ID laws. Indiana allows voters to use any photo ID that has “expired after the date of the most recent general election.” Ind. Code Ann. § 3-5-2-40.5(a)(3). Georgia allows voters to present any expired driver’s license at the polls. Ga. Code Ann. § 21-2-417(a)(l); see also Georgia Secretary of State, Georgia Voter Identification Requirements, available online at http://sos.georgia.gov/Gaphotoid/ (last visited August 28, 2012) (listing as “acceptable” voter ID “[a] Georgia Driver’s License, even if expired”). By contrast, SB 14 prohibits the use of an ID which has expired “more than 60 days before the date of presentation” at the polls. Tex. Elec. Code § 63.0101 (January 1, 2012).

Moreover, the burdens associated with obtaining a purportedly “free” voter ID card will be heavier under SB 14 than under either Indiana or Georgia law. This is true for at least two reasons. The first relates to out-of-pocket cost. Under SB 14, EIC applicants will have to present DPS officials with a government-issued form of ID, the cheapest of which, a certified copy of a birth certificate, costs $22. By contrast, Georgia residents may present a wide range of documents to obtain a voter ID card, including a student ID, paycheck stub, Medicare or Medicaid statement, or certified school transcript. See Ga. Elec. Code § 183-1-20-.01. The diverse range of documents accepted by Georgia (24 categories in all) means that few voters are likely to incur out-of-pocket costs to obtain a voter ID. And although Indiana law, like SB 14, requires voters to present a government-issued document (such as a birth certificate) to obtain a “free” photo ID, in Indiana the “fee for obtaining a copy of one’s birth certificate” is significantly lower than in Texas, ranging from $3 to $12, depending on the county. See Crawford, 553 U.S. at 198 n. 17, 128 S.Ct. 1610.

The second cost SB 14 will impose on EIC applicants is the burden associated with traveling to a DPS office. The United States submitted unrebutted evidence showing that “81 Texas counties have no [DPS] office, and 34 additional counties have [DPS] offices open two days per week or less.” Proposed Findings of Fact by Eric Himpton Holder, Jr. (“U.S. Proposed Findings”) Doc. 223 at 6, see also Am. Compl., ECF No. 25 Ex. 7 at 4. This means that in at least one-third of Texas’s counties, would-be voters will have to travel out-of-eounty merely to apply for an EIC. Georgia and Indiana voters face no such burdens. Indeed, Georgia law requires each county to “provide at least one place in the county at which it shall accept applications for and issue [free] Georgia voter identification cards.” Ga. Code Ann. § 21-2-417.1(a). Similarly, every Indiana county has a BMV office that is required by law to disperse “free” photo IDs. See Indiana Bureau of Motor Vehicles, Branch Locations and Hours, available online at http://www.in.gov/bmv/2337.htm (last visited August 28, 2012).

Given all this, we have little trouble finding that SB 14 will be far more burdensome than either Indiana’s or Georgia’s voter ID laws. And because the laws are so different, we place very little stock in Dr. Shaw’s comparisons among these three states.

We briefly note two additional, independent reasons to reject Dr. Shaw’s conclusions. First, Dr. Shaw’s expert report concludes that “Indiana and Georgia provide ... relevant comparisons” to Texas because both states, like Texas “have substantial minority populations.” TA 931. But these minority populations are different. As Dr. Shaw himself notes, although Indiana and Georgia both have “a sizable black population,” neither state has “Hispanic populations on the order of those in Texas.” Id. at n. 3. Of course, different minority groups have different cultural and historical experiences, and may accordingly be affected differently by similar laws. Indeed, the Supreme Court has emphasized the unique position of Texas’s Hispanic community, explaining that the “political, social, and economic legacy of past discrimination for Latinos in Texas may well hinder their ability to participate effectively in the political process.” LULAC v. Perry, 548 U.S. 399, 440, 126 S.Ct. 2594, 165 L.Ed.2d 609 (2006) (internal quotation marks and citations omitted) (emphasis added). We thus find it completely inappropriate to compare Hispanics in Texas with African Americans in Indiana or Georgia.

Second, Dr. Shaw failed to conduct any further demographic comparisons among the three states. For example, nothing in Dr. Shaw’s expert report or in his trial testimony speaks to poverty rates in Indiana or Georgia, much less whether such poverty rates break down along racial lines. As we explain below, record evidence in this case demonstrates that poverty rates in Texas do in fact break down along racial lines. Thus, without more, we have no basis for concluding that Indiana and Georgia are “relevant comparisons” to Texas. TA931.

Finally, and quite apart from the methodological flaws in Dr. Shaw’s study, his conclusion that voter ID laws do not depress voter turnout suffers from an additional weakness. Although Dr. Shaw’s expert report suggests that very few voters in Indiana and Georgia were turned away at the polls, his report does indicate that photo ID laws might dissuade some voters from attempting to cast a ballot in the first place. In the 2008 survey of Indiana voters relied upon by Dr. Shaw, 7% of eligible voters who failed to vote gave “I did not have the correct form of identification” as at least one of their reasons for not voting. See TA 933. (By way of comparison, only 2% of Texas non-voters in 2008 gave that answer. Id.). Asked about this statistic at trial, Dr. Shaw testified only that “it’s not clear to me how to treat” this response. Trial Tr. 7/11/2012 (AM) 107:15-16. “[F]or instance,” Dr. Shaw testified, “if someone says I wasn’t registered to vote, and then says [as an additional reason], and I lacked proper identification, it’s not clear to me how to treat those as a cumulative estimate of the effect of photo ID.” Id. 107:17-20. We take this point, but reiterate once again that Texas has the burden of proof, and if Indiana’s voter ID law might have discouraged up to 7% of eligible non-voters from even going to the polls, we cannot accept Texas’s proposed finding that “generally ... photo ID laws do not decrease voter turnout.” Texas Proposed Findings, ECF No. 202 at 7.

We turn next to Texas’s second major line of evidence: ID possession rates. Relying largely on telephone surveys of voters, Texas contends that Hispanics, African Americans, and whites in Texas all possess photo ID at roughly the same rates and that SB 14 will thus impose equal burdens on all voters. In response, the United States argues that Texas’s studies are defective. It further offers a study of its own, which it claims shows that Hispanic and African American registered voters are in fact nearly twice as likely as white registered voters to lack photo ID.

This discussion proceeds in four parts. In subsection 1 we describe Texas’s first survey, which allegedly shows no racial disparity in ID possession rates. In subsection 2 we summarize the United States’s study of ID possession rates, not just because the United States offers it as affirmative evidence, but also because Texas uses the study as a starting point for its second set of surveys, which, in turn, we discuss in subsection 3. Finally, in subsection 4 we address an analysis submitted by Defendant-Intervenors. For reasons explained in detail below, we find none of these studies reliable.

1.

The first relevant attempt to determine the number of Texas voters who lack photo ID came in January 2012 when Texas submitted to the Attorney General a computer-generated list of 795,955 voters it was unable to “match” with entries in the DPS ID database. As mentioned earlier, supra Part I, this “no-match” list provided a partial basis for the Attorney General’s denial of preclearance because it suggested that “Hispanic registered voters are more than twice as likely as non-Hispanic registered voters to lack” a DPS-issued driver’s license or ID card. Am. Compl., ECF No. 25 Ex. 7 at 2.

After the Attorney General denied preclearance, Texas retained Dr. Shaw (the University of Texas political science professor) to survey the individuals on the January no-match list and determine whether they actually lacked valid photo ID. Contacted by telephone, survey respondents were asked questions about their race, whether they were registered to vote, whether they possessed any form of identification required by SB 14, and whether they were disabled. TA 937. Because “certain kinds of people — those of higher socio-economic status, especially— are more likely to actually respond to poll-takers,” Dr. Shaw “weighted” his results. TA 937 & n. 5. “Weighting” involves applying a statistical formula to the final data to correct for “response bias.” As Dr. Shaw explained, “Weighting allows groups who are less likely to respond to pollsters to be properly represented in the poll.” TA 937 n. 5.

In relevant part, Dr. Shaw’s report concluded that, contrary to the Attorney General’s conclusions, Hispanic respondents lacked “any form of identification necessary for voting” at “the same rate as for white respondents,” i.e. “5% of the time (un-weighted), or 6% of the time (weighted).” TA 940. Dr. Shaw further reported that just over 9% of African Americans in the general sample (18 of 196 voters) “do not have an acceptable form of voter ID as per SB 14.” TA 941. Seeking to minimize the disparity between white and African American ID possession rates, Dr. Shaw noted that

[t]en of those 18 are over 65 years of age, however, and thus qualify for absentee ballots. And another four of the remaining eight self-identify as disabled, and thus also qualify for absentee ballots. In short, four out of 196 African Americans [or approximately 2%] reputedly ‘at risk’ due to Texas’s new voter ID law could potentially be affected by SB 14.

TA 941. Based in part on Dr. Shaw’s survey results, Texas argues that “SB 14’s photo identification requirement will not have a disparate impact on Hispanic or African-American voters in Texas.” Texas Proposed Findings, ECF No. 202 at 13.

The reliability of Dr. Shaw’s study, however, is seriously undermined by his surveys’ extraordinarily low response rates. Just over 2% of the individuals Dr. Shaw attempted to contact ultimately responded to his questions. TA 967-68. As explained in greater detail in Part III.B.3, infra, uncontested record evidence suggests that such low response rates render telephone surveys scientifically invalid. We thus find that Dr. Shaw’s survey of those on the January no-match list is methodologically unsound and therefore unreliable.

Dr. Shaw’s study suffers from at least one additional defect. At first blush, his survey suggests that African Americans in Texas are disproportionately likely to lack photo ID: over 9% of African American respondents reported lacking ID, compared to 5-6% of the general population. In fact, the actual disparity may be even larger since Dr. Shaw reported only unweighted numbers — not the weighted percentage — for African Americans. In response to this disparity, Dr. Shaw attempted to reduce the African American figure by removing disabled people and those over 65. But disabled voters and those over 65 will not be “exempt” from SB 14. Although they will be able to vote by mail without a photo ID, SB 14 imposes obligations on such voters who choose to vote in-person. Specifically, voters over the age of 65 who want to vote in person will have to present acceptable photo ID. And disabled voters who wish to identify themselves using their picture-less voter ID certificate will first have to obtain written documentation of disability from either the Social Security Administration or the Department of Veterans Affairs. Tex. Elec. Code § 13.002(i). Moreover, although Dr. Shaw removed these voters from the pool of African American voters, he failed to perform any similar reductions for the Hispanic or general samples, thus making cross-racial comparisons impossible. Dr. Shaw’s selective reduction was therefore both inappropriate and methodologically unsound.

2.

The second data set at issue is a computer-generated “no-match” list compiled in May, 2012 by the United States’s expert, Harvard political scientist Stephen Ansolabehere. Much like Texas’s January no-match list, Dr. Ansolabehere’s list attempts to show the number of Texas voters who lack state-issued photo ID. Dr. Ansolabehere compiled his list by cross-referencing Texas’s voter registry with (1) the DPS ID database (which contains both state-issued driver’s licenses and personal ID cards), and (2) Texas’s license-to-carry database.

Prior to cross-referencing these databases, Dr. Ansolabehere “cleaned” the driver’s license and license-to-carry-databases to remove duplicative and immaterial entries. Specifically, Dr. Ansolabehere removed from these databases all entries with identical social security numbers. He also removed records for driver’s licenses that were either expired or belonged to somebody marked “deceased.” U.S. Ex. 544 11-13. As Dr. Ansolabehere explained:

[Driver’s license] records that correspond to deceased persons, expired licenses, and other cases may very well match to individuals on the [voter registry]. They should not be considered valid matches as they are not valid voters or do not have a valid state identification for purposes of voter identification. Keeping these cases in the matching process would create false positives in the match and lead to “too many” matches.

Id. at 13. Dr. Ansolabehere also performed a parallel “cleaning” of the voter registry list to remove duplicate entries, although (and we shall say more about this later) he failed to remove voters who had been designated “deceased.” Id. at 11-12.

Dr. Ansolabehere then began constructing his no-match list. He defined a “matched” voter — i.e., a voter who presumably possesses state-issued ID — as “any [voter] on the Voter Registration database with the same identifying information ... on the Drivers’ License or License to Carry Database.” Id. at 5. “Same identifying information,” for the purposes of Dr. Ansolabehere’s study, constituted one of two things: a full 9-digit social security number match; or an identical first name, last name, and date of birth. Id. at 15.

Through this process, Dr. Ansolabehere constructed a no-match list consisting of nearly 1.9 million voters who apparently lacked Texas-issued photo ID. Almost 20% of these no-matches were marked by Dr. Ansolabehere as “ambiguous.” This meant one of two things. First, the no-match could be a voter with a very common name who had been matched to a state ID but without “high probabilities of certainty.” Id. (The State of Texas, for example, could well contain multiple people named “Michael Jones” born on January 6, 1981, so it would be unclear whether a “match” would be for the right Michael Jones). Second, the no-match could be a voter with a driver’s license marked “Not Eligible” to drive. As Dr. Ansolabehere noted, it is “unclear” how the “Not Eligible” notation “affects the validity of the identification for purposes of voting” under SB 14. Id. at 14.

Because Texas does not track voters by race, Dr. Ansolabehere cross-referenced his no-match list with a database provided by Catalist, LLC. A private vendor specializing in voter registration data, Catalist attempts to determine voters’ race by applying a predictive algorithm that uses “name dictionaries and residential area information.” Id. at 6, 8. As Dr. Ansolabehere explains in his expert report:

A name dictionary would identify someone named Greg Jones as 60% likely to be White based on the frequency with which that name is used in the population. Someone named Greg Bernard Jones who lives in an area that is 31% Black has an 83% probability of being Black. The combination of name information and local area information, then, sharpens the algorithm for identifying race considerably.

Id. at 8.

After cross-referencing his no-match list with Catalist’s database, and thus assigning a racial classification to every individual on the no-match list, Dr. Ansolabehere reached two conclusions:

1) If ambiguous cases are treated as no-matches, 20.71% of registered African American voters, 17.49% of registered Hispanic voters, and 10.85% of registered white voters cannot be matched with IDs in the Texas databases;

2) If ambiguous cases are treated as matches, 15.97% of registered African American voters, 14.32% of registered Hispanic voters, and 9.65% of registered white voters cannot be matched with IDs in the Texas databases.

Id. at 31. In either event, the United States contends, Dr. Ansolabehere’s data conclusively shows that registered African American and Hispanic voters in Texas are disproportionately likely to lack photo ID. For several reasons, we find that Dr. Ansolabehere’s study cannot support this conclusion.

First, Dr. Ansolabehere’s study, even were it methodologically flawless, is of limited value because it fails to examine all SB 14-qualifying ID. Recall that SB 14 permits in-person voters to present one of five types of photo ID. Two are issued by the state: a DPS-issued driver’s license or ID card, or a license to carry a concealed handgun. Three are issued by the federal government: a passport, military ID, or citizenship certificate with photograph. Tex. Elec. Code § 63.0101 (January 1, 2012). Yet Dr. Ansolabehere limited his study to the two state forms of ID. This limitation is significant because the United States relies on Dr. Ansolabehere’s study to support its broad conclusion that there exists “a substantial racial disparity in the possession of identification required by SB Bp.” U.S. Proposed Findings, ECF No. 223 at 11 (emphasis added, some capitalization altered). But for this to be true, Dr. Ansolabehere’s study would have to have considered all forms of “identification required by SB 14,” not just the two state-issued qualifying forms. Yet despite the fact that Dr. Ansolabehere was retained by the United States and expressly asked for access to the databases regarding the three forms of federal ID, Trial Tr. 7/12/2012 (AM) 98:13-21, his expert report states that “[n]o federal lists were provided to me for the sake of this analysis,” U.S. Ex. 544 at 9-10. Unless the United States is able to present a full picture of precisely who lacks any form of SB 14-approved ID, it has no basis for asking us to affirmatively find that SB 14 will discriminate against anyone — at least not on the basis of disparate ID possession rates.

The failure to analyze federal data is not the only problem with Dr. Ansolabehere’s study. It is also plagued by several methodological flaws that make it impossible to rely on it even for the more limited proposition that there exists a racial disparity in the possession of state-issued ID. First, concerned that his algorithm might result in “ ‘too many’ matches,” Dr. Ansolabehere “cleaned” the DPS database by removing 779,918 deceased drivers prior to creating his no-match list. U.S. Ex. 544 at 13. But Dr. Ansolabehere failed to remove nearly 50,000 of those exact same individuals from the voter rolls, thus virtually ensuring that these deceased voters would be included on the no-match list. We can think of no good reason for their inclusion. After all, Lyndon Johnson’s 1948 Senate race notwithstanding, the dead cannot vote in Texas. See Robert Caro, Means to an Ascent 329 (First Vintage Books 1991). True, nothing in the record suggests that the dead voters on Dr. Ansolabehere’s no-match list were disproportionately minorities. But his failure to remove them means, at the very least, that his no-match list overstates SB 14’s effect.

Second, we have serious doubts as to whether Catalist’s algorithm accurately identified the racial composition of voters in this case. Although Dr. Ansolabehere’s expert report states that Catalist is an industry leader in “identifying races based on names and Census data,” placed second in a “Multi-Cultural Name Matching Challenge,” and has been used in several academic studies, U.S. Ex. 544 at 8-9, the record contains no direct evidence as to the accuracy of Catalist’s algorithm. To the contrary, record evidence suggests-— albeit not conclusively — that Catalist’s error rate in this case may be quite high. When cross-examining Dr. Ansolabehere, Texas’s counsel demonstrated anecdotally that a number of voters on his no-match list do, in fact, possess state-issued photo ID, and further showed that the race listed on many of those voters’ IDs differed from Catalist’s racial classification. See Trial Tr. 7/12/2012 (PM) 57:24-58:5; 83:4-21; 84:11-85:7. Moreover, when Texas’s expert Dr. Shaw conducted a follow-up survey of voters on Dr. Ansolabehere’s no-match list, only 68% of respondents identified as “black” by Catalist actually self-identified as “black.” TA 973. As discussed in more detail below, see infra Section III.B.3, we have serious qualms about the methodological rigor of Dr. Shaw’s surveys and so cannot conclude that this 68% figure is accurate. Nevertheless, this number raises a red flag. In litigation like this, which depends largely on accurate racial classification, even the possibility that 32% of those classifications are wrong is simply too high.

Finally, at trial Texas demonstrated how Dr. Ansolabehere’s name matching algorithm, which requires an exact match, could lead to a number of false no-matches. For example, under Dr. Ansolabehere’s algorithm, a “Bob Thomas” on the voter registry cannot be matched with an ID for “Robert Thomas.” Nor can Juan Gonzalez be matched with Juan Gonzales. And any woman who changes her last name at marriage is a no-match if her voter registration remains in her maiden name. Importantly, however, SB 14 will permit voters to cast a ballot if the “name on the documentation is substantially similar to ... the name on the list.” Tex. Elec. Code § 63.001 (January 1, 2012) (emphasis added). Given this, Dr. Ansolabehere’s failure to match voters to state-issued IDs with “substantially similar” names undermines his conclusions. This is especially true given that English spellings of Spanish names often vary slightly (Gonzalez v. Gonzales; Delacruz v. De la Cruz)— which, in this context, would lead to a disproportionate number of no-matches for Hispanics. And notably, some of these false no-matches could probably have been eliminated. Dr. Ansolabehere testified that it is possible to employ a “fuzzy matching” subroutine, “where you take a list of potential nicknames and so forth and try to match on the basis of those, or initials and things like that.” Trial Tr. 7/12/2012 (PM) 50:18-21.

To sum up, Dr. Ansolabehere’s study excludes federal forms of ID and uses an uncertain racial classification algorithm. Moreover, his no-match list is inflated: it includes both deceased voters and voters who may have ID under a “substantially similar” name. We therefore find Dr. Ansolabehere’s study unreliable.

3.

Texas’s final study of ID possession rates rests on yet another set of telephone surveys undertaken by Dr. Shaw. Following the submission of Dr. Ansolabehere’s expert report, Texas retained Dr. Shaw to perform a second set of surveys — this time of the individuals on Dr. Ansolabehere’s no-match list. Dr. Shaw conducted three surveys: (1) a general sample of 1,000 individuals from the no-match list; (2) a sample of 600 individuals Catalist identified as African American; and (3) a sample of 600 individuals Catalist identified as Hispanic. TA 973. As in Dr. Shaw’s first survey, respondents were asked whether they possessed any form of ID required by SB 14 — both state and federal — as well as questions about their race, voting habits, disability status, and opinions on voter ID laws. TA 978-82.

Although Dr. Shaw’s second round of surveys resembled his first, two methodological features are worth highlighting. First, as mentioned above, Dr. Shaw concluded that Catalist’s racial identification algorithm was largely inaccurate. TA 973. As a result, Dr. Shaw categorized respon dents based on their self-identified race— i.e., how they answered the question “[w]hat would you say is your main race?” — rather than how they were classified by Catalist. Id.; see also TA 981. Second, unlike in his first study, Dr. Shaw “did not report weighted survey results” to the court, although he testified that he “[had] in fact, weighted” his results on his own. Trial Tr. 7/11/2012 (AM) 81:19-22.

Based on this second round of surveys, Dr. Shaw concluded that “there is no statistically significant difference in ID possession rates amongst whites, blacks, and Hispanics when the appropriate universe of ID types is accounted for.” TA 976-77. Specifically, Dr. Shaw reported that in his general sample, 9.38% of whites lacked any of the forms of ID listed in SB 14, compared with 9.30% of African Americans, and 6.18% of Hispanics. TA 976. Asking about federal forms of ID apparently made a difference: Dr. Shaw reported that “amongst Hispanics, possession of a valid passport or citizenship certificate is higher than that for the Anglo population.” TA 975. Overall, though, Dr. Shaw testified that in the general sample, race-based “difference between these rates of [ID] possession” is statistically insignificant. Trial Tr. 7/11/2012 (AM) 35:21-36:2. Moreover, Dr. Shaw noted that “[i]f we rely on data from the Hispanic and black surveys (respectively), 6.72% of Hispanics and 7.64% of blacks do not possess one or more of the photo IDs identified in SB 14.” TA 976. Both of these rates are lower than the percentage of whites who reported lacking ID in the general sample. Id.

Again, however, Dr. Shaw’s surveys were plagued by low response rates. His survey of the general sample from Dr. Ansolabehere’s list had a response rate of only 2.0%, meaning that just one out of every 50 voters he attempted to reach actually answered his questions. TA 983. Response rates for the African American and Hispanic surveys were hardly any better-just 2.5% and 2.1%, respectively. TA 984-85. Significantly, Dr. Shaw conceded that he had never obtained such low response rates during any of the live interview telephone surveys he conducted over the course of his career. Trial Tr. 7/11/2012 (AM) 71:21-25. A low response rate, he testified, is “always a concern for surveys.” Id. at 44:6. Dr. Shaw explained:

There’s always the possibility that the people you contacted are systematically different from the people you couldn’t contact either because you couldn’t reach them or because you reached them and they didn’t want to participate in the survey. This is broadly referred to as non-response bias.

Id. 44:1-5. In other words, a low response rate increases the probability that the people who actually responded are in some relevant way different than the target population. This is hardly an ancillary concern. After all, the entire point of a telephone survey in this context is to obtain a representative sample of a target population.

Here, Dr. Shaw’s concerns are well justified. Response rates of 2.0, 2. 1, and 2.5% fall far short of anything deemed acceptable in the polling industry. True, as Dr. Shaw testified, the Pew Research Center recently concluded that telephone surveys with response rates as low as 9% may be deemed reliable in some circumstances. See Pew Research Center, Assessing the Representativeness of Public Opinion Surveys, (May 15, 2012) at 8-10, available online at http://www.peoplepress.org/filesAegacy-pdf/Assessing% 20the% 20Representativeness% 20of% 20Public% 20Opinion % 20Surveys.pdf (“Pew Study”) (last visited August 28, 2012). But a 9% response rate, while low, is far higher than the response rates ob tained by Dr. Shaw — 4.5 times the rates he achieved for the general sample. Moreover, the Pew study, the validity of which is unquestioned by Texas, sets the industry floor: neither Dr. Shaw nor any other witness cited a single study that suggests that a response rate any lower than 9% is satisfactory. To the contrary, Dr. David Marker, a statistician who performs surveys for the federal government, testified that a 2% response rate is “exceedingly low,” and “out of bounds with the surveys that are requested and used by the government.” Trial Tr. 7/11/2012 (PM) 108:6-8. Dr. Marker further testified that “I don’t believe a two percent survey can provide statistically valid estimates,” id. 116:5-6, because such estimates “will not at all reflect the underlying true overall population estimates that you’re trying to understand.” Id. 108:21-23. The results of surveys with such low response rates, Dr. Marker bluntly concluded, are “really irrelevant.” Id. 108:20-21. Thus, taking our cues from the polling industry itself, we have little trouble finding that Dr. Shaw’s response rates fall well short of acceptable.

In any event, even had Dr. Shaw obtained an adequate response rate, his survey would still lack certain hallmarks of reliability. The 2012 Pew study, for example, concluded that “telephone surveys that [1] include landlines and cell phones and [2] are weighted to match the demographic composition of the population ... provide accurate data.” Pew Study at 1 (emphasis added); see also Trial Tr. 7/11/2012 (PM) 113:21-22 (Dr. Marker testifying as to the fact that the Pew study “uses cell phones and land lines”). Dr. Shaw himself testified that these are “protocols that Pew assumes and that need to be in place in order to get that sort of quality.” Trial Tr. 7/11/2012 (AM) 127:10-11 (emphasis added). Despite this, Dr. Shaw failed to include any cell phone numbers; instead, he contacted only individuals with landline phones. Id. 127:12-17. Nor did Dr. Shaw produce any “weighted results of the survey to the Court.” Id. 127:24-25. Given that these “protocols ... need to be in place” to ensure quality, Dr. Shaw’s failure to abide by them further undermines the validity of his surveys. Id. 71:21-25 (emphasis added).

At trial, Dr. Shaw sought to assure us that his weighted surveys, had they been submitted as evidence, would “look[ ] much the same” as the non-weighted surveys. Id. 130:20. Dr. Shaw testified that his survey found that groups that are harder to contact — “younger groups ... people with lower income, et cetera” — tend to possess ID at the same rate as easier-to-contact groups. Id. 40:20-41:1. Thus, Dr. Shaw testified, weighting for age or socioeconomic status “simply would have replaced a group with 90 percent identification rates with another group with 90 percent identification rates.” Id. 40:23-25. As Dr. Shaw put it:

[I]n that sense the discussion, the conversation about response rates ... while it’s important and while it’s interesting, I mean, ultimately the proof is in the pudding [I]f we were to [weight], what you’re going to find is no change. I mean, these are 90 percent possession rates. The 90 percent possession rates, self-professed possession rates occur across a lot of different groups.

Id. 130:21-131:5.

This point merits two responses. First, we are unable to accept Dr. Shaw’s vague representations about weighted data — representations that were not presented to the court and subjected to cross-examination. Second, even if ID possession rates are broadly the same across different age and socioeconomic groups, Dr. Shaw still has no basis for concluding that he obtained a representative sample within those groups. As Dr. Marker explained:

[Weighting] only goes so far. The issue is, are the young people who responded, the two percent, are they like all of the other young people? Are the elderly who responded like the elderly who didn’t? Are the blacks who responded like the typical blacks? And that doesn’t get help by weighting.

Trial Tr. 7/11/2012 (PM) 112:22-113:1.

To take but one example of this point, Dr. Shaw surveyed only people with land-line phones. And though it may be true that 18-to-29-year-old landline users have photo ID at the same rates as other land-line users, it may also be true that 18-to-29-year-olds who use only cell phones are disproportionately likely to lack photo ID. For that matter, it may be that 65-to-80-year-old Hispanic women who use only cell phones are also disproportionately likely to lack photo ID. Because Dr. Shaw declined to survey these individuals, he simply has no way of knowing.

One final note. The Pew study concluded that “[o]ne significant area of potential non-response bias ... is that survey participants tend to be significantly more engaged in civic activity than those who do not participate.” Pew Study at 2. Notably, this is true even at the 9% response level deemed acceptable by Pew. See id. It thus seems probable that, as the Pew report suggests, the voters who responded to Dr. Shaw’s survey were “significantly more” civically engaged than the general population. And it may also be that civically engaged voters of all races and income levels are significantly more likely to report possession of state-issued photo ID. This might be because the civically engaged are more likely to travel internationally (and thus have a passport); volunteer in their community (which often requires a car and driver’s license); or simply renew their existing ID on time. Or it could be, as Dr. Shaw suggested at trial, that those who care about what their neighbors think “would be embarrassed to say they don’t have a driver’s license,” even if they actually lack one. Trial Tr. 7/11/2012 (AM) 111:18-19.

True, all of this is speculative and uncertain. But that is precisely the point. We cannot tell from Dr. Shaw’s surveys precisely who he contacted, if they were representative of the target population, and whether their answers were skewed. As Dr. Shaw himself testified, this is the inherent danger in low response rates. See id. 44:1-5. Because Dr. Shaw failed to achieve a minimally acceptable response rate and to apply the polling industry’s standard practices, we find his study unreliable.

4.

DefendanWIntervenors point out that, effective January 1, 2004, Texas’s voter registration form includes a field where voters are asked to write down their driver’s license or personal ID number. Texas Elec. Code § 13.002(c)(8)(A). “[I]f the applicant has not been issued [such] a number,” the applicant instead must provide “the last four digits of the applicant’s social security number.” Id. § 13.002(c)(8)(B). If the applicant lacks a driver’s license number, personal ID number, and a social security number, the applicant then must include “a statement ... that the applicant has not been issued [any such] number.” Id. at § 13.002(c)(8)(C). According to the record, 56.4% of Texas voters were registered on or after January 1, 2004, and those registered voters were thus required to use this voter registration form. Trial Tr. 7/13/2012 106:20-107:1-3. As Defendanfr-Intervenors point out, 9.7% of those voters with Spanish surnames failed to provide a driver’s license or personal ID number on their registration form, compared with 7.5% of the general population. Id. 107:21-108:1. Relying on this “very substantial facial difference,” Defendant-Intervenors argue that Hispanic voters in Texas disproportionately lack photo ID. Id. 108:1. We cannot infer from this data anything of the sort.

The first problem is the same one that plagued Dr. Ansolabehere’s no-match list: voters who lack a driver’s license or personal ID card may nonetheless possess one of the three federal forms of identification acceptable under SB 14. Indeed, Defendant>-Intervenors’ dataset is even more limited than Dr. Ansolabehere’s. For all its problems, at least Dr. Ansolabehere’s no-match list purported to cover the entire voter registry. By contrast, Defendants Intervenors’ data covers just 56.4% of Texas voters. Furthermore, unlike Dr. Ansolabehere’s study, Defendant-Intervenors’ data fails to reflect the possibility that a voter may possess a license to carry a handgun.

In any event, faced with a registration form requesting either an 8-digit driver’s license/personal ID number or the last four digits of a social security number, even voters who possess a driver’s license may opt to provide a social security number. After all, four digits are easier to write than eight. Furthermore, many voters likely memorize their social security number but not their driver’s license number. Thus, what voters write on their registration form is barely probative of whether they actually possess a state-issued ID card — much less whether they possess any SB-14 approved form of ID.

C.

We pause to summarize the evidentiary findings we have made so far. Contrary to Texas’s contentions, nothing in existing social science literature speaks conclusively to the effect of photo ID requirements on voter turnout. Moreover, scant lessons, if any, can be drawn from Indiana and Georgia, largely because SB 14 is more restrictive than the photo ID laws adopted by either of those states. Finally, no party has submitted reliable evidence as to the number of Texas voters who lack photo ID, much less the rate of ID possession among different racial groups.

Given this, we could end our inquiry here. Texas bears the burden of proving that nothing in SB 14 “would lead to a retrogression in the position of racial minorities with respect to their effective exercise of the electoral franchise.” Beer, 425 U.S. at 141, 96 S.Ct. 1357. Because all of Texas’s evidence on retrogression is some combination of invalid, irrelevant, and unreliable, we have little trouble concluding that Texas has failed to carry its burden.

Significantly, however, this case does not hinge merely on Texas’s failure to “prove a negative.” See Bossier Parish I, 520 U.S. at 480, 117 S.Ct. 1491 (internal quotation marks omitted). To the contrary, record evidence suggests that SB 14, if implemented, would in fact have a retrogressive effect on Hispanic and African American voters. This conclusion flows from three basic facts: (1) a substantial subgroup of Texas voters, many of whom are African American or Hispanic, lack photo ID; (2) the burdens associated with obtaining ID will weigh most heavily on the poor; and (3) racial minorities in Texas are disproportionately likely to live in poverty. Accordingly, SB 14 will likely “lead to a retrogression in the position of racial minorities with respect to their effective exercise of the electoral franchise.” Beer, 425 U.S. at 141, 96 S.Ct. 1357.

The first of these facts — that there exists a subgroup of registered voters, including minorities, who lack SB 14-ap-proved photo ID — is undisputed by Texas and finds support in its own expert’s study. Although we are unable to conclude that the individuals Dr. Shaw surveyed were representative of the target population, he did find — and this finding is unaffected by the flaws in his study — a subset of Texas voters who have none of the SB 14-quali-fying IDs, neither state nor federal. Moreover, as Dr. Shaw testified, his surveys revealed that the race-based “difference between these rates of [ID] possession” is statistically insignificant. Trial Tr. 7/11/2012 (AM) 35:21-36:2. Citing Dr. Shaw’s surveys, Texas itself urges us to conclude that “there is no racial or ethnic disparity among the group of registered voters who currently lack a photo ID and would be required to obtain one under SB 14.” Texas Proposed Findings, ECF No. 202 at 19 (emphasis added). And during closing arguments, Texas’s counsel conceded that “[t]he record does tell us that there is a subset of registered voters who lack the ID,” and that “there’s a significant percentage of ... minorities and Anglos who don’t have SB 14 qualifying IDs.” See Trial Tr. 7/13/2012 (AM) 24:3-6, 25:2-6. Thus, based on Texas’s own evidence, we find that there is a subset of Texas voters who lack SB 14-approved ID' — again, both state and federal — and that, at minimum, racial minorities are proportionately represented within this subgroup.

Equally uncontested is the proposition that, for members of this subgroup to cast a regular in-person ballot under SB 14, they will have to obtain an acceptable form of photo ID, and that the cheapest option is an EIC. In order to obtain an EIC, would-be voters will need to present one of several underlying documents, and as Texas concedes, the least expensive option for most prospective voters who lack supporting identification will be a certified copy of their birth certificate — which costs at least $22. See Advisory Regarding Election Identification Certificates, ECF No. 308 at 2.

But this is not all. Recall that would-be voters will need to apply for an EIC at a DPS office, and that almost one-third of Texas’s counties (81 of 254) lack one. Supra Part I at 7. This means that many would-be voters who need to obtain an EIC — individuals who by definition have no valid driver’s license — will have to find some way to travel long distances to obtain one. This is hardly an insignificant concern, especially given that “everything is bigger in Texas.” See, e.g., Rick Perry, Amid a Dim National Economy Texas Remains in the Spotlight, October 31, 2008, available at http://www.tradeand industrydev.com/region/texas/amid-a-dimnationaleconomy-texas-remains-spotlighF554 (last visited August 28, 2012).

Reinforcing this proposition, Texas Representative Trey Martinez Fischer, who represents a district which includes the city of San Antonio and its outskirts, testified that “you will not find a DPS office from downtown San Antonio to the western boundary, which is heavily concentrated with African-Americans, and particularly Hispanics.” Trial Tr. 7/10/2012 (AM) 119:23-25. State Senator Carlos Uresti echoed this concern, testifying that in his district — -which is “70 percent Hispanic, about 5 percent African American”— “[t]here are some towns ... where the nearest DPS office is about a 100 to 125 mile[] one way” trip away. Trial Tr. 7/12/2012 (AM) 7:16-8:1. And far from disputing the long travel times imposed by the dearth of DPS offices, Texas’s counsel told us that “I don’t think that the facts of the geographic distances [between DPS offices] are necessarily contested.” Trial Tr. 7/13/2012 52:4-5.

Of course, we remain cognizant of the Supreme Court’s holding in Crawford that “for most voters ... the inconvenience of making a trip to the BMV ... does not qualify as a substantial burden on the right to vote.” Crawford, 553 U.S. at 198, 128 S.Ct. 1610. But Crawford was a facial challenge, and this general principle yields when the closest office is 100 to 125 miles away. Even the most committed citizen, we think, would agree that a 200 to 250 mile round trip — especially for would-be voters having no driver’s license — constitutes a “substantial burden” on the right to vote. Our own Federal Rules of Civil Procedure support this conclusion, specifying that witnesses are unavailable to testify if they must travel more than 100 miles to do so. See Fed.R.Civ.P. 45(c)(3).

Significantly, these burdens will fall most heavily on the poor. Like any fixed cost, the $22 (minimum) EIC applicants will have to pay to obtain prerequisite documentation weighs disproportionately on those living in poverty. Moreover, while a 200 to 250 mile trip to and from a DPS office would be a heavy burden for any prospective voter, such a journey would be especially daunting for the working poor. Poorer citizens, especially those working for hourly wages, will likely be less able to take time off work to travel to a DPS office — a problem exacerbated by the fact that wait times in DPS offices can be as long as three hours during busy months of the year. U.S. Ex. 10 at 1. This concern is especially serious given that none of Texas’s DPS offices are open on weekends or past 6:00 PM, eliminating for many working people the option of obtaining an EIC on their own time. See U.S. Ex. 361. A law that forces poorer citizens to choose between their wages and their franchise unquestionably denies or abridges their right to vote. The same is true when a law imposes an implicit fee for the privilege of casting a ballot, like the $22 many would-be voters who lack the required underlying documentation will have to pay to obtain an EIC. “[Wjealth or fee paying has ... no relation to voting qualifications; the right to vote is too precious, too fundamental to be so burdened or conditioned.” Harper, 383 U.S. at 670, 86 S.Ct. 1079.

To be sure, a section 5 case cannot turn on wealth alone. In Texas, however, the poor are disproportionately racial minorities. According to undisputed U.S. Census data, the poverty rate in Texas is 25.8% for Hispanies and 23.3% for African Americans, compared to just 8.8% for whites. Mot. to take Judicial Notice of Census Data, ECF No. 219 Ex. 4 at 7, 16. This means that the burdens of obtaining an EIC will almost certainly fall more heavily on minorities, a concern well recognized by those who work in minority communities. Lydia Camarillo, a Texas voter education specialist who has worked for over 35 years in the Hispanic community, testified that because “Latinos are often among the working poor[,] ... Latinos struggling to afford groceries, rent, and child care may not be able to afford ... a copy of a birth certificate in order to get a voter ID.” Defendant-Intervenors Ex. 9 at 224. Moreover, Camarillo testified, “[f]or working class Latinos, the requirement of travelling to the DPS during regular business hours may prevent them from obtaining ID because their work hours are not flexible.” Id.

Again, this is not all. Undisputed census data shows that in Texas, 13.1% of African Americans and 7.3% of Hispanies live in households without access to a motor vehicle, compared with only 3.8% of whites. Mot. to take Judicial Notice of Census Data, ECF No. 219 Ex. 4 at 8, 17. If traveling over 200 miles constitutes a substantial burden on people without driver’s licenses who can nonetheless find a ride to a DPS office, supra at 47-48, imagine the burden for the predominantly minority population whose households lack access to any car at all. In fact, in some places it may be impossible to get to a DPS office without a car. State Senator Rodney Ellis testified that in his “inner city district” in Houston, DPS offices are not “easily accessible by public transporta tion.” Trial Tr. 7/11/2012 (PM) 34:8-12. If DPS offices are inaccessible by public transportation in Houston, Texas’s largest city, we seriously doubt their accessibility in rural areas — which, presumably, are less likely to have public transportation infrastructure. Unfortunately, we are left to wonder. Texas, which bears the burden of proof, submitted absolutely nothing to counter this testimony or to show whether its DPS offices are reachable via public transportation.

None of the burdens associated with obtaining an EIC has ever before been imposed on Texas voters. Based on the record evidence before us, it is virtually certain that these burdens will disproportionately affect racial minorities. Simply put, many Hispanics and African Americans who voted in the last election will, because of the burdens imposed by SB 14, likely be unable to vote in the next election. This is retrogression. See Bossier Parish II, 528 U.S. at 324, 120 S.Ct. 866.

Significantly, Texas disputes none of the facts underlying this conclusion — not the $22 cost for a birth certificate, not the distance between DPS offices, not the poverty rates for minorities in Texas, not the disproportionate vehicle access rates. Instead, in a hodgepodge of arguments, Texas seeks to downplay SB 14’s impact, contending, in essence, that the law’s retrogressive effect will not be particularly severe. In addition, Texas insists that courts may not legally consider nonracial factors like poverty when determining whether a law warrants section 5 preclearance. We address each argument in turn.

In support of its assertion that the burdens imposed by SB 14 are less onerous than they may seem, Texas contends that because disabled voters and those over the age of 65 will be able to vote absentee, they are “[un]affected by SB 14’s photo ID requirement.” Texas Proposed Findings, ECF No. 202 at 18. Texas also points out that the $22 fee for a birth certificate will only affect voters who lack the underlying documentation needed for an EIC. And, the state assures us, Texans will not really mind traveling long distances to obtain an EIC. For “the people who choose to live in that part of Texas,” Texas’s counsel stated during closing arguments, “it’s just a reality of life that they have to drive long distances.” Trial Tr. 7/13/2012 52:16-18.

These arguments lack merit. To begin with, voters eligible for absentee ballots will not be “exempt” from SB 14. See supra at part III.B.1. Some voters over age 65 will undoubtedly prefer to cast their ballots at the polls — perhaps out of habit, a sense of civic pride, or simply because they wish to follow the news all the way up to Election Day before selecting a candidate. Reverend Peter Johnson, an African American clergyman and a leader in Texas’s civil rights community testified:

I have a group of African-American senior citizen women, they want to go to the voting polls and stand in line and vote at the voting polls. There’s a certain degree of dignity for them to do this.... Because these people appreciate this sacred right to vote, and they’re not going to vote absentee.

Trial Tr. 7/11/2012 (PM) 6:1-9. If these “senior citizen women” — or any other voters over 65 — wish to cast an in-person ballot, they will not be permitted to do so unless they can produce an SB 14-qualify-ing photo ID.

In any event, even if, as Texas asserts, this so-called “allowance” for elderly voters, together with SB 14’s special procedures for disabled voters, were to reduce the overall number of people affected by the law, we have no reason to believe that they would mitigate SB 14’s disproportionate effect on racial minorities. To the contrary, at least with respect to voters over 65, record evidence indicates that any “benefit” will disproportionately accrue to white Texans. Undisputed census data shows that 19.4% of the white voting age population in Texas is over 65, compared to just 10.6% of African Americans and 8.7% of Hispanics. DefendanNIntervenors’ Ex. 118. This suggests that far more white voters than minorities will be eligible to cast absentee ballots. Texas offers nothing to counter this evidence, much less provides any reason to believe that SB 14’s absentee provisions will mitigate its retrogressive effect on minorities.

Texas’s failure to support its factual assertions dooms its other claims as well. If the state believes that only a few Texans lack certified copies of their birth certificate, it should have produced evidence to that effect. The same is true with respect to its claim that Texans are preternaturally unperturbed by the prospect of traveling 200 to 250 miles. And for either of these propositions to have been relevant to the issue before us, Texas would have to have provided evidence that they mitigate SB 14’s retrogressive effect on minority voters. Because Texas has offered no record evidence for either the truth of its contentions or their effect on racial minorities, we give no credence to this line of argument.

This brings us to Texas’s second — and primary — argument. Relying on the literal language of section 5, which prohibits states from “denying or abridging the right to vote on account of race or color,” or “because [a voter] is a member of a language minority group,” 42 U.S.C. §§ 1973c(a), 1973b(f)(2) (emphasis added), Texas argues:

The “effects” prong of section 5 does not extend to laws that merely have a disparate impact on the races. It allows courts to deny preclearance only if the effect of SB 14 is to deny or abridge the right to vote “on account of” race or color, or “because of” one’s membership in a language minority group.

Texas Proposed Findings, EOF No. 202 at 44. According to Texas, if SB 14 denies or abridges the right to vote at all, it does so “on account of’ factors like poverty or lack of vehicular access. To be sure, these factors may correlate with racial minority status, but because disenfranchisement is proximately caused by something other than race, Texas contends, this court may not deny preclearance under section 5’s effect element. Indeed, Texas believes that its reading of section 5 is constitutionally compelled because “the Constitution does not allow ... federal officials to deny preclearance to State laws that do not violate the Fifteenth Amendment” — i.e., that do not directly discriminate on the basis of race. Texas Proposed Findings, ECF No. 202 at 46. “At the very least,” Texas concludes, “Northwest Austin compels courts to construe section 5 to avoid this grave constitutional question absent specific and unambiguous statutory language to the contrary.” Id.

For several reasons, we find this argument entirely unpersuasive. To begin with, as explained above, Congress passed the Voting Rights Act precisely to prohibit election devices proximately based on something other than race — “notorious devices” such as “poll taxes, literacy tests, grandfather clauses, and property qualifications.” Shelby Cnty., 679 F.3d at 853; see also Allen, 393 U.S. at 565, 89 S.Ct. 817 (holding that the purpose of the Voting Rights Act was to eliminate “the subtle, as well as the obvious, state regulations which have the effect of denying citizens their right to vote because of their race”). In fact, the very point of such devices was that they were supposedly “race neutral,” thus giving states an end-run around the Fifteenth Amendment’s prohibition on racial discrimination in voting. Yet under Texas’s interpretation, section 5’s effect element could not have reached any of these laws. As Texas’s counsel stated, “certainly I’m not promoting a literacy test, but if the evidence was that it didn’t deny or abridge the right to vote on account of color and there was no racially discriminatory purpose ... you could have no violation under [section 5’s] standards.” Trial Tr. 7/13/2012 30:15-20. We cannot accept an interpretation of section 5 that would so severely constrain courts’ ability to block precisely the type of “evil” that the Voting Rights Act was meant to address. Katzenbach, 383 U.S. at 328, 86 S.Ct. 803.

Moreover, Texas’s reading of section 5 collapses its effect element into its purpose element. After all, any law that would deny or abridge the right to vote directly on account of race — e.g., a law that disenfranchises African Americans because they are African Americans — would have been enacted with a discriminatory purpose. And, to be clear, this is the only type of law Texas thinks section 5’s effect element reaches, thus rendering the purpose and effect elements coterminous. But section 5 quite clearly has two separate and distinct requirements: covered states must prove that a change in voting procedures “neither has the purpose nor will have the effect of denying or abridging the right to vote on account of race.” 42 U.S.C. § 1973c(a) (emphasis added). As the Supreme Court has repeatedly explained, “[w]e must have regard to all the words used by Congress, and as far as possible give effect to them.” United States v. Atl. Research Corp., 551 U.S. 128, 137, 127 S.Ct. 2331, 168 L.Ed.2d 28 (2007) (quoting Louisville & Nashville R. Co. v. Mottley, 219 U.S. 467, 475, 31 S.Ct. 265, 55 L.Ed. 297 (1911)). Interpreting “purpose” and “effect” as synonymous would run afoul of this principle. The Supreme Court has said as much, emphasizing that “Congress plainly intended that a voting practice not be precleared unless both discriminatory purpose and effect are absent.” City of Rome v. United States, 446 U.S. 156, 172, 100 S.Ct. 1548, 64 L.Ed.2d 119 (1980) (first emphasis added).

Finally — and unsurprisingly given both the purpose of the Voting Rights Act and section 5’s plain text — Texas cites no authority for its novel reading of the statute. For decades, courts have applied the Supreme Court’s longstanding interpretation of section 5’s effect element, requiring covered states to prove that none of their “voting-procedure changes ... would lead to a retrogression in the position of racial minorities with respect to their effective exercise of the electoral franchise.” Beer, 425 U.S. at 141, 96 S.Ct. 1357. Never has a court excused “retrogression in the position of racial minorities” because that retrogression was proximately caused by something other than race.

As we have indicated throughout this opinion, we are sensitive to the concerns raised in Northwest Austin. See supra Part I at 9-11, Part II at 19-23. But to hold, as Texas urges, that section 5 applies only to voting changes that themselves violate the Fifteenth Amendment would require us to ignore section 5’s purpose and structure, as well as decades of Supreme Court decisions interpreting its language. We see nothing in Northwest Austin that would permit, much less require, this “inferior” court to undertake such a dramatic departure from well-established law.

IV.

To sum everything up: section 5 prohibits covered states from implementing voting laws that will have a retrogressive effect on racial minorities. See Beer, 425 U.S. at 141, 96 S.Ct. 1357. Texas, seeking to implement its voter ID law, bears the burden of proof and must therefore show that SB 14 lacks retrogressive effect. Georgia, 411 U.S. at 538, 93 S.Ct. 1702. But as we have found, everything Texas has submitted as affirmative evidence is unpersuasive, invalid, or both. Moreover, uneontested record evidence conclusively shows that the implicit costs of obtaining SB 14-qualifying ID will fall most heavily on the poor and that a disproportionately high percentage of African Americans and Hispanics in Texas live in poverty. We therefore conclude that SB 14 is likely to lead to “retrogression in the position of racial minorities with respect to their effective exercise of the electoral franchise.” Beer, 425 U.S. at 141, 96 S.Ct. 1357. Given this, and given that Texas must show that SB 14 lacks both discriminatory purpose and effect, we have no need to examine whether the law was enacted with discriminatory purpose. Accordingly, we shall deny Texas’s request for declaratory relief.

In reaching This conclusion, we emphasize the narrowness of this opinion. Specifically, we have decided nothing more than that, in this particular litigation and on this particular record, Texas has failed to demonstrate that its particular voter ID law lacks retrogressive effect. Nothing in this opinion remotely suggests that section 5 bars all covered jurisdictions from implementing photo ID laws. To the contrary, under our reasoning today, such laws might well be precleared if they ensure (1) that all prospective voters can easily obtain free photo ID, and (2) that any underlying documents required to obtain that ID are truly free of charge. Indeed, Georgia’s voter ID law was precleared by the Attorney General — and probably for good reason. Unlike SB 14, the Georgia law requires each county to provide free election IDs, and further allows voters to present a wide range of documents to obtain those IDs. Ga. Code Ann. § 21-2-417.1(a); Ga. Elec. Code 183-1-20-.01. The contrast with Senate Bill 14 could hardly be more stark.

Finally, during closing arguments, Texas’s counsel complained that they had been shouldered with an “impossible burden” in this litigation. Trial Tr. 7/13/2012 27:14. This may well be correct, but Texas’s lawyers have only their client to blame. The State of Texas enacted a voter ID law that — at least to our knowledge — is the most stringent in the country. That law will almost certainly have retrogressive effect: it imposes strict, unforgiving burdens on the poor, and racial minorities in Texas are disproportionately likely to live in poverty. And crucially, the Texas legislature defeated several amendments that could have made this a far closer case. Ignoring warnings that SB 14, as written, would disenfranchise minorities and the poor, see, e.g., JA 1300-03; 1329, the legislature tabled or defeated amendments that would have:

• waived all fees for indigent persons who needed the underlying documents to obtain an EIC, Trial Tr. 7/12/2012 (AM) 30:17-31:7, 33:23-24;

• reimbursed impoverished Texans for EIC-related travel costs, JA 2139^12;

• expanded the range of identifications acceptable under SB 14 by allowing voters to present student or Medicare ID cards at the polls, Trial Tr. 7/12/2012 (AM) 34:21-24; JA 1246^7;

• required DPS offices to remain open in the evening and on weekends, JA 1337; and

• allowed indigent persons to cast provisional ballots without photo ID. Trial Tr. 7/12/2012 (AM) 35:3-37:1.

Put another way, if counsel faced an “impossible burden,” it was because of the law Texas enacted — nothing more, nothing less.

y.

For the foregoing reasons, we deny Texas’s request for a declaratory judgment. The parties are hereby ordered to meet and confer as to a schedule to govern the constitutional issue and to file an advisory within 14 days with a proposed schedule. A separate order has been filed on this date.