Fish v. Schwab

FILED United States Court of Appeals Tenth Circuit PUBLISH April 29, 2020 Christopher M. Wolpert UNITED STATES COURT OF APPEALS Clerk of Court TENTH CIRCUIT STEVEN WAYNE FISH, on behalf of himself and all others similarly situated; DONNA BUCCI, on behalf of herself and all others similarly situated; CHARLES STRICKER, on behalf of himself and all others similarly situated; THOMAS J. BOYNTON, on behalf of himself and all others similarly situated; DOUGLAS HUTCHINSON, on behalf of himself and all others similarly situated; LEAGUE OF WOMEN VOTERS OF KANSAS, Plaintiffs - Appellees, v. No. 18-3133 SCOTT SCHWAB, in his official capacity as Secretary of State for the State of Kansas, Defendant - Appellant. ___________________________ STATE OF TEXAS; STATE OF ARKANSAS; STATE OF OKLAHOMA; STATE OF WEST VIRGINIA; PAUL LEPAGE, † Governor of Maine; STATE OF MISSOURI; EAGLE FORUM EDUCATION & LEGAL DEFENSE FUND, Amici Curiae. ____________________________ CODY KEENER; ALDER CROMWELL, Plaintiffs, and PARKER BEDNASEK, Plaintiff - Appellee, v. No. 18-3134 SCOTT SCHWAB, Kansas Secretary of State, Defendant - Appellant. -------------------------------------- † Paul LePage is no longer Governor of Maine. In a letter dated March 15, 2019, counsel for Maine’s Attorney General, informed the court that Maine’s current Governor, Janet T. Mills, “does not support the position taken [by several states and then-Governor LePage] in the amicus brief filed in this case on October 5, 2018,” and that “the position taken in the amicus [brief] does not reflect the position of the State of Maine.” Letter of Susan P. Herman, Deputy Attorney General to Tenth Circuit Clerk, Case Nos. 18-3133 & 18-3134, at 1 (Mar. 15, 2019). Although we acknowledge this correspondence, it has no impact on our resolution of this appeal. 2 STATE OF TEXAS; STATE OF ARKANSAS; STATE OF OKLAHOMA; STATE OF WEST VIRGINIA; PAUL LEPAGE, Governor of Maine; STATE OF MISSOURI; EAGLE FORUM EDUCATION & LEGAL DEFENSE FUND, Amici Curiae. Appeals from the United States District Court for the District of Kansas (D.C. No. 2:16-CV-02105-JAR) (D.C. No. 2:15-CV-09300-JAR) Toby Crouse, Solicitor General of Kansas, (Derek Schmidt, Attorney General of Kansas, Jeffrey A. Chanay, Chief Deputy Attorney General, Bryan C. Clark, Assistant Solicitor General, Dwight R. Carswell, Assistant Solicitor General, with him on the briefs), Office of Attorney General, Topeka, Kansas, for Defendant- Appellant. Dale Ho, American Civil Liberties Union Foundation, New York, New York, (R. Orion Danjuma and Sophia Lin Lakin, American Civil Liberties Union Foundation, New York, New York; Mark P. Johnson, Curtis E. Woods and Samantha Wenger, Dentons US LLP, Kansas City, Missouri; Neil A. Steiner and Rebecca Kahan Waldman, Dechert LLP, New York, New York; Angela M. Liu, Dechert LLP, Chicago, Illinois; Lauren Bonds and Zal K. Shroff, ACLU Foundation of Kansas, Overland Park, Kansas; Lino S. Lipinsky De Orlov, Dentons US LLP, Denver, Colorado; Mark T. Emert, Fagan, Emert & Davis LLC, Lawrence, Kansas; Shannon Wells Stevenson, Davis Graham & Stubbs LLP, Denver, Colorado, with him on the brief), for Plaintiffs-Appellees. Ken Paxton, Attorney General of Texas, Jeffrey C. Mateer, First Assistant Attorney General, Kyle D. Hawkins, Solicitor General, Matthew H. Frederick, Deputy Solicitor General, Beth Klusmann, Assistant Solicitor General, filed an amicus curiae brief for the States of Texas, Arkansas, Missouri, Oklahoma, West 3 Virginia and Paul R. LePage, Governor of Maine, in support of Defendant- Appellant. Lawrence J. Joseph, Washington, D.C., filed an amicus curiae brief on behalf of Eagle Forum Education & Legal Defense Fund, in support of Defendant- Appellant. Before BRISCOE, McKAY, * and HOLMES, Circuit Judges. HOLMES, Circuit Judge. In these two consolidated appeals, we must determine whether a Kansas law requiring documentary proof of citizenship (“DPOC”) for voter registration is preempted by section 5 of the National Voter Registration Act (“NVRA”), 52 U.S.C. § 20504, or violates the Fourteenth Amendment’s Equal Protection Clause. We addressed the first of these questions, i.e., whether Kansas’s DPOC requirement is preempted by section 5 of the NVRA, in Fish v. Kobach (“Fish I”), 840 F.3d 710 (10th Cir. 2016). Section 5 of the NVRA mandates that a voter- * The late Honorable Monroe G. McKay was a member of the three- judge panel assigned to this case and heard the parties’ oral arguments, but he passed away on March 28, 2020. He took no part in the final disposition of this case, including the preparation of this opinion. “The practice of this court permits the remaining two panel judges if in agreement to act as a quorum in resolving the appeal.” United States v. Wiles, 106 F.3d 1516, 1516 n.* (10th Cir. 1997); see also 28 U.S.C. § 46(d) (providing that “[a] majority of the number of judges authorized to constitute a court or panel thereof . . . shall constitute a quorum”). The remaining two panel members have acted as a quorum with respect to this appeal, and, for the reasons stated herein, have voted to affirm the decision of the district court. 4 registration form must be a part of any application to obtain or renew a driver’s license. Such a registration form “‘may require only the minimum amount of information necessary to’ prevent duplicate registrations and to ‘enable State election officials to assess the eligibility of the applicant and to administer voter registration and other parts of the election process.’” Id. at 715–16 (footnotes omitted) (quoting 52 U.S.C. § 20504(c)(2)(B)). In Fish I, we held that—on the factual record then before this court—“the DPOC required by Kansas law [was] more than the minimum amount of information necessary [to perform the Kansas Secretary of State’s eligibility-assessment and registration duties] and, therefore, [was] preempted by the NVRA.” Id. at 717. Thus, we held that “the district court did not abuse its discretion in granting [a] preliminary injunction [against the enforcement of the DPOC law] because the NVRA preempts Kansas’s DPOC law as enforced against those applying to vote while obtaining or renewing a driver’s license.” Id. at 716. We remanded for a trial on the merits where Kansas’s Secretary of State would have an opportunity to demonstrate that the DPOC requirement was not more than the minimum amount of information necessary to perform his eligibility-assessment and registration duties. On remand, the district court consolidated that statutory challenge with a related case that raises the second aspect of this appeal, i.e., whether the DPOC requirement violates the Fourteenth Amendment’s Equal Protection Clause. The 5 Supreme Court and this court have evaluated challenges to state-voter- identification requirements under the Equal Protection Clause. See Crawford v. Marion Cty. Election Bd., 553 U.S. 181, 189S91 (2008) (plurality opinion of Stevens, J.); ACLU of N.M. v. Santillanes, 546 F.3d 1313, 1320 (10th Cir. 2008). Proceeding under that framework, the Equal Protection Clause challenge to the DPOC requirement is predicated on the idea that the DPOC requirement unconstitutionally burdens the right to vote because the interests asserted by the Kansas Secretary of State (“the Secretary”) are insufficient to justify the burden it imposes on that right. After holding a joint bench trial, the district court entered a permanent injunction against the enforcement of the DPOC requirement under both section 5 of the NVRA and the Equal Protection Clause. The Secretary has appealed. His appeal raises the two fundamental questions outlined above. First, in Bednasek v. Schwab, No. 18-3134, does the DPOC requirement violate the Equal Protection Clause? Second, in Fish v. Schwab, No. 18-3133, does section 5 of the NVRA preempt the DPOC requirement? Exercising jurisdiction under 28 U.S.C. § 1291, we answer both questions in the affirmative and thus affirm the district court’s judgment enjoining enforcement of the DPOC requirement. In doing so, we summarize the relevant background, assure ourselves that the challengers possess 6 standing, and then discuss both challenges to the DPOC requirement, taking up first (for organizational convenience) the constitutional challenge. I. Background A. Kansas’s DPOC Requirement Both suits on appeal challenge Kansas’s DPOC requirement, and so we start by recounting Fish I’s summary of the statute and regulations that constitute Kansas’s DPOC requirement: Kansas adopted its DPOC requirement for voter registration on April 18, 2011. Secure and Fair Elections (“SAFE”) Act, ch. 56, § 8(l), 2011 Kan. Sess. Laws 795, 806, 809S11 (codified at Kan. Stat. Ann. § 25S2309(l)). The requirement took effect January 1, 2013. Id. at § 8(u), 2011 Kan. Sess. Laws at 812. The SAFE Act requires that (l) The county election officer or secretary of state’s office shall accept any completed application for registration, but an applicant shall not be registered until the applicant has provided satisfactory evidence of United States citizenship. Evidence of United States citizenship as required in this section will be satisfied by presenting one of the documents listed . . . in person at the time of filing the application for registration or by including a photocopy of one of the following documents with a mailed registration application. After a person has submitted satisfactory evidence of citizenship, the county election officer shall indicate this information in the person’s permanent voter file. Kan. Stat. Ann. § 25S2309(l). The statute then lists thirteen forms of documentation acceptable to prove U.S. citizenship, including a birth certificate or passport. See § 25S2309(l)(1)S(13). For citizens unable to present 7 DPOC, subsection (m) provides an alternate means to prove citizenship by the submission of evidence to the state election board followed by a hearing. See § 25S2309(m). The state election board is composed of “the lieutenant governor, the secretary of state and the attorney general.” § 25S2203(a). [Then-serving Kansas] Secretary [of State Kris W.] Kobach promulgated regulations for the DPOC requirement on October 2, 2015. Kan. Admin. Regs. § 7S23S15 (the “90-day regulation”). Those regulations provide that applications unaccompanied by DPOC are deemed to be “incomplete.” § 7S23S15(a). Once an application is designated as incomplete, a voter has ninety days to provide DPOC or else the application is canceled and a new voter-registration application is required to register. See § 7S23S15(b)S(c). 840 F.3d at 717. B. Factual Background 1. Bednasek v. Schwab, No. 18-3134 Mr. Parker Bednasek—the only remaining plaintiff in Bednasek v. Schwab, No. 18-3134—moved from Texas to Kansas in order to attend the University of Kansas. While he was a full-time student at the University of Kansas, he canceled his Texas voter registration and applied to register to vote in Kansas. He did so because he “considered [him]self to be a resident in Kansas, and [he] wanted to vote in Kansas elections.” Aplt.’s App., Vol. 38, at 9339 (Tr. of Bench Trial, Day 2, P.M. Session, filed Mar. 30, 2018). In applying, he swore that he was a Kansas resident and that he had abandoned his former residence. He later 8 swore that he had “no intent to leave Kansas in the future.” Id., Vol. 48, at 11692 (Aff. of Parker Bednasek, filed Apr. 21, 2016). While at the University of Kansas, Mr. Bednasek paid out-of-state tuition, had a vehicle that he jointly owned with his parents that was registered in Texas, had a car insurance policy on that vehicle registered to his parents’ Texas home, and applied for and received a Texas driver’s license. When he submitted his application to register to vote, Mr. Bednasek did not provide DPOC. He did not do so because (1) his birth certificate was at his parent’s home in Texas, and (2) “he [did] not agree with the law” and was attempting to challenge it. Id., Vol. 47, at 11466 (Findings of Fact & Conclusions of Law, filed June 18, 2018); see id., Vol. 38, at 9368S69. Mr. Bednasek would later acquire a copy of his DPOC in order to apply to the Navy but did not submit it to Kansas. Because he never submitted DPOC to Kansas, his application was canceled under the DPOC requirement. 2. Fish v. Schwab, No. 18-3133 In Fish v. Schwab, No. 18-3133, multiple plaintiffs attempted to register to vote as “motor voters” under section 5 of the NVRA, but their applications to register were denied because of the DPOC requirement. We briefly recount some of their experiences. Mr. Steven Fish applied for a driver’s license and to register as a voter. The driver’s license examiner did not inform him that he needed to 9 provide DPOC, but he subsequently received notices informing him that he needed to submit DPOC. However, he had difficulty locating his birth certificate because “he was born on a decommissioned Air Force base.” Id., Vol. 47, at 11460. He thus was unable to vote in the 2014 general election. A family member subsequently found his birth certificate, and he has now registered. Similarly, Ms. Donna Bucci applied to vote while renewing her driver’s license. She was not told that she needed to provide DPOC, but, like Mr. Fish, she later received a notice informing her that she needed to provide DPOC in order to register. However, Ms. Bucci did not possess a copy of her birth certificate. The district court found that “[s]he [could not] afford the cost of a replacement birth certificate from Maryland and she credibly testified that spending money to obtain one would impact whether she could pay rent.” Id. at 11461. Her application was canceled for failure to provide DPOC, and she was unable to vote in the 2014 election. A third plaintiff, Mr. Douglas Hutchinson, likewise applied to register to vote while renewing his driver’s license but did not provide DPOC. His application was also later canceled. Other plaintiffs did bring DPOC to register, but various errors in the administration of the DPOC requirement prevented their registration. Mr. Charles Stricker applied to vote while renewing his driver’s license and brought DPOC with him, but the clerk told him that he did not need to provide anything. He only 10 learned that his application had been canceled for lack of DPOC when he was not allowed to vote at the polls. Similarly, Mr. Thomas Boynton applied to vote, and his application was suspended for failure to provide DPOC. He had, however, brought DPOC with him to register and provided the requested documentation. Nevertheless, he was told that he was not registered when he showed up at the polls. After the election, he received a notice that he needed to resubmit DPOC in order to complete the voter-registration process. The final plaintiff is the League of Women Voters of Kansas (“Kansas League”), “a nonpartisan, nonprofit volunteer organization that encourages informed and active participation of citizens in government.” Id. at 11452. The district court found that “the DPOC requirement significantly hampered the Kansas League’s voter registration work,” id. at 11453, “the DPOC requirement forced the Kansas League to devote substantial resources to assist voters whose applications are in suspense due to the failure to provide DPOC,” id. at 11455, and “the DPOC requirement has forced the Kansas League to spend a considerable amount of member resources—including volunteer time—and money to educate the public about registering under the DPOC law,” id. 11 C. Procedural Background 1. Fish I The Fish plaintiffs brought suit seeking a preliminary injunction against the enforcement of the DPOC requirement. The district court granted the preliminary injunction and “required [the Secretary] to register to vote any applicants previously unable to produce DPOC and to cease enforcement of Kansas’s DPOC requirement with respect to individuals who apply to register to vote at the Kansas Department of Motor Vehicles (‘DMV’) through the motor voter process.” Fish I, 840 F.3d at 716. The Secretary appealed the entry of the preliminary injunction, and we affirmed. Id. at 716–17. As recounted in more depth below, we held that section 5 of the NVRA preempted Kansas’s DPOC requirement. Id. at 716. The relevant portion of section 5 of the NVRA, known as the “motor-voter” provision, states: (2) The voter registration application portion of an application for a State motor vehicle driver’s license— (A) may not require any information that duplicates information required in the driver’s license portion of the form (other than a second signature or other information necessary under subparagraph (C)); (B) may require only the minimum amount of information necessary to— (i) prevent duplicate voter registrations; and (ii) enable State election officials to assess the eligibility of the applicant and to administer voter registration and other parts of the election process; 12 (C) shall include a statement that— (i) states each eligibility requirement (including citizenship); (ii) contains an attestation that the applicant meets each such requirement; and (iii) requires the signature of the applicant, under penalty of perjury . . . . 52 U.S.C. § 20504(c)(2)(A)S(C) (emphases added). In Fish I, we read subparagraph (C)’s “attestation requirement” as establishing “the presumptive minimum amount of information necessary for a state to carry out its eligibility-assessment and registration duties [under subparagraph (B)].” 840 F.3d at 737. However, we acknowledged that “whether the attestation requirement actually satisfies the minimum-information principle in a given case turns on the factual question of whether the attestation requirement is sufficient for a state to carry out these duties.” Id. at 738. We held that “in order for a state advocating for a DPOC regime to rebut the presumption that the attestation requirement is the minimum information necessary for it to carry out its eligibility-assessment and registration duties,” section 5 of the NVRA requires “[the] state to show that ‘a substantial number of noncitizens have successfully registered’ notwithstanding the attestation requirement.” Id. at 738–39 (quoting Kobach v. U.S. Election Assistance Comm’n (“EAC”), 772 F.3d 1183, 1198 (10th Cir. 2014)). We held that the Secretary had failed to demonstrate that a substantial number of noncitizens had successfully registered. Id. at 746–47. In particular, 13 the Secretary had only shown that between 2003 and 2013 “thirty noncitizens registered to vote.” Id. at 746. “These numbers [fell] well short of the showing necessary to rebut the presumption that attestation constitutes the minimum amount of information necessary for Kansas to carry out its eligibility-assessment and registration duties.” Id. at 747. Thus, we concluded that the plaintiffs’ challenge was likely to succeed on the merits. Id. at 750. We went on to address the remaining preliminary-injunction factors and concluded that the district court did not err in concluding that they, too, favored a preliminary injunction. Id. at 751S56. We thus affirmed the district court’s grant of the preliminary injunction. Id. at 756. 2. The Joint Fish and Bednasek Bench Trial On remand, the district court consolidated Fish—the statutory case—with Bednasek—the constitutional case—for trial. We summarize the district court’s factual findings, its legal conclusions in both cases, and the remedies it imposed. a. Factual Findings In addition to the above evidence about individual plaintiffs with suspended or canceled applications, the plaintiffs put forward statistical evidence about the overall number of suspended applications. The district court found that, before the preliminary injunction in Fish was issued, there were 14,770 individuals who had applied to vote but whose applications were suspended for failure to provide 14 DPOC. Of these, 5,655 were motor-voter applicants. Another 16,319 individuals had their applications canceled for failure to provide DPOC. Of these, 11,147 were motor-voter applicants. That “amount[ed] to 31,089 total applicants who were denied registration for failure to provide DPOC.” Aplt.’s App., Vol. 47, at 11447. Other numbers and expert testimony demonstrated that “[c]anceled or suspended applicants represented 12.4% of new voter registrations between January 1, 2013[,] and December 11, 2015.” Id. at 11448. An expert opined that the total number of applicants with suspended or canceled applications would have increased but for the injunction, “in part because voter registration activity typically increases in the months leading up to a presidential election.” Id. at 11449. However, while there was significant evidence that would-be voters had their applications suspended and canceled by the DPOC requirement, the court acknowledged that “[t]here was little admissible evidence presented at trial about the rate of DPOC possession by suspended and canceled applicants.” Id. at 11457. The district court also made factual findings about the number of noncitizens who had applied to register to vote. The district court found that, “at most, 67 noncitizens registered or attempted to register in Kansas over the last 19 years.” Id. at 11519. Of these, “there [were] only 39 confirmed noncitizens who successfully registered to vote between 1999 and 2013 when the DPOC law 15 became effective.” Id. at 11508. Those 39 individuals represented 0.002% of all registered voters in Kansas as of January 1, 2013—the date the DPOC requirement went into effect. Id. And specifically as to those applications that were suspended, the court found that the estimated number of suspended applications that belonged to noncitizens was “statistically indistinguishable from zero,” while “more than 99% of the individuals” whose voter-registration applications were suspended were citizens who would have been able to vote but for the DPOC requirement. Id. at 11491S92; see id. at 11481. Moreover, even those few instances of noncitizens attempting to register to vote may be explained by “administrative anomalies.” Id. at 11520. For example, Kansas’s voter- registration database included 100 individuals with purported birth dates in the 19th century and 400 individuals with purported birth dates after their date of voter registration. While the district court only found these 39 instances of noncitizen registration, the Secretary presented several anecdotes regarding noncitizens who purportedly attempted to register to vote. There was evidence that some noncitizens with temporary driver’s licenses had been registered to vote and that one citizen told Kansas officials that she had voted before becoming a citizen. The Secretary also identified an incident, summarized in a letter in the record, where employees of a hog farm “were transported to [a county] office by their 16 employer to register to vote” even though a county clerk stated that “some of these employees felt they were pressured to register even though they may not be legal.” Id., Vol. 30, at 7668 (Letter from Seward Cty. Clerk to Kan. S. Ethics & Elections Comm., dated Mar. 3, 2011). In its preliminary injunction analysis, the district court had concluded that the evidence submitted about this incident was “insufficient to show that noncitizens actually voted.” Id., Vol. 4, at 867 (Mem. & Order, filed May 17, 2016). While the Secretary returns to these examples in this court, it is unclear how they relate to the 39 instances of noncitizen registration that the district court identified. Apart from the 39 confirmed instances of noncitizen registration, Kansas presented expert testimony on statistical estimates of noncitizen registration and on noncitizen registration outside Kansas. But the district court either gave little weight or entirely rejected Kansas’s expert testimony on these topics. The court found that one expert’s testimony contained “myriad misleading statements” and “preordained opinions.” Id., Vol. 47, at 11474; see id. at 11476 (“The record is replete with further evidence of [the expert]’s bias.”). The court also concluded that studies offered by a different expert were “confusing, inconsistent, and methodologically flawed.” Id. at 11491. “[L]ooking beyond Kansas, [the Secretary’s] evidence of noncitizen registration at trial was weak.” Id. at 11519. The district court explained at length why it excluded large portions of the 17 Secretary’s expert testimony on statistical estimates of noncitizen registration in Kansas and found much of the remaining testimony unpersuasive. Id. (explaining that one of the Secretary’s experts was “credibly dismantled” by the architect of the survey upon which the expert had relied). b. Legal Conclusions in Bednasek v. Schwab, No. 18-3134 In Bednasek, the court carefully evaluated these facts under the equal- protection rubric established by the Supreme Court’s opinion in Crawford v. Marion County Election Board, supra. As discussed at length below, Crawford instructs that we are to examine the burden that a state law places on the right to vote and then weigh the government’s asserted interests for imposing that law against that burden. Guided by Crawford, the district court here balanced the burdens imposed by the DPOC requirement against the state’s interests in preventing noncitizen voter registration, maintaining accurate voter rolls, and maintaining confidence in elections. The court concluded that “the magnitude of potentially disenfranchised voters impacted by the DPOC law and its enforcement scheme cannot be justified by the scant evidence of noncitizen voter fraud before and after the law was passed, by the need to ensure the voter rolls are accurate, or by the State’s interest in promoting public confidence in elections.” Aplt.’s App., Vol. 47, at 11526. 18 c. Legal Conclusions in Fish v. Schwab, No. 18-3133 In Fish, the court held that the preemption framework established in Fish I was the law of the case. Applying that framework and considering the evidence summarized above, the court concluded that there was “no credible evidence that a substantial number of noncitizens registered to vote under the attestation regime.” Id. at 11507. Instead, it found that the “evidence of a small number of noncitizen registrations in Kansas . . . is largely explained by administrative error, confusion, [and] mistake.” Id. at 11509. The court concluded that “[the Secretary] ha[d] failed to rebut the presumption that the attestation clause meets the minimum information principle in § 5 of the NVRA, and [it] therefore order[ed] judgment in favor of [the Fish] Plaintiffs.” Id. at 11515. d. Remedies Finding that the DPOC requirement violated the Equal Protection Clause and section 5 of the NVRA, the court ordered the Secretary “not [to] enforce the DPOC law and accompanying regulation against voter registration applicants in Kansas.” Id. at 11529. The court also ordered various specific forms of relief not at issue here, e.g., ordering the Secretary to update websites and provide applicants with certificates of registration. Id. at 11530. The Secretary timely appealed. 19 II. Standing The Secretary first argues that Mr. Bednasek—the only remaining plaintiff in Bednasek v. Schwab, No. 18-3134—lacks standing. 1 We summarize the relevant legal principles and conclude Mr. Bednasek has standing. A. Legal Principles Governing Standing Article III of the United States Constitution restricts the jurisdiction of federal courts to the adjudication of “Cases” or “Controversies.” U.S. C ONST . art. III, § 2, cl. 1. To satisfy Article III’s case-or-controversy requirement, a plaintiff must demonstrate standing by establishing “(1) an ‘injury-in-fact,’ (2) a sufficient ‘causal connection between the injury and the conduct complained of,’ and (3) a ‘likel[ihood]’ that the injury ‘will be redressed by a favorable decision.’” Susan B. Anthony List v. Driehaus, 573 U.S. 149, 157S58 (2014) (alteration in original) (quoting Lujan v. Defs. of Wildlife, 504 U.S. 555, 560S61 (1992)); accord Spokeo, Inc. v. Robins, --- U.S. ----, 136 S. Ct. 1540, 1547 (2016). “Put simply, a plaintiff must establish three elements: an injury-in-fact, causation, and redressability.” Bronson v. Swensen, 500 F.3d 1099, 1106 (10th Cir. 2007). 1 The Secretary does not challenge the Fish plaintiffs’ standing. Nevertheless, “[w]e have an obligation to assure ourselves of litigants’ standing under Article III.” Frank v. Gaos, --- U.S. ----, 139 S. Ct. 1041, 1046 (2019) (per curiam) (quoting DaimlerChrysler Corp. v. Cuno, 547 U.S. 332, 340 (2006)). We concluded that the Fish plaintiffs had standing in Fish I, see 840 F.3d at 716 n.5 (“We are confident on the current record that Plaintiffs–Appellees have standing to sue.”), and we see no reason to depart from that conclusion now. 20 We are focused on the first two of these requirements. As to the first, “[t]o establish injury in fact, a plaintiff must show that he or she suffered ‘an invasion of a legally protected interest’ that is ‘concrete and particularized’ and ‘actual or imminent, not conjectural or hypothetical.’” Spokeo, 136 S. Ct. at 1548 (quoting Lujan, 504 U.S. at 560); accord People for the Ethical Treatment of Prop. Owners v. U.S. Fish & Wildlife Serv., 852 F.3d 990, 996–97 (10th Cir. 2017). And as to the second, causation is established when the injury “is fairly traceable to the challenged conduct of the defendant.” Spokeo, 136 S. Ct. at 1547 (citing Lujan, 504 U.S. at 560S61); accord Benham v. Ozark Materials River Rock, LLC, 885 F.3d 1267, 1273 (10th Cir. 2018) (“To satisfy the traceability requirement, the defendant’s conduct must have caused the injury.”). This requires proving that there is “a substantial likelihood that the defendant’s conduct caused plaintiff’s injury in fact.” Bronson, 500 F.3d at 1109–10 (quoting Nova Health Sys. v. Gandy, 416 F.3d 1149, 1156 (10th Cir. 2005)). “We . . . review the district court’s rulings on standing de novo.” Niemi v. Lasshofer, 770 F.3d 1331, 1344 (10th Cir. 2014); accord People for the Ethical Treatment of Prop. Owners, 852 F.3d at 996. However, we review the factual findings underlying the district court’s standing determination for clear error. See Protocols, LLC v. Leavitt, 549 F.3d 1294, 1298 (10th Cir. 2008); see also McCormack v. Herzog, 788 F.3d 1017, 1024 (9th Cir. 2015) (“Questions of 21 standing are . . . reviewed de novo, but underlying factual findings are reviewed for clear error.”); ASPCA v. Feld Entm’t, Inc., 659 F.3d 13, 19 (D.C. Cir. 2011) (same); Me. People’s All. v. Mallinckrodt, Inc., 471 F.3d 277, 283 (1st Cir. 2006) (same). B. Application The Secretary raises two arguments that Mr. Bednasek lacks standing. First, the Secretary argues that Mr. Bednasek did not suffer an injury-in-fact because he was not a Kansas resident and so he was ineligible to vote. Second, the Secretary argues that Mr. Bednasek’s harm was not caused by the DPOC requirement and was instead self-inflicted. We reject both arguments before briefly assuring ourselves that Mr. Bednasek’s injury-in-fact is redressable. 1. Injury-in-fact The Secretary argues that Mr. Bednasek did not suffer “an invasion of a legally protected interest,” Spokeo, 136 S. Ct. at 1548 (quoting Lujan, 504 U.S. at 560), when his application was cancelled because Mr. Bednasek ostensibly was not a resident of Kansas and so was ineligible to vote there. For purposes of “determining the residence of a person offering to vote” under Kansas law, a person’s residence is “the place adopted by a person as such person’s place of habitation, and to which, whenever such person is absent, such person has the intention of returning.” Kan. Stat. Ann. § 25-407. “That this determination is not 22 based solely on intent has been recognized by the Kansas Supreme Court, which has stated that a citizen has the right to change his residence permanently or temporarily, and that ‘whether he does so, or which he does, is determined by his acts and his intentions.’” Warren v. Gaston, 55 F. Supp. 2d 1230, 1237 (D. Kan. 1999) (quoting State ex rel. Parker v. Corcoran, 128 P.2d 999, 1003 (Kan. 1942)). While Mr. Bednasek’s family is from Texas, he was enrolled as a full-time student at the University of Kansas when he attempted to register to vote. Moreover, in the fall of 2015, he cancelled his Texas voter registration and attempted to register in Kansas because he “considered [him]self to be a resident in Kansas, and [he] wanted to vote in Kansas elections.” Aplt.’s App., Vol. 38, at 9339. In applying to register in Kansas, he swore that he was a Kansas resident and that he had abandoned his former residence. And, as part of this lawsuit, he submitted an affidavit wherein he stated that he was a Kansas resident and had “no intent to leave Kansas in the future.” Id., Vol. 48, at 11692. The district court thus concluded that Mr. Bednasek was a resident of Kansas. The Secretary, however, points out that Mr. Bednasek paid out-of-state tuition at the University of Kansas, had a car that he jointly owned with his parents that was registered in Texas, had a car insurance policy—for which his parents are listed as the policy holders—that was also registered to his parents’ Texas home, and applied for and received a Texas driver’s license. The Secretary 23 argues that these actions undermine Mr. Bednasek’s declared intentions. But in rejecting a motion to dismiss on standing grounds, the district court stated: “the Court finds that neither Bednasek’s Texas driver’s license, nor his Texas automobile registration and insurance, objectively disprove Bednasek’s repeated attestations of Kansas residency.” Id., Vol. 49, at 11825 (Mem. & Order, filed July 29, 2016). This factual finding about Mr. Bednasek’s residence was incorporated in the summary judgment and trial rulings. We only review the district court’s factual findings for clear error. See Protocols, 549 F.3d at 1298; McCormack, 788 F.3d at 1024; ASCPA, 659 F.3d at 19; Me. People’s All., 471 F.3d at 283. And this factual finding was not clearly erroneous: Mr. Bednasek had chosen to live in Kansas, canceled his Texas voter registration, and sworn that he intended to remain in Kansas. The district court could weigh all of this evidence and make a credibility determination. 2 We conclude that the district 2 Moreover, the standard for paying in-state tuition is more onerous than that for voter residency. Kansas’s regulations concerning in-state tuition state that “[v]oting or registration for voting in Kansas,” “standing alone, ordinarily shall not constitute sufficient evidence of a change to Kansas residence.” Kan. Admin. Regs. § 88S3S2(c)(1). The regulations also create a presumption that individuals enrolled in academic programs are not residents for purposes of in-state tuition. Id. § 88S3S2(d). No similar presumption is in the voter-residency statute. And the relevant statute requires an individual to be a domiciliary resident of Kansas for twelve months prior to receiving in-state tuition, see Kan. Stat. Ann. § 76S729(a)(1); no such requirement is in the voter-registration residency definition. Thus, the fact that Mr. Bednasek paid out- of-state tuition does not tell us much. Additionally, Kansas’s then-existing (continued...) 24 court’s finding that Mr. Bednasek was a Kansas resident for purposes of voter registration was not clearly erroneous. The Secretary counters by citing two Kansas cases: Willmeth v. Harris, 403 P.2d 973 (Kan. 1965), and Gleason v. Gleason, 155 P.2d 465 (Kan. 1945). But these cases acknowledge that residency is a “question of fact” that turns on “all the surrounding facts and circumstances,” Gleason, 155 P.2d at 467; see also Willmeth, 403 P.2d at 978 (concluding “[t]here is ample evidence in the record to support the trial court’s findings” about an individual’s residency); thus, those authorities do nothing to change our conclusion that the district court’s factual finding that Mr. Bednasek was a Kansas resident was not clearly erroneous. Moreover, neither case was interpreting residency under the relevant voter- registration statute, and so neither could provide a persuasive reason to depart from the above analysis in any event. In sum, we conclude that district court did not clearly err in concluding that Mr. Bednasek was a resident of Kansas and thus suffered an injury-in-fact when the Secretary cancelled his registration application. 2 (...continued) guidance on “Election Standards” stated that residency for purposes of voter registration “is not related to or affected by vehicle registration,” and so the Texas vehicle registration also is far from dispositive here. Aplt.’s App., Vol. 49, at 11825. Likewise, we do not find the fact that his parents linked the auto insurance policy with their house to be telling. 25 2. Causation The Secretary next argues that Mr. Bednasek did not have standing to sue because his injury was “self-inflicted” and thus not fairly traceable to Kansas’s conduct. Aplt.’s Opening Br. at 23. The Secretary’s argument turns on the following facts: that Mr. Bednasek possessed DPOC that was located at his parent’s home in Texas, that he acquired a copy of this DPOC at some point during his Kansas residency in order to apply to the Navy, and that he stated that he did not bring the DPOC when he attempted to register to vote because he did not agree with the law. But we rejected similar arguments in Fish I. See 840 F.3d at 716 n.5, 753S54. We stated that “our cases show that typically a finding of self-inflicted harm results from either misconduct or something akin to entering a freely negotiated contractual arrangement, not from a failure to comply with an allegedly unlawful regime.” Id. at 753 (collecting cases). “[W]e reject[ed] the notion that the source of an injury is a litigant’s decision not to comply with an allegedly unlawful state regime, rather than the regime itself.” Id. at 754 (emphasis added). “Were this notion to apply in a case like this one, a court could never enjoin enforcement of an unlawful statute if the plaintiffs could have complied with the statute but elected not to; this hypothetical scenario borders on the absurd.” Id. This all remains true now. 26 Our opinion in ACLU of New Mexico v. Santillanes, supra, underscores the point. There, in-person voters were required to produce voter identification. We concluded that the plaintiffs who intended to vote in future elections and would be required to present identification in those elections had standing. 546 F.3d at 1319. This was the case even though “no individual [plaintiff] lack[ed] photo identification.” Id. Kansas’s denial of Mr. Bednasek’s application to register to vote for lack of DPOC functions analogously to the voter-identification law in Santillanes; the requirement that Mr. Bednasek provide DPOC is sufficient to establish standing. Cf. Common Cause/Georgia v. Billups, 554 F.3d 1340, 1351S52 (11th Cir. 2009) (“Requiring a registered voter either to produce photo identification to vote in person or to cast an absentee or provisional ballot is an injury sufficient for standing.”). In arguing to the contrary, the Secretary relies most heavily on Clapper v. Amnesty International USA, 568 U.S. 398 (2013). In that case, the plaintiffs asserted they were “suffering ongoing injuries” because they had chosen “to take costly and burdensome measures to protect the confidentiality of their international communications” from the potential of surveillance under the Foreign Intelligence Surveillance Act. Id. at 402, 415. The Supreme Court rejected the notion that these expenditures established standing: “respondents cannot manufacture standing merely by inflicting harm on themselves based on 27 their fears of hypothetical future harm that is not certainly impending.” Id. at 416. But unlike the hypothetical enforcement in Clapper, the DPOC requirement here actually was enforced against Mr. Bednasek. And it was that enforcement that caused Mr. Bednasek’s injury-in-fact, not any decision he made to spend money to avoid a hypothetical injury. Furthermore, the suit here concerns the fundamental right to vote; it does not concern “[the] actions of the political branches in the fields of intelligence gathering and foreign affairs,” where the Court engages in an especially rigorous standing inquiry. Id. at 408S09. We thus conclude Mr. Bednasek has demonstrated that there is a substantial likelihood that his injury-in-fact was caused by the Secretary’s actions. See Bronson, 500 F.3d at 1109–10. 3. Redressability And while not challenged by the Secretary, we have no doubt that there is “a ‘likel[ihood]’ that the injury ‘will be redressed by a favorable decision.’” Susan B. Anthony List, 573 U.S. at 158 (alteration in original) (quoting Lujan, 504 U.S. at 560S61). An order enjoining the Secretary from enforcing the DPOC requirement will ensure that approximately 30,000 voters whose applications were canceled or suspended are registered to vote. Cf. Santillanes, 546 F.3d at 1319. We thus conclude Mr. Bednasek has established standing. 28 III. Discussion Turning to the merits of the two challenges to the DPOC requirement, we summarize our standard of review before considering whether the DPOC requirement unconstitutionally burdens the right to vote and whether the DPOC requirement is preempted by section 5 of the NVRA. We conclude that the DPOC requirement is both unconstitutional and preempted by section 5 of the NVRA, and thus affirm the district court’s judgment. A. Standard of Review “We review the district court’s legal conclusions in a bench trial de novo; findings of fact will not be set aside unless clearly erroneous.” FTC v. Chapman, 714 F.3d 1211, 1216 (10th Cir. 2013) (quoting Ryan v. Am. Nat. Energy Corp., 557 F.3d 1152, 1157 (10th Cir. 2009)); accord United States v. Estate of St. Clair, 819 F.3d 1254, 1264 (10th Cir. 2016). Because the Secretary only challenges the court’s legal conclusions, see Aplt.’s Opening Br. at 21 (explaining that “the State asserts only legal error”), our review is de novo. B. Whether Kansas’s DPOC Requirement Violates the Fourteenth Amendment’s Equal Protection Clause We first ask whether Kansas’s DPOC requirement violates the Fourteenth Amendment’s Equal Protection Clause. We structure this inquiry by setting forth the Anderson-Burdick balancing test that governs our analysis, see Burdick v. Takushi, 504 U.S. 428 (1992); Anderson v. Celebrezze, 460 U.S. 780 (1983), and 29 then analyzing the constitutionality of the DPOC requirement under that test by weighing the burdens the voting requirement imposes against the interests the government asserts. We conclude that the DPOC requirement is unconstitutional and uphold the district court’s injunction against its enforcement. 1. Governing Legal Principles Crawford v. Marion County Election Board, supra, instructs that when examining a claim that a state law unconstitutionally burdens a plaintiff’s right to vote under the Equal Protection Clause, we are to examine the burden that the state law places on the right to vote and then weigh the government’s asserted interests for imposing that law against the burden. In the following discussion, we introduce the so-called Anderson-Burdick balancing test before summarizing how it was applied in Crawford. a. The Anderson-Burdick Balancing Test The right to vote is “a fundamental political right, . . . preservative of all rights.” Dunn v. Blumstein, 405 U.S. 330, 336 (1972) (omission in original) (quoting Reynolds v. Sims, 377 U.S. 533, 562 (1964)); Fish I, 840 F.3d at 752 (“There can be no dispute that the right to vote is a constitutionally protected fundamental right.”); see also Harper v. Va. State Bd. of Elections, 383 U.S. 663, 670 (1966) (“[T]he right to vote is too precious, too fundamental to be so burdened or conditioned.”). “No right is more precious in a free country than that 30 of having a voice in the election of those who make the laws under which, as good citizens, we must live. Other rights, even the most basic, are illusory if the right to vote is undermined.” Wesberry v. Sanders, 376 U.S. 1, 17 (1964). Nevertheless, “[i]t does not follow . . . that the right to vote in any manner and the right to associate for political purposes through the ballot are absolute.” Burdick, 504 U.S. at 433; accord Utah Republican Party v. Cox, 892 F.3d 1066, 1076 (10th Cir. 2018), cert. denied, 139 S. Ct. 1290 (2019). Instead, the Constitution allows states to “prescribe ‘the Times, Places and Manner of holding Elections for Senators and Representatives,’ and the Court therefore has recognized that States retain the power to regulate their own elections.” Burdick, 504 U.S. at 433 (citation omitted) (quoting U.S. C ONST . art. I, § 4, cl. 1). These regulations can help protect the right to vote by ensuring that elections are “fair and honest” and that “some sort of order, rather than chaos, . . . accompan[ies] the democratic processes.” Id. (quoting Storer v. Brown, 415 U.S. 724, 730 (1974)); accord Campbell v. Buckley, 203 F.3d 738, 745 (10th Cir. 2000); see Timmons v. Twin Cities Area New Party, 520 U.S. 351, 358 (1997) (“States may, and inevitably must, enact reasonable regulations of parties, elections, and ballots to reduce election- and campaign-related disorder.”). However, there is “[n]o bright line,” Timmons, 520 U.S. at 359, that separates those regulations that properly impose order—thereby protecting the 31 fundamental right to vote—from those that unduly burden it—thereby undermining it. See Crawford, 553 U.S. at 190 (plurality opinion of Stevens, J.) (explaining that instead of applying a “litmus test” courts must make a “hard judgment”); Anderson, 460 U.S. at 789 (explaining that instead of applying a “litmus-paper test” “a court must resolve such a challenge by an analytical process that parallels its work in ordinary litigation”). Instead, the Supreme Court has instructed that: A court considering a challenge to a state election law must weigh “the character and magnitude of the asserted injury to the rights protected by the First and Fourteenth Amendments that the plaintiff seeks to vindicate” against “the precise interests put forward by the State as justifications for the burden imposed by its rule,” taking into consideration “the extent to which those interests make it necessary to burden the plaintiff’s rights.” Burdick, 504 U.S. at 434 (quoting Anderson, 460 U.S. at 789); accord Santillanes, 546 F.3d at 1320 (“[T]he appropriate test when addressing an Equal Protection challenge to a law affecting a person’s right to vote is to ‘weigh the asserted injury to the right to vote against the precise interests put forward by the State as justifications for the burden imposed by its rule.’” (quoting Crawford, 553 U.S. at 190)). This balancing test has become known as the Anderson-Burdick balancing 32 test. See Cox, 892 F.3d at 1077 (“[W]e analyze electoral regulations using the now-familiar Anderson-Burdick balancing test.”). 3 Both parties agree that this Anderson-Burdick balancing test applies here. We thus turn to examine how the Anderson-Burdick test was applied to a related voting restriction in Crawford v. Marion County Election Board, supra. 3 The Supreme Court’s cases have equivocated over which provision of the Constitution mandates this balancing test. Compare Anderson, 460 U.S. at 786 n.7 (“[W]e base our conclusions directly on the First and Fourteenth Amendments and do not engage in a separate Equal Protection Clause analysis. We rely, however, on the analysis in a number of our prior election cases resting on the Equal Protection Clause of the Fourteenth Amendment. These cases, applying the ‘fundamental rights’ strand of equal protection analysis, have identified the First and Fourteenth Amendment rights implicated by restrictions on the eligibility of voters and candidates, and have considered the degree to which the State’s restrictions further legitimate state interests.” (collecting cases)), with Crawford, 553 U.S. at 207 n.* (plurality opinion of Scalia, J.) (“A number of our early right-to-vote decisions, purporting to rely upon the Equal Protection Clause, strictly scrutinized nondiscriminatory voting laws requiring the payment of fees.” (collecting cases)). Our court, however, has traced the Anderson-Burdick balancing test to the Equal Protection Clause. See Santillanes, 546 F.3d at 1319S20; see also Edward B. Foley, Due Process, Fair Play, and Excessive Partisanship: A New Principle for Judicial Review of Election Laws, 84 U. C HI . L. R EV . 655, 674 (2017) (explaining that the Supreme Court’s “equal-protection-for-voting jurisprudence has evolved into what is known as the ‘Anderson-Burdick balancing test’”). Yet, as explained above, the Anderson- Burdick balancing test does not entail “a traditional equal-protection inquiry.” Democratic Exec. Comm. of Fla. v. Lee, 915 F.3d 1312, 1319 (11th Cir. 2019); see also Green Party of Tenn. v. Hargett, 791 F.3d 684, 692 (6th Cir. 2015) (discussing relationship between Anderson-Burdick and the Equal Protection Clause, and collecting cases). 33 b. Crawford v. Marion County Election Board Crawford involved “the constitutionality of an Indiana statute requiring citizens voting in person on election day, or casting a ballot in person at the office of the circuit court clerk prior to election day, to present photo identification issued by the government.” 553 U.S. at 185 (plurality opinion of Stevens, J.). The Court—in two plurality opinions, one by Justice Stevens and one by Justice Scalia—upheld the constitutionality of the statute under the Anderson-Burdick balancing test. Id. at 189S90 (plurality opinion of Stevens, J.); id. at 204S05 (plurality opinion of Scalia, J.). We have held that Justice Stevens’s opinion is controlling, and so we focus on that opinion here. See Santillanes, 546 F.3d at 1321 (“Following Crawford, it appears that Justice Stevens’s plurality opinion controls, a position advocated by the Plaintiffs in the present case because it is the narrowest majority position.”); id. at 1321S25 (citing, in the analysis, only to Justice Stevens’s opinion); see also Foley, supra, at 676 (“The controlling opinion in [Crawford] was written by Stevens and joined by Roberts and Kennedy.”). Justice Stevens’s opinion—joined by Chief Justice Roberts and Justice Kennedy—started its discussion of the relevant legal standard with Harper v. Virginia State Board of Elections, supra, a case that invalidated a poll-tax under the Equal Protection Clause. Justice Stevens explained that Harper had not 34 invalidated the poll-tax because of any racial classification or animus; instead, it invalidated the poll-tax as “invidious because it was irrelevant to the voter’s qualifications.” Crawford, 553 U.S. at 189; see Harper, 383 U.S. at 668 (“To introduce wealth or payment of a fee as a measure of a voter’s qualifications is to introduce a capricious or irrelevant factor. The degree of the discrimination is irrelevant. In this context—that is, as a condition of obtaining a ballot—the requirement of fee paying causes an ‘invidious’ discrimination that runs afoul of the Equal Protection Clause.” (citation omitted) (quoting Skinner v. Okla. ex rel. Williamson, 316 U.S. 535, 541 (1942))). Justice Stevens thus deduced the rule that “even rational restrictions on the right to vote are invidious if they are unrelated to voter qualifications.” Crawford, 553 U.S. at 189. At the same time, Justice Stevens acknowledged “that ‘evenhanded restrictions that protect the integrity and reliability of the electoral process itself’ are not invidious and satisfy the standard set forth in Harper.” Id. at 189S90 (quoting Anderson, 460 U.S. at 788 n.9). But determining how to “neatly separate valid from invalid restrictions” is not always an easy task. Id. at 190. Notably, Justice Stevens applied the Anderson-Burdick balancing test even though the voter-identification law at issue was not “unrelated to voter qualifications.” Id. at 189; see id. at 193 (noting that Congress had indicated “that photo identification is one effective method of establishing a voter’s 35 qualification to vote”). Instead, “[h]owever slight th[e] burden [imposed on the right to vote] may appear, as Harper demonstrates, it must be justified by relevant and legitimate state interests ‘sufficiently weighty to justify the limitation.’” Id. at 191 (quoting Norman v. Reed, 502 U.S. 279, 288S89 (1992)). In other words, Anderson-Burdick scrutiny is required even when the burden imposed by a voter- identification law has some relationship to voter qualifications and even when the burden imposed may appear slight. An important question then is exactly what this scrutiny entails. Justice Stevens’s opinion explained that the degree of scrutiny was “flexible.” Id. at 190 n.8; see Burdick, 504 U.S. at 434 (explaining that under Anderson the mere fact that a burden was imposed does not itself mandate strict scrutiny; instead, “a more flexible standard applies”); accord Campbell v. Davidson, 233 F.3d 1229, 1233 (10th Cir. 2000) (applying “the flexible standard of Burdick”). This “flexib[ility]” acknowledges that we must engage in a case-specific inquiry based on (1) “the character and magnitude of the asserted injury to the rights . . . that the plaintiff seeks to vindicate,” (2) “the precise interests put forward by the State as justifications for the burden imposed by its rule,” and (3) “the extent to which those [state] interests make it necessary to burden the plaintiff’s rights.” Burdick, 504 U.S. at 434 (quoting Anderson, 460 U.S. at 789). Thus, the scrutiny we apply will wax and wane with the severity of the burden imposed on the right to vote in 36 any given case; heavier burdens will require closer scrutiny, lighter burdens will be approved more easily. Id. (“Under this standard, the rigorousness of our inquiry into the propriety of a state election law depends upon the extent to which a challenged regulation burdens First and Fourteenth Amendment rights.”). We, and our sister circuits and commentators, have referred to this as a “sliding scale” test. Navajo Nation v. San Juan Cty., 929 F.3d 1270, 1283 (10th Cir. 2019); see Ariz. Libertarian Party v. Hobbs, 925 F.3d 1085, 1090 (9th Cir. 2019) (“We have described this approach as a ‘sliding scale’—the more severe the burden imposed, the more exacting our scrutiny; the less severe, the more relaxed our scrutiny.” (quoting Ariz. Green Party v. Reagan, 838 F.3d 983, 988 (9th Cir. 2016))); Foley, supra, at 675 (“Anderson and Burdick create a kind of sliding-scale balancing test, whereby the strength of the state’s justification required to defend its law depends on the severity of the burden that the law imposes on the would-be voter’s opportunity to cast a ballot . . . .”); see also Ne. Ohio Coal. for the Homeless v. Husted, 696 F.3d 580, 592 (6th Cir. 2012) (per curiam) (“While a rational basis standard applies to state regulations that do not burden the fundamental right to vote, strict scrutiny applies when a state’s restriction imposes ‘severe’ burdens. For the majority of cases falling between 37 these extremes, we apply the ‘flexible’ Anderson/Burdick balancing test.” (citations omitted)). 4 Justice Stevens—applying this “flexible” approach—then concluded that the photographic-identification requirement at issue in Crawford “impose[d] only a limited burden on voters’ rights.” 553 U.S. at 203 (quoting Burdick, 504 U.S. at 439). In particular, Justice Stevens relied heavily on the district court’s conclusion that the challengers “had ‘not introduced evidence of a single, individual Indiana resident who will be unable to vote’” as a result of the law “or who will have his or her right to vote unduly burdened by its requirements.” Id. at 187 (quoting Ind. Democratic Party v. Rokita, 458 F. Supp. 2d 775, 783 (S.D. 4 Justice Scalia would have abandoned this traditional understanding of the Anderson-Burdick balancing test as a flexible, sliding scale test. Instead, he characterized Burdick as “forg[ing] Anderson’s amorphous ‘flexible standard’ into” a more rigid “two-track approach.” Crawford, 553 U.S. at 205 (plurality opinion of Scalia, J.). Under his view, strict scrutiny would be applied to “severe” burdens while “nonsevere, nondiscriminatory restrictions” would be reviewed under “a deferential ‘important regulatory interests’ standard.” Id. at 204S05 (quoting Burdick, 504 U.S. at 433S34 ). Applying this framework to the law at issue in Crawford, Justice Scalia would have concluded that the burden imposed by Indiana’s photo-identification law was “simply not severe” and thus justified by Indiana’s generalized interests under a deferential standard. Id. at 209. But six Members of the Court—Chief Justice Roberts, Justice Stevens, Justice Kennedy, see id. at 190 n.8, and all three dissenters, see id. at 210 (Souter, J., joined by Ginsburg, J., dissenting); id. at 237 (Breyer, J., dissenting)—rejected Justice Scalia’s reframing of the Anderson-Burdick balancing test as involving only two distinct tracks. And, as we have previously held that Justice Stevens’s opinion is controlling, we must apply the “flexible” approach to the Anderson- Burdick balancing test that was used by Justice Stevens, see Crawford, 553 U.S. at 190 n.8, as well as by the Court’s earlier cases, see Burdick, 504 U.S. at 434. 38 Ind. 2006)). Additionally, while Justice Stevens acknowledged that the law could impose a burden on those without photo identification, “the evidence in the record d[id] not provide [the Court] with the number of registered voters without photo identification.” Id. at 200. “Much of the argument about the numbers of such voters c[ame] from extrarecord, postjudgment studies, the accuracy of which ha[d] not been tested in the trial court.” Id. Thus, he concluded that any costs associated with requiring voters to acquire freely provided photographic identification did not “qualify as a substantial burden on the right to vote, or even represent a significant increase over the usual burdens of voting.” Id. at 198. This determination that the burden was not substantial was “record-based.” Id. at 208 (plurality opinion of Scalia, J.); see id. at 189 (plurality opinion of Stevens, J.) (“[T]he evidence in the record is not sufficient to support a facial attack on the validity of the entire statute . . . .” (emphasis added)); id. at 200 (plurality opinion of Stevens, J.) (“[O]n the basis of the evidence in the record it is not possible to quantify either the magnitude of the burden on this narrow class of voters or the portion of the burden imposed on them that is fully justified.” (emphasis added)). In light of this weak evidence of a burden, Justice Stevens concluded that the state’s justifications were “sufficiently weighty to justify the limitation.” Id. at 191 (quoting Norman, 502 U.S. at 288S89). In particular, Indiana had “unquestionably relevant” interests in election modernization, preventing voter 39 fraud, and safeguarding voter confidence. Id. Justice Stevens determined that these interests were “legitima[te]” and “importan[t],” even though, for example, “[t]he record contain[ed] no evidence of any [voter-impersonation] fraud actually occurring in Indiana.” Id. at 194, 196. Thus, the Crawford opinion demonstrates that when there is limited evidence of a burden on the right to vote, the state need not present concrete evidence to justify its assertion of legitimate or important generalized interests. See Santillanes, 546 F.3d at 1323 (“In requiring the City to present evidence of past instances of voting fraud, the district court imposed too high a burden on the City.”). However, when a more substantial burden is imposed on the right to vote, our review of the government’s interests is more “rigorous[].” Burdick, 504 U.S. at 434; see also Anderson, 460 U.S. at 789 (“In passing judgment, the Court must not only determine the legitimacy and strength of each of those interests; it also must consider the extent to which those interests make it necessary to burden the plaintiff’s rights.”); League of Women Voters of N.C. v. North Carolina, 769 F.3d 224, 246 (4th Cir. 2014) (“North Carolina asserts goals of electoral integrity and fraud prevention. But nothing in the district court’s portrayal of the facts suggests that those are anything other than merely imaginable.”); Obama for Am. v. Husted, 697 F.3d 423, 433S34 (6th Cir. 2012) (holding voting regulation was not justified by “vague interest[s]” when the state had submitted “no evidence” to 40 justify its invocation of the interests); cf. Cal. Democratic Party v. Jones, 530 U.S. 567, 584 (2000) (explaining that the determination of whether the “four asserted state interests—promoting fairness, affording voters greater choice, increasing voter participation, and protecting privacy—are . . . compelling . . . is not to be made in the abstract, by asking whether fairness, privacy, etc., are highly significant values; but rather by asking whether the aspect of fairness, privacy, etc., addressed by the law at issue is highly significant”). Finally, in addition to Justice Stevens’s consideration of the “facial attack on the validity of the entire statute” in Crawford based on “the statute’s broad application to all Indiana voters,” 553 U.S. at 189, 202–03, he also parsed the burden imposed on specific categories of voters: “elderly persons born out of State, who may have difficulty obtaining a birth certificate; persons who because of economic or other personal limitations may find it difficult either to secure a copy of their birth certificate or to assemble the other required documentation to obtain a state-issued identification; homeless persons; and persons with a religious objection to being photographed,” id. at 199 (footnote omitted). But the record evidence concerning the burdens imposed on these voters was once again lackluster: “on the basis of the evidence in the record it is not possible to quantify either the magnitude of the burden on this narrow class of voters or the portion of the burden imposed on them that is fully justified.” Id. at 200. “[T]he evidence 41 in the record d[id] not provide [the Court] with the number of registered voters without photo identification,” “the deposition evidence presented in the District Court d[id] not provide any concrete evidence of the burden imposed on voters who currently lack photo identification,” and the “record sa[id] virtually nothing about the difficulties faced by either indigent voters or voters with religious objections to being photographed.” Id. at 200S01. Furthermore, the Court noted that the unique burdens on these classes of individuals would be mitigated by a provision in the Indiana law that allowed certain “voters without photo identification [to] cast provisional ballots that will ultimately be counted.” Id. at 199. 5 In sum, Crawford teaches that we must balance any burden on the right to vote imposed by the DPOC requirement against the government’s asserted interests as justifications for imposing that burden. We must apply the Anderson- 5 Justice Scalia criticized the “lead opinion” for this focus on the burden imposed on specific populations. 553 U.S. at 204S05. In his view, traditional equal-protection principles required evaluating the burden the voter-identification requirement imposed on all voters, even if different populations felt the impacts of that single burden differently. Id. at 205; see also id. at 206 (arguing that, in its prior opinions, “when [the Court] began to grapple with the magnitude of burdens, [it] did so categorically and did not consider the peculiar circumstances of individual voters or candidates”). He particularly thought Justice Stevens’s focus on the requirement’s impact on the poor, disabled, and elderly was problematic because these are not suspect classifications. Id. at 207. He would have avoided “[t]he lead opinion’s record-based resolution” and instead held that “[t]he burden of acquiring, possessing, and showing a free photo identification is simply not severe.” Id. at 208S09. 42 Burdick balancing test flexibly, i.e., “the rigorousness of our inquiry into the propriety of [the DPOC requirement] depends upon the extent to which [it] burdens” voters’ rights. Burdick, 504 U.S. at 434. And, while we are to evaluate “the statute’s broad application to all . . . voters” to determine the magnitude of the burden, Crawford, 553 U.S. at 202–03 (plurality opinion of Stevens, J.), we may nevertheless specifically consider the “limited number of persons” on whom “[t]he burdens that are relevant to the issue before us” will be “somewhat heavier,” id. at 198S99 (plurality opinion of Stevens, J.); see also Harper, 383 U.S. at 668 (holding poll tax facially unconstitutional while identifying the specifically pernicious effect such a tax has on those unable to pay it); Nathaniel Persily, Fig Leaves and Tea Leaves in the Supreme Court’s Recent Election Law Decisions, 2008 S UP . C T . R EV . 89, 101 (2008) (“Voting laws that are unconstitutional on their face are usually so because a minority (even a nonsuspect one) is disadvantaged.”). 2. Application In the following discussion, we (a) set out the burdens imposed on the right to vote by the DPOC requirement, (b) discuss the interests that the Secretary offers to justify those burdens, and then (c) arrive at the “hard judgment” Crawford demands. We conclude that the significant burden quantified by the 31,089 voters who had their registration applications canceled or suspended 43 requires us to increase the “rigorousness of our inquiry,” Burdick, 504 U.S. at 434, and that such an inquiry demonstrates that the precise interests put forward by the Secretary do not justify the burden imposed on the right to vote. Thus, we conclude the DPOC requirement is unconstitutional. a. Burden on Voters Based primarily on the district court’s finding that 31,089 applicants were prevented from registering to vote because of the DPOC requirement, we conclude that the burden imposed on the right to vote by the DPOC requirement was significant and requires heightened scrutiny. 6 In arriving at this conclusion, we reject the Secretary’s counterargument that the burden imposed here compares favorably to the burden that the Supreme Court found insubstantial in Crawford. The district court found that—before the preliminary injunction in Fish I was issued—14,770 individuals had applied to vote but had their applications suspended by the Secretary for failure to provide DPOC. In addition to these suspended applications, another 16,319 individuals had their applications canceled by the Secretary for failure to provide DPOC. The district court thus found that “31,089 total applicants . . . were denied registration for failure to 6 In labeling the burden “significant,” we—applying Justice Stevens’s “flexible” approach in Crawford—conclude that the burden is at least somewhere in between the two poles identified by Justice Scalia, i.e., “severe” and “nonsevere.” 44 provide DPOC.” Aplt.’s App., Vol. 47, at 11447. Expert testimony demonstrated that “[c]anceled or suspended applicants represented 12.4% of new voter registrations between January 1, 2013[,] and December 11, 2015,” or “approximately 12% of the total voter registration applications submitted since the law was implemented.” Id. at 11448S49. And an expert opined that the total number of applicants with suspended or canceled applications would have increased but for the injunction. And despite the eventual injunction, many of the would-be voters—including both Mr. Fish and Ms. Bucci—actually were disenfranchised and “were not registered in time to vote in the [intervening] 2014 election by operation of the DPOC law.” Id. at 11450S51. The Secretary challenges none of these findings on appeal. These factual findings create a fundamental distinction between this case and Crawford: based on an extensive record, the district court here concluded that the Kansas Secretary of State actually denied approximately thirty thousand would-be voters’ registration applications in his implementation of the DPOC requirement, while, in Crawford, the scant evidence before the Court left it with the unenviable task of attempting to estimate the magnitude of the burden on voting rights, largely from untested extra-record sources. The district court in Crawford had found that the challengers “had ‘not introduced evidence of a single, individual Indiana resident who w[ould] be unable to vote’” as a result of 45 the law. 553 U.S. at 187 (quoting Ind. Democratic Party, 458 F. Supp. 2d at 783). The Court concluded that “the evidence in the record d[id] not provide [it] with the number of registered voters without photo identification.” Id. at 200. This was because the district court had found the challengers’ experts “to be ‘utterly incredible and unreliable.’” Id. (quoting Ind. Democratic Party, 458 F. Supp. 2d at 803). Contrast that with the district court’s factual finding here that 31,089 total applicants were denied registration for failure to provide DPOC. While the record in Crawford led Justice Stevens to conclude that the burden there was slight, the record before us instead leads us to conclude that the burden on the right to vote here was significant. Moreover, in finding Indiana’s statute constitutional, Justice Stevens relied on the fact that—even for that unquantified number of voters who would lack a photo identification at the next election—Indiana’s statute provided that, “if eligible, voters without photo identification may cast provisional ballots that will ultimately be counted.” Id. at 199. In order to have such a provisional ballot counted, an eligible voter without photographic identification only needed to execute the required affidavit at the local circuit court clerk’s office. Id. Additionally, “[p]resumably most voters casting provisional ballots w[ould] be able to obtain photo identifications before the next election,” and thus would not bear this burden going forward. Id. at 199 n.19. Thus, the Court concluded that 46 the statute as a whole “impose[d] only a limited burden on voters’ rights.” Id. at 203 (quoting Burdick, 504 U.S. at 439). But as the district court here concluded, Kansas’s DPOC requirement offered no similar safety valve. See Aplt.’s App., Vol. 47, at 11520 (“[U]nlike the Indiana law in Crawford, an eligible Kansas applicant on the suspense or cancellation list does not have the option to fill out a provisional ballot, produce DPOC after the election, and have their ballot counted.”). The Kansas law allows voters who submit an application without DPOC to supplement that application at any point within the following ninety days. See Kan. Admin. Regs. § 7S23S15(b). Counties have been instructed to contact each voter that has submitted an application without DPOC three times within this ninety-day period. See Aplt.’s App., Vol. 47, at 11446. And, if they do not have DPOC, they may apply for a hearing where they are otherwise able to prove their citizenship. See Kan. Stat. Ann. § 25S2309(m). But these provisions are significantly less effective than the safety valve in Crawford because that safety valve allowed certain individuals to actually cast provisional votes while these provisions do not. Thus, voters who were not registered due to a lack of DPOC—including those who did not know that they were not registered, like Mr. Boynton and Mr. Stricker—could show up to vote but be turned away without a backup option for them to cast votes. 47 Thus, we conclude that this case presents fundamental differences with Crawford—differences that make the burden on the right to vote more substantial here than in Crawford. In sum, the burden imposed on the right to vote by the DPOC requirement was significant and requires heightened scrutiny. 7 The Secretary offers several arguments to the contrary, but we do not find any of them persuasive. First, the Secretary argues that—as a matter of law—the DPOC requirement here only imposes a limited burden on voters. He bases his 7 We acknowledge that the record also reflects certain instances where the disenfranchisement of voters arguably was not directly related to the burdens associated with the legal requirement that voters produce DPOC. That is because, in some instances, the voters in fact produced DPOC to Kansas state employees and their disenfranchisement apparently stemmed from bureaucratic snafus in Kansas’s implementation of the DPOC regime. Notably, in some instances, disenfranchisement was the result of the apparently inadvertent failures of state employees to record and to give effect to the DPOC that prospective voters provided. For example, Mr. Stricker and Mr. Boynton both brought DPOC when they went to register and either provided it to the clerk or were told that they did not need to. The Secretary does not argue that disenfranchisement stemming from such bureaucratic or administrative failures in the implementation of the DPOC regime should not be considered by us to be part and parcel of that regime, nor, relatedly, that such disenfranchisement should be excluded from our analysis of the alleged burden caused by the DPOC regime under Anderson-Burdick. Cf. Democratic Exec. Comm. of Fla., 915 F.3d at 1319S21 (considering “the way in which Florida implements the scheme” when determining the severity of the burden imposed under Anderson-Burdick); Ne. Ohio Coal. for the Homeless, 696 F.3d at 591S95 (considering burden imposed on the right to vote by “poll-worker error” under Anderson-Burdick). In particular, the Secretary does not argue that prospective voters disenfranchised through such bureaucratic snafus should not be counted, for purposes of the burden analysis, among the approximately 30,000 would-be voters that the district court found to be disenfranchised. Therefore, we have no occasion to consider any such argument further. 48 argument on Justice Stevens’s statement in Crawford that “the inconvenience of making a trip to the [Bureau of Motor Vehicles], gathering the required documents, and posing for a photograph surely does not qualify as a substantial burden on the right to vote, or even represent a significant increase over the usual burdens of voting.” 553 U.S. at 198. But Justice Stevens made that statement in light of the record then before the Court. Id. at 189 (concluding “the evidence in the record is not sufficient to support a facial attack on the validity of the entire statute” (emphasis added)); id. at 200 (“[O]n the basis of the evidence in the record it is not possible to quantify either the magnitude of the burden on this narrow class of voters or the portion of the burden imposed on them that is fully justified.” (emphasis added)). But while there was no evidence in Crawford that the inconvenience of going to the Bureau of Motor Vehicles constituted a significant burden, the approximately 30,000 would-be voters disenfranchised in this case provide a concrete evidentiary basis to find that a significant burden was imposed by the DPOC requirement. Thus, the fundamental differences between the record in this case and in Crawford lead us to a different conclusion. Furthermore, we also reject the Secretary’s argument that features of the DPOC requirement made it—as a matter of law—less burdensome than the law in Crawford. While the Secretary notes that Kansas allows those without DPOC to register by meeting with him and other officials, this procedure has only been 49 used five times, and we agree with the district court’s finding that its byzantine nature “adds, not subtracts, from the burdensomeness of the law.” Aplt.’s App., Vol. 47, at 11526; cf. Harman v. Forssenius, 380 U.S. 528, 541–42 (1965) (finding unconstitutional what was “plainly a cumbersome procedure” to avoid a poll tax). The Secretary also argues that Kansas accepts DPOC through e-mail, fax, or mail, while the law at issue in Crawford required voters without photo identification to travel to the Bureau of Motor Vehicles. But the provisions the Secretary cites do not mention e-mail or fax. See Kan. Stat. Ann. § 25S2309(l), (t). Moreover, it is far from clear that requiring all registrants to provide DPOC—even though they need only submit it once—is less burdensome than Crawford’s requirement that voters bring identification to the polls. After all, many Indiana voters in Crawford no doubt would have driven to the polls and therefore already would have had their driver’s licenses with them; consequently, they would not have been obliged to take any further steps to vote. The Secretary also argues that there is no way to determine how many of the 31,089 would-be voters whose applications were suspended or denied “were actually unable (as opposed to just unwilling) to obtain a birth certificate or other evidence of citizenship.” Aplt.’s Opening Br. at 32; cf. Frank v. Walker, 768 F.3d 744, 749 (7th Cir. 2014) (“If people who already have copies of their birth certificates do not choose to get free photo IDs, it is not possible to describe the 50 need for a birth certificate as a legal obstacle that disfranchises them.”). In support, he notes that the district court acknowledged that “[t]here was little admissible evidence presented at trial about the rate of DPOC possession by suspended and canceled applicants.” Aplt.’s App., Vol. 47, at 11457; see also id. (“There is no evidence about how many canceled and suspended applicants in fact lack DPOC . . . .”). And, as his argument goes, if voters simply did not want to be inconvenienced by providing DPOC, this inconvenience does not necessarily constitute a cost that is beyond the “usual burdens of voting.” Crawford, 553 U.S. at 198 (plurality opinion of Stevens, J.). But the concrete record evidence of the disenfranchisement of the 31,089 would-be voters again provides reason to believe that the DPOC requirement does impose a significant burden on Kansas voters, even if some of those voters could have registered with DPOC. While the district court was unable to determine what percentage of the disenfranchised voters lacked DPOC, it did recount extensive testimony about individual voters like Mr. Fish and Ms. Bucci who lacked DPOC or faced significant costs to obtain it. When this testimonial evidence was combined with the statistical evidence of disenfranchised voters, the district court could properly conclude here that the DPOC requirement imposed a significant burden on the right to vote. 8 8 We note that a sister circuit has concluded—in the NVRA context—that this very DPOC requirement burdens the right to vote by imposing (continued...) 51 The Secretary further argues that Crawford is analogous to this case because the district court in Crawford had “estimated” that, “when the statute was enacted, around 43,000 Indiana residents lacked a state-issued driver’s license or identification card.” Id. at 187S88 (plurality opinion of Stevens, J.) (citing Ind. Democratic Party, 458 F. Supp. 2d at 807). The Secretary thus argues that, like here, there was widespread and quantified evidence of disenfranchisement in Crawford, and so we should reach the same result as the Court did there. But this argument ignores that the Supreme Court in Crawford found that this 43,000 number was meaningless because it told the Court “nothing about the number of free photo identification cards [that had been] issued since” the statute’s enactment and thus nothing about how many voters actually would be turned away from the polls. Id. at 202 n.20. We thus do not view the Court in Crawford as implying that the disenfranchisement of 43,000 voters would not be significant. Furthermore, the Court in Crawford concluded that “the evidence in the record d[id] not provide [it] with the number of registered voters without photo identification” and thus the number of voters who might be unable to vote. Id. at 8 (...continued) “onerous” processes that can lead would-be voters to give up. See League of Women Voters of the U.S. v. Newby, 838 F.3d 1, 13 (D.C. Cir. 2016) (“It does not matter whether [would-be voters were being held on a suspension list] because they lack access to the requisite documentary proof or simply because the process of obtaining that proof is so onerous that they give up. The outcome is the same—the abridgment of the right to vote.” (citations omitted)). 52 200 (emphasis added). In contrast, we know here that approximately 30,000 Kansans took affirmative and concrete steps to register to vote and were disenfranchised by application of the DPOC requirement. Therefore, we are not persuaded by the Secretary’s attempt here to analogize Crawford to this case. The Secretary relatedly argues that the district court overstated the burden by considering all 31,089 would-be voters with canceled or suspended applications because some of them might have later submitted DPOC to cure their applications if the district court had not enjoined the DPOC requirement. It is true that in Crawford Justice Stevens suggested that certain estimates of the number of voters without photographic identification were likely overstated because some voters may have obtained identification in the intervening months. Id. at 188 n.6, 202 n.20. However, the Secretary’s argument disregards the expert’s opinion here that the total number of applicants with suspended or canceled applications would have increased but for the injunction. The Secretary’s argument, in contrast, is based on sheer speculation: he makes no attempt to estimate how many of the would-be Kansas voters with canceled or suspended applications would have taken the step to submit DPOC. And, as Justice Stevens rightly pointed out in Crawford, “[s]upposition . . . is not an adequate substitute for admissible evidence subject to cross-examination in constitutional adjudication.” Id. at 202 n.20. And the concrete, admissible 53 evidence here indicates that—because of Kansas’s DPOC requirement—31,089 would-be voters were not permitted to vote; without doubt, that is a significant burden. In sum, we conclude that the DPOC requirement imposed a significant burden on the right to vote. 9 b. Asserted State Interests We now turn to “evaluate the interests put forward by the State as justifications for the burden imposed by its rule.” Id. at 190 (plurality opinion of Stevens, J.). The Secretary puts forward four interests: “(i) protecting the integrity of the electoral process, (ii) ensuring the accuracy of voter rolls, (iii) 9 Mr. Bednasek also argues that the DPOC requirement is suspect because it applies different registration requirements to those who registered before the statute was enacted than to those who registered after. See Kan. Stat. Ann. § 25-2309(n) (“Any person who is registered in this state on the effective date of this amendment to this section is deemed to have provided satisfactory evidence of citizenship and shall not be required to resubmit evidence of citizenship.”). But Mr. Bednasek concedes that this distinction is justified in part by Kansas’s “legitimate state interest” in protecting the reliance interests of voters who were registered under the previous regime, Aplee.’s Resp. Br. at 76, and the authority he provides that found certain grandfather clauses unconstitutional in the voting context concluded that the statutes were unconstitutional because they sought to avoid the import of the Fifteenth Amendment. See Lassiter v. Northampton Cty. Bd. of Elections, 360 U.S. 45, 49S50 (1959); Guinn v. United States, 238 U.S. 347, 363 (1915). No similar allegations have been made here, and so we do not base our analysis of the burden on this argument. 54 safeguarding voter confidence, and (iv) preventing voter fraud.” Aplt.’s Opening Br. at 33. We agree with the Secretary that each of these interests is legitimate in the abstract. However, the Secretary points to no concrete evidence that “those interests make it necessary to burden the plaintiff’s rights” in this case, Burdick, 504 U.S. at 434 (quoting Anderson, 460 U.S. at 789), and so—in the following section—we conclude that, on this record, these legitimate interests are insufficiently weighty to justify the limitations on the right to vote imposed by the DPOC requirement. While nominally distinct interests, three of the Secretary’s asserted interests—protecting the integrity of the electoral process, ensuring the accuracy of voter rolls, and preventing vote fraud—largely overlap. Each fundamentally can be boiled down to Kansas’s interest in making sure that only eligible voters vote in its elections. And we agree with the Secretary that “[t]here is no question about the legitimacy or importance of the State’s interest in counting only the votes of eligible voters.” Crawford, 553 U.S. at 196 (plurality opinion of Stevens, J.); see id. (“[T]he interest in orderly administration and accurate recordkeeping provides a sufficient justification for carefully identifying all voters participating in the election process.”). Likewise, when put in terms of electoral integrity, “[a] State indisputably has a compelling interest in preserving the integrity of its election process.” Purcell v. Gonzalez, 549 U.S. 1, 4 (2006) 55 (per curiam) (quoting Eu v. S.F. Cty. Democratic Cent. Comm., 489 U.S. 214, 231 (1989)). And we agree that “[t]he State’s interest is particularly strong with respect to efforts to root out fraud,” Doe v. Reed, 561 U.S. 186, 197 (2010), because fraud “drives honest citizens out of the democratic process and breeds distrust of our government,” id. (quoting Purcell, 549 U.S. at 4). “While the most effective method of preventing election fraud may well be debatable, the propriety of doing so is perfectly clear.” Crawford, 553 U.S. at 196 (plurality opinion of Stevens, J.). Thus, we agree with the Secretary that Kansas’s interest in counting only the votes of eligible voters is legitimate in the abstract, but, on this record, we do not see any evidence that such an interest made it necessary to burden voters’ rights here. The Secretary argues that—even if there is no inaccuracy or fraud to correct—the DPOC requirement furthers Kansas’s interest in increasing public confidence and participation in the democratic process. Aplt.’s Opening Br. at 35–36; see also id. at 34 (“Preventing Kansas from verifying the bedrock voter qualification of United States citizenship would undermine the integrity of the electoral process—whether widespread voter fraud exists or not.” (emphasis added)). Again, we agree with the Secretary that “[c]onfidence in the integrity of our electoral processes is essential to the functioning of our participatory democracy.” Purcell, 549 U.S. at 4. Such confidence is vital because “[v]oter 56 fraud drives honest citizens out of the democratic process and breeds distrust of our government.” Id. Voters “will feel disenfranchised” when they have reason to “fear their legitimate votes will be outweighed by fraudulent ones.” Id. Thus, the Supreme Court has recognized that “public confidence in the integrity of the electoral process has independent significance[] because it encourages citizen participation in the democratic process.” Crawford, 553 U.S. at 197 (plurality opinion of Stevens, J.). While we agree with the Secretary that Kansas has a legitimate interest in safeguarding voter confidence, we explain below why there is no evidence that the DPOC requirement furthers that interest. Thus, we agree that each of the interests asserted by the Secretary is legitimate in the abstract. However, we now turn to explain why—due to the significant burden that the DPOC requirement imposes on the right to vote and the lack of concrete evidence supporting the relevance of these interests in this case—we cannot conclude “those interests make it necessary to burden the plaintiff’s rights.” Burdick, 504 U.S. at 434 (quoting Anderson, 460 U.S. at 789). c. Balancing As we have discussed, “the rigorousness of our inquiry into the propriety of [the DPOC requirement] depends upon the extent to which [it] burdens” voters’ rights. Id. Here, the evidence of the approximately 30,000 disenfranchised voters means that heightened scrutiny is appropriate. Thus, we must look at more than 57 whether the proffered interests are legitimate in the abstract; we must ask whether the concrete evidence demonstrates that “those interests make it necessary to burden the plaintiff’s rights” in this case. Id. (quoting Anderson, 460 U.S. at 789); see Cal. Democratic Party, 530 U.S. at 584; League of Women Voters of N.C., 769 F.3d at 246; Obama for Am., 697 F.3d at 433S34. Doing so, we conclude that the Secretary has failed to introduce sufficiently weighty evidence to justify the burdens imposed on voters. To start, the district court found essentially no evidence that the integrity of Kansas’s electoral process had been threatened, that the registration of ineligible voters had caused voter rolls to be inaccurate, or that voter fraud had occurred. In particular, it found that, “at most, 67 noncitizens registered or attempted to register in Kansas over the last 19 years.” Aplt.’s App., Vol. 47, at 11519. Of these, “[a]t most, 39 noncitizens have found their way onto the Kansas voter rolls in the last 19 years.” Id. at 11520. The Secretary does not argue that these factual findings are clearly erroneous. Thus we are left with this incredibly slight evidence that Kansas’s interest in counting only the votes of eligible voters is under threat. Indeed, even as to those 39 noncitizens who appear on the Kansas voter rolls, the district court effectively found that “administrative anomalies” could account for the presence of many—or perhaps even most—of them there. Id. Supporting this determination is the fact that Kansas’s voter-registration 58 database included 100 individuals with purported birth dates in the 19th century and 400 individuals with purported birth dates after their date of voter registration. And so it is quite likely that much of this evidence of noncitizen registration is explained by administrative error. The Secretary also presented the district court with out-of-state evidence about election fraud and noncitizen registration. But the district court concluded that, “looking beyond Kansas, [the Secretary’s] evidence of noncitizen registration at trial was weak.” Id. at 11519. It explained at length why it excluded large portions of the Secretary’s expert testimony and found much of the remaining testimony unpersuasive. Id. (explaining that one of the Secretary’s experts was “credibly dismantled” by the architect of the survey upon which the expert had relied and that the court “d[id] not fully credit” a second expert’s testimony “given its inclusion of misleading and false assertions”). We have no doubt that inaccurate voter registrations exist in our country, see, e.g., Husted v. A. Philip Randolph Inst., --- U.S. ----, 138 S. Ct. 1833, 1838 (2018) (“It has been estimated that 24 million voter registrations in the United States—about one in eight—are either invalid or significantly inaccurate.”), but the Secretary fails to connect this generalized information to the DPOC requirement at issue here or to argue that the district court clearly erred in finding that “the trial evidence did not demonstrate the largescale problem urged by [the Secretary].” Aplt.’s App., 59 Vol. 47, at 11520. In light of the significant burden on the right to vote, we thus do not rely on the Secretary’s out-of-state evidence of voter-fraud and nonvoter registration. Finally and relatedly, while much of the above discussion focused on Kansas’s interest in counting only the votes of eligible voters, it is also true that the evidence did not demonstrate that Kansas’s interest in safeguarding voter confidence made it necessary to enact the DPOC requirement. In particular, the district court found that “the evidence in this case d[id] not show that the DPOC law furthers” Kansas’s “significant interest” in “maintaining confidence in the electoral process.” Id. at 11527S28. The district court found that, even under calculations from one of the Secretary’s experts, the estimated number of suspended applications that belonged to noncitizens was “statistically indistinguishable from zero,” while “more than 99% of the individuals” whose voter-registration applications were suspended were citizens who presumably would have been able to vote but for the DPOC requirement. Id. at 11491S92; id. at 11528 (“[T]he DPOC law disproportionately impacts duly qualified registration applicants, while only nominally preventing noncitizen voter registration.”). Thus, the district court found that this disproportionate impact on qualified registration applicants “also may have the inadvertent effect of eroding, instead of maintaining, confidence in the electoral system.” Id. at 11528. Again, while the 60 Secretary casts aspersions on these factual findings, he does not contest them as clearly erroneous. See Aplt.’s Opening Br. at 21 (“[T]he State asserts only legal error . . . .”). In sum, the burden on the right to vote evinced by the approximately 30,000 disenfranchised voters elevates “the rigorousness of our inquiry into the propriety of [the DPOC requirement].” Burdick, 504 U.S. at 434. And when we look at “‘the precise interests put forward by the State as justifications for the burden imposed by its rule,’ taking into consideration ‘the extent to which those interests make it necessary to burden the plaintiff’s rights,’” id. (quoting Anderson, 460 U.S. at 789), we see that the Secretary’s proffered justifications are not supported—and indeed in several places are undercut—by the facts found by the district court. We thus make the “hard judgment,” Crawford, 553 U.S. at 190 (plurality opinion of Stevens, J.), that the DPOC requirement unconstitutionally burdens the right to vote and uphold the district court’s judgment. To be sure, the Secretary argues that Crawford and our opinion in Santillanes prohibit us from examining whether there is any evidence behind his proffered justifications. In Crawford, Justice Stevens’s opinion accepted three justifications—election modernization, protection against voter fraud, and safeguarding voter confidence—in the abstract, i.e., without requiring evidence that these interests were at risk in Indiana or remedied by the photographic- 61 identification requirement at issue. Id. at 191S97 (plurality opinion of Stevens, J.). Similarly, in Santillanes, we concluded that “[i]n requiring the City to present evidence of past instances of voting fraud, the district court imposed too high a burden on the City.” 546 F.3d at 1323. Guided by Crawford, we relied on the city’s general invocation of its interest and Crawford’s citation to “flagrant examples of such fraud in other parts of the country,” id. (quoting Crawford, 553 U.S. at 195 (plurality opinion of Stevens, J.)), in concluding that “[p]revention of voter fraud and voting impersonation as urged by the City are sufficient justifications for a photo identification requirement for local elections,” id. But in Crawford and Santillanes, the Supreme Court and our court had concluded that there was only a light burden on the right to vote. Crawford, 553 U.S. at 202S03 (plurality opinion of Stevens, J.); Santillanes, 546 F.3d at 1322–23. While both of those decisions acknowledged that the laws at issue there could impose a burden on the right to vote, those burdens were limited and were further mitigated by the possibility of casting provisional ballots. But the evidence here that 31,089 voter-registration applications were either suspended or canceled differentiates the magnitude of the burden here from the magnitude of the burdens in those cases. Because “the rigorousness of our inquiry into the propriety of [the challenged law] depends upon the extent to which [it] burdens” voters’ rights, Burdick, 504 U.S. at 434, and neither of those decisions confronted 62 the level of disenfranchisement that the record evidence establishes that the DPOC requirement produced here, neither of those decisions applied the more robust scrutiny that we must here. And it is only after engaging in that heightened scrutiny that we find that the facts in this case do not support the conclusion that the Secretary’s legitimate interests justify the burdens that the DPOC law imposes on the right to vote. Thus, our approach here is entirely consistent with Crawford and Santillanes; the different result is driven by the different record of burdens and, consequently, the different level of judicial scrutiny applied. In sum, we conclude that the DPOC requirement is unconstitutional and uphold the district court’s injunction in Bednasek v. Schwab, No. 18-3134. 10 C. Whether Section 5 of the NVRA Preempts the DPOC Requirement We also uphold the district court’s entry of the injunction against the enforcement of the DPOC requirement with regard to motor-voter registrants (i.e., 10 While the Secretary argues that we should not facially invalidate the statute based on “the allegedly ‘confusing’ implementation of Kansas law,” Aplt.’s Opening Br. at 33, or “the unique circumstances of one individual,” Aplt.’s Reply Br. at 16, he does not offer a broader challenge to the scope of the relief that the district court ordered. We do not rely here on the confusing implementation of the DPOC law or the circumstances of any individual plaintiff, instead finding that the demonstrated disenfranchisement of approximately 30,000 would-be voters demonstrates that the “broad application” of the DPOC requirement imposed an unjustified burden on “all [Kansas] voters.” Crawford, 553 U.S. at 202S03 (plurality opinion of Stevens, J.). 63 the appeal in Fish v. Schwab, No. 18-3133) because section 5 of the NVRA preempts Kansas’s DPOC requirement. In coming to that conclusion, we (1) summarize our opinion in Fish I, (2) explain why that opinion establishes the law of this case, and (3) hold that, under the Fish I framework, Kansas’s DPOC requirement is preempted by section 5 of the NVRA because Kansas failed to demonstrate substantial numbers of non-citizen voters attempted to register or vote. Thus, even apart from our constitutional holding in Bednasek v. Schwab, No. 18-3134, we affirm the district court’s injunction—as to the class of voters who sought to register under section 5 of the NVRA. 1. Preemption Framework Established in Fish I In Fish I, we addressed the fundamental question of whether Congress—through the NVRA—had utilized the authority vested in it by the Elections Clause to preempt state regulations—namely, the DPOC requirement. The Constitution’s Elections Clause states: The Times, Places and Manner of holding Elections for Senators and Representatives, shall be prescribed in each State by the Legislature thereof; but the Congress may at any time by Law make or alter such Regulations, except as to the Places of chusing Senators. U.S. C ONST . art. I, § 4, cl. 1. This provision makes clear that while states must set the “Times, Places and Manner” of their elections, “Congress can step in, either making its own regulations that wholly displace state regulations or else 64 modifying existing state regulations.” Fish I, 840 F.3d at 724; see Arizona v. Inter Tribal Council of Ariz., Inc., 570 U.S. 1, 8 (2013) (“Upon the States [the Elections Clause] imposes the duty (‘shall be prescribed’) to prescribe the time, place, and manner of electing Representatives and Senators; upon Congress it confers the power to alter those regulations or supplant them altogether.”). In other words, “[t]he Clause is a default provision; it invests the States with responsibility for the mechanics of congressional elections, but only so far as Congress declines to preempt state legislative choices.” Foster v. Love, 522 U.S. 67, 69 (1997) (citation omitted); accord Fish I, 840 F.3d at 725S26. In Fish I, we explained that, in order to answer the fundamental preemption question, “the Elections Clause requires that we straightforwardly and naturally read the federal and state provisions in question as though part of a unitary system of federal election regulation but with federal law prevailing over state law where conflicts arise.” Fish I, 840 F.3d at 729; accord Gonzalez v. Arizona, 677 F.3d 383, 394 (9th Cir. 2012) (en banc), aff’d sub nom. Inter Tribal, 570 U.S. 1 (2013). We then turned to the relevant portions of Kansas’s DPOC requirement and the NVRA. Kansas’s DPOC requirement provides: “The county election officer or secretary of state’s office shall accept any completed application for registration, but an applicant shall not be registered until the applicant has provided satisfactory evidence of United States citizenship.” Kan. Stat. Ann. § 25–2309(l). 65 The statute then “enumerates thirteen forms of documentation, including a birth certificate and a passport, that meet this requirement.” Fish I, 840 F.3d at 732. We then turned to the relevant portion of section 5 of the NVRA, which states: (2) The voter registration application portion of an application for a State motor vehicle driver’s license— (A) may not require any information that duplicates information required in the driver’s license portion of the form (other than a second signature or other information necessary under subparagraph (C)); (B) may require only the minimum amount of information necessary to— (i) prevent duplicate voter registrations; and (ii) enable State election officials to assess the eligibility of the applicant and to administer voter registration and other parts of the election process; (C) shall include a statement that— (i) states each eligibility requirement (including citizenship); (ii) contains an attestation that the applicant meets each such requirement; and (iii) requires the signature of the applicant, under penalty of perjury . . . . 52 U.S.C. § 20504(c)(2)(A)S(C). We read subparagraph (B) as “restricting states’ discretion in creating their own DMV voter-registration forms” by establishing what we referred to as the “minimum-information principle.” Fish I, 840 F.3d at 733. The minimum- information principle “establishes a ceiling on what information the states can require” on the motor-voter form. Id.; see id. (“Under NVRA section 5, a state 66 motor voter form ‘may require only the minimum amount of information necessary’ for state officials to carry out their eligibility-assessment and registration duties.” (quoting 52 U.S.C. § 20504(c)(2)(B))). This principle “calls on states to include the least possible amount of information necessary on the motor voter form.” Id. at 736. We also noted subparagraph (C)’s command to “states to list qualifications and also to require applicants to attest that they meet them and to sign the attestation under penalty of perjury.” Id. Subparagraph (C) “mandates that states include an attestation requirement” on the motor-voter form. Id. Reading subparagraphs (B) and (C) harmoniously, see FDA v. Brown & Williamson Tobacco Corp., 529 U.S. 120, 133 (2000) (“A court must . . . interpret the statute ‘as a symmetrical and coherent regulatory scheme,’ and ‘fit, if possible, all parts into an harmonious whole.’” (citations omitted) (first quoting Gustafson v. Alloyd Co., 513 U.S. 561, 569 (1995); then quoting FTC v. Mandel Bros., Inc., 359 U.S. 385, 389 (1959))), we concluded that “section 5 is reasonably read to establish [subparagraph (C)’s] attestation requirement as the presumptive minimum amount of information necessary for a state to carry out its eligibility-assessment and registration duties [under subparagraph (B)],” Fish I, 840 F.3d at 737; id. at 737S38 (inferring “from the statutory structure that Congress contemplated that the attestation requirement would be regularly used 67 and would typically constitute the minimum amount of information necessary for state officials to carry out their eligibility-assessment and registration duties”). While subparagraph (C)’s attestation requirement provides the presumptive minimum amount of information necessary for a state to carry out its subparagraph (B) duties, we “recognize[d] that in a given case [the attestation requirement] may not be sufficient for a state to carry out its eligibility- assessment and registration duties.” Id. at 737. “[W]hether the attestation requirement actually satisfies the minimum- information principle in a given case turns on the factual question of whether the attestation requirement is sufficient for a state to carry out these duties.” Id. at 738. Thus, “in order for a state advocating for a DPOC regime to rebut the presumption that the attestation requirement is the minimum information necessary for it to carry out its eligibility-assessment and registration duties, it must make a factual showing that the attestation requirement is insufficient for these purposes.” Id. “More specifically, in order to rebut the presumption as it relates to the citizenship criterion, we interpret[ed] the NVRA as obliging a state to show that ‘a substantial number of noncitizens have successfully registered’ notwithstanding the attestation requirement.” Id. at 739 (quoting EAC, 772 F.3d at 1198). And so we observed that “if Kansas fails to rebut this presumption that attends the attestation regime, then DPOC necessarily requires more information 68 than federal law presumes necessary for state officials to meet their eligibility-assessment and registration duties (that is, the attestation requirement).” Id. In that circumstance, “Kansas’s DPOC law would be preempted.” Id. Finally, we then applied this preemption framework and concluded that the Secretary had failed to demonstrate that a substantial number of noncitizens had successfully registered. Id. at 747–48. In particular, the Secretary had only shown that between 2003 and 2013 “thirty noncitizens registered to vote.” Id. at 746. “These numbers [fell] well short of the showing necessary to rebut the presumption that attestation constitutes the minimum amount of information necessary for Kansas to carry out its eligibility-assessment and registration duties.” Id. at 747. Thus, we concluded that the plaintiffs’ challenge was likely to succeed on the merits. Id. at 750S51. We then concluded that the district court did not err in holding that the remaining preliminary-injunction factors also favored a preliminary injunction, and thus affirmed the district court’s grant of a preliminary injunction. Id. at 751S56. However, we acknowledged that if, on remand, “evidence comes to light that a substantial number of noncitizens have registered to vote in Kansas during a relevant time period, inquiry into whether DPOC is the minimum amount of information necessary for Kansas to carry out its eligibility-assessment and registration duties would then be appropriate.” Id. 69 at 750–51. In other words, while the Secretary had not demonstrated that a substantial number of noncitizens had registered to vote at the preliminary- injunction stage of the litigation, we left open the possibility that he would be able to make that showing at a trial on the merits. As noted, however, the district court subsequently concluded that the Secretary failed to make that showing. 2. The Fish I Framework Governs We conclude that the Fish I framework governs our inquiry into whether section 5 of the NVRA preempts Kansas’s DPOC requirement. However, the Secretary argues against this conclusion, claiming that Fish I “is not binding on this panel” and that we should reevaluate the Fish I framework. Aplt.’s Opening Br. at 18, 41S42. Thus, before we turn to evaluating whether the Secretary has demonstrated that a substantial number of noncitizens registered to vote, we explain why, under the law-of-the-case doctrine, Fish I’s legal determinations are the law of the case. The law-of-the-case doctrine provides that, “when a court rules on an issue of law, the ruling ‘should continue to govern the same issues in subsequent stages in the same case.’” Bishop v. Smith, 760 F.3d 1070, 1082 (10th Cir. 2014) (quoting United States v. Graham, 704 F.3d 1275, 1278 (10th Cir. 2013)); accord Arizona v. California, 460 U.S. 605, 618 (1983). “Under this doctrine, ‘the decision of the appellate court establishes the law of the case and ordinarily will 70 be followed by both the trial court on remand and the appellate court in any subsequent appeal.’” Cressman v. Thompson, 798 F.3d 938, 946 (10th Cir. 2015) (quoting Zinna v. Congrove, 755 F.3d 1177, 1182 (10th Cir. 2014)). The doctrine serves “important” functions. Entek GRB, LLC v. Stull Ranches, LLC, 840 F.3d 1239, 1240 (10th Cir. 2016). “Without something like it, an adverse judicial decision would become little more than an invitation to take a mulligan, encouraging lawyers and litigants alike to believe that if at first you don’t succeed, just try again.” Id.; see also 18B C HARLES A LAN WRIGHT ET AL ., F EDERAL P RACTICE & P ROCEDURE § 4478, Westlaw (2d ed., database updated Apr. 2020) (“Law-of-the-case rules have developed to maintain consistency and avoid reconsideration of matters once decided during the course of a single continuing lawsuit.”). However, despite its importance, “the decision whether to apply law of the case doctrine remains a matter of judicial discretion.” Entek GRB, 840 F.3d at 1242; see Rimbert v. Eli Lilly & Co., 647 F.3d 1247, 1251 (10th Cir. 2011) (“This doctrine is designed to promote finality and prevent re-litigation of previously decided issues, but does not serve to limit a court’s power.”). “Although the law of the case doctrine is not a limit on our power, nor ‘an inexorable command,’” United States v. Irving, 665 F.3d 1184, 1192 (10th Cir. 2011) (citations omitted) (quoting United States v. Alvarez, 142 F.3d 1243, 1247 71 (10th Cir. 1998)), it “is subject to very limited exceptions,” id. In particular, “[w]e will only deviate from the law of the case ‘(1) when the evidence in a subsequent trial is substantially different; (2) when controlling authority has subsequently made a contrary decision of the law applicable to such issues; or (3) when the decision was clearly erroneous and would work a manifest injustice.’” Cressman, 798 F.3d at 946 (quoting Irving, 665 F.3d at 1192 n.12); accord United States v. Trent, 884 F.3d 985, 995 (10th Cir. 2018). In light of all this, it is clear that, at least ordinarily, a panel of our court would follow any rulings made in prior appeals in the same case; thus, we ordinarily would follow Fish I’s framework. But the Secretary argues that “Fish I does not bind this panel” because that opinion was considering a preliminary injunction while we are now considering the merits. Aplt.’s Opening Br. at 41S42 (bold-faced font omitted). His argument relies centrally on Homans v. City of Albuquerque, 366 F.3d 900 (10th Cir. 2004). In that case, a two-judge motions panel of this court granted an emergency motion for a preliminary injunction that enjoined the relevant law pending review of the merits. Id. at 903. “In doing so, the motions panel held that [the plaintiff] established a likelihood of success on the merits” on his claim that the relevant statute was unconstitutional. Id. On remand, the district court stated its own view that the statute was constitutional, but it nevertheless—under its perceived duty—entered judgment enjoining 72 enforcement of the statute based on the motions panel’s ruling. Id. In an appeal of that order, i.e., a second appeal, we held that the motions panel’s preliminary ruling was not the law of the case and thus did not bind the district court’s or our subsequent merits determinations. Id. at 904–05. This was because “the two judge panel decision of our court constituted an interlocutory ruling, and its holding was limited to the conclusion that [the plaintiff] had shown a likelihood of success on the merits of his claim.” Id. at 904. Furthermore, “[t]o the extent that any language in [the motions-panel’s decision] c[ould] be read as an assessment of the actual merits of [the plaintiff’s] claim, as opposed to his likelihood of success on the merits, such language [was] dicta” and “not subject to the law of the case doctrine.” Id. at 904 n.5; see Univ. of Tex. v. Camenisch, 451 U.S. 390, 395 (1981) (“[T]he findings of fact and conclusions of law made by a court granting a preliminary injunction are not binding at trial on the merits.” (citations omitted)). The Secretary argues, based on Homans and Camenisch, that Fish I is not entitled to law-of-the-case effect. We disagree. As Camenisch indicates, the normal rule is that “[r]ulings—predictions—as to the likely outcome on the merits made for preliminary injunction purposes do not ordinarily establish the law of the case, whether the ruling is made by a trial court or by an appellate court.” 18B WRIGHT ET AL ., supra, at § 4478.5 (footnotes omitted); cf. Attorney Gen. of 73 Okla. v. Tyson Foods, Inc., 565 F.3d 769, 776 (10th Cir. 2009) (“In affirming the district court’s denial of a preliminary injunction, we do not address the underlying merits of Oklahoma’s ultimate claims at trial.”). This rule exists because “the court of appeals must often consider such preliminary relief without the benefit of a fully developed record and often on briefing and argument abbreviated or eliminated by time considerations.” Sherley v. Sebelius, 689 F.3d 776, 782 (D.C. Cir. 2012). And so our sister circuits and a leading treatise agree that “[a] fully considered appellate ruling on an issue of law made on a preliminary injunction appeal . . . become[s] the law of the case for further proceedings in the trial court on remand and in any subsequent appeal.” 18B WRIGHT ET AL ., supra, at § 4478.5 (emphasis added) (footnote omitted); see Rodriguez v. Robbins, 804 F.3d 1060, 1080S81 (9th Cir. 2015) (explaining that conclusions on pure issues of law made during appeal of preliminary injunction constitute law of the case), rev’d on other grounds sub nom. Jennings v. Rodriguez, 138 S. Ct. 830 (2018); Howe v. City of Akron, 801 F.3d 718, 740 (6th Cir. 2015) (collecting cases from other circuits reaching the conclusion that “when a court reviewing the propriety of a preliminary injunction issues a fully considered ruling on an issue of law with the benefit of a fully developed record, then the conclusions with respect to the likelihood of success on the merits are the law of the case in any subsequent appeal”). 74 Guided by this authority, we note that, in Homans, the motions panel granted the emergency motion two days after it was made, and so it made sense there that we would not afford law-of-the-case effect to the motion panel’s decision—made, as it was, under (necessarily) severe time constraints. 366 F.3d at 903. Here, however, the Fish I panel was able to consider the issue fully and issue a lengthy opinion discussing pure issues of law. See Sherley, 689 F.3d at 783 (“The time constraints and limited record available to the court in those cases are not present here. We therefore follow the other circuits in concluding that the exception [to the law-of-the-case doctrine] is not present either.”). “[W]here the earlier ruling, though on preliminary-injunction review, was established in a definitive, fully considered legal decision based on a fully developed factual record and a decisionmaking process that included full briefing and argument without unusual time constraints, why should we not follow the usual law-of-the-case jurisprudence?” Id. at 782. We, like our sister circuits, think that it makes eminent sense to apply the law-of-the-case doctrine in these circumstances. Arguing against this conclusion, the Secretary also cites to the concurring opinion in Prairie Band Potawatomi Nation v. Wagnon, 402 F.3d 1015, 1029 (10th Cir.) (McConnell, J., concurring), vacated, 546 U.S. 1072 (2005). But the majority opinion in that case adopted the same position set out above—namely 75 that an opinion that robustly addresses legal issues at the preliminary-injunction stage of the litigation will provide “the district court with the appropriate legal framework within which to view th[e] case.” Id. at 1020; see id. at 1021 (“In so ruling [on an issue in the earlier appeal of the preliminary injunction], it would appear that [the prior ruling] is the law of this case . . . .”). In responding to the concurring opinion cited by the Secretary, “we specifically recognized that the principles of law of the case are flexible,” and so “while it [wa]s true that we [we]re not bound by Prairie Band I, it [wa]s not true that ‘the law of the case doctrine does not apply’ merely because Prairie Band I dealt with a preliminary injunction.” Id. at 1026 (quoting id. at 1030 (McConnell, J., concurring)). We think this gets it exactly right: because the Fish I panel was able to consider the issue fully and issue a lengthy opinion discussing pure issues of law, we conclude that the law-of-the-case doctrine applies to Fish I’s legal conclusions. Of course, this doctrine “does not serve to limit a court’s power.” Rimbert, 647 F.3d at 1251. And exceptions to the doctrine’s effect exist when there is new and different evidence, an intervening change in controlling authority, or the prior ruling was clearly erroneous and would work a manifest injustice. Id. But it is undisputed that none of these exceptions applies here, and thus we adhere to the “thorough and sound” discussion of the applicable 76 preemption framework found in our prior opinion, Fish I. Prairie Band Potawatomi Nation, 402 F.3d at 1026. In sum, we conclude that Fish I’s preemption framework is the law of the case. 3. Whether the Secretary Presented Sufficient Evidence to Satisfy Fish I The Secretary argues that he satisfied the Fish I framework on remand. But the district court’s factual findings undermine this argument, and the Secretary’s remaining arguments turn on his already-rejected view that Kansas must be afforded “sufficient discretion [for it] to determine whether the problem of unqualified voters becoming registered is ‘substantial.’” Aplt.’s Opening Br. at 57. We thus conclude that the Secretary has failed to show that a substantial number of noncitizens registered to vote. In Fish I, we held that “to overcome the presumption that attestation constitutes the minimum amount of information necessary for a state to carry out its eligibility-assessment and registration duties, the state must show that a substantial number of noncitizens have successfully registered to vote under the attestation requirement.” 840 F.3d at 739. At that stage, “[t]he district court found that between 2003 and the effective date of Kansas’s DPOC law in 2013, only thirty noncitizens registered to vote,” and we concluded that “[t]hese numbers [fell] well short of the showing necessary to rebut the presumption that 77 attestation constitutes the minimum amount of information necessary for Kansas to carry out its eligibility-assessment and registration duties.” Id. at 746S47. But we acknowledged that “[i]f evidence comes to light that a substantial number of noncitizens have registered to vote in Kansas during a relevant time period, inquiry into whether DPOC is the minimum amount of information necessary for Kansas to carry out its eligibility-assessment and registration duties would then be appropriate.” Id. at 750S51. Thus, to overcome the presumption that the attestation requirement satisfies the minimum-information principle, the Secretary needed to provide greater evidence of noncitizens registering to vote. But, at the trial on the merits, the district court found that only 39 noncitizens “successfully registered to vote despite the attestation requirement.” Aplt.’s App., Vol. 47, at 11472; see id. at 11508. Yet recall that, even as to those 39 noncitizens who appear on the Kansas voter rolls, the district court effectively found that “administrative anomalies” could account for the presence of many—or perhaps even most—of them there. Id. at 11520. In other words, even the figure of 39 registered noncitizens could be more the product of such anomalies than of the voluntary registration actions of noncitizens in the face of the attestation requirement. But, accepting that figure at face value, the confirmed noncitizens who successfully registered to vote from 1999 to 2013 was equivalent to less than three noncitizen voters a year. In Fish I we concluded that the 30 noncitizens 78 who had registered to vote between 2003 and 2013—which also equated to “no more than three per year”—were “well short” of “substantial.” 840 F.3d at 746–47. Following Fish I’s guidance, we reach a similar conclusion here. Specifically, we conclude that the addition of nine voters spread over four more years means that the Secretary has still failed to demonstrate that substantial numbers of noncitizens successfully registered to vote notwithstanding the attestation requirement. Thus, we conclude that the Secretary has failed to rebut the presumption that the attestation requirement satisfies the minimum-information principle. The Secretary does not argue that the district court’s factual findings concerning the number of noncitizens who registered to vote were clearly erroneous. Nevertheless, he references statements made in the legislative record by certain legislators who believed noncitizens had registered to vote, cites a letter from a court clerk asserting that employees of a hog farm “were transported to [the county clerk’s] office by their employer to register to vote” and that “some of these employees felt they were pressured to register even though they may not be legal,” Aplt.’s App., Vol. 30, at 7668, and notes possible instances of noncitizens registering to vote, including one where a citizen told Kansas officials that she had voted before becoming a citizen, see Aplt.’s Opening Br. at 57S58 (citing Aplt.’s App., Vol. 38, at 9460S65, 9522S23). But because the Secretary 79 does not directly contest the district court’s factual findings about how many noncitizens registered to vote, he may not wage a guerilla war on the district court’s factual findings through these ad hoc, anecdotal references to other purported incidents of noncitizen registration. This evidence does not establish that substantial numbers of noncitizens registered to vote. The Secretary also cites to two cases from the Seventh Circuit that identified individual noncitizens who registered to vote using motor-voter forms, see Kimani v. Holder, 695 F.3d 666, 668, 671 (7th Cir. 2012); Keathley v. Holder, 696 F.3d 644, 645 (7th Cir. 2012), and provides a citation to an internet story about another noncitizen whom Illinois officials registered to vote, see Aplt.’s Opening Br. at 58S60. These anecdotes, even were we to consider them, do not establish that “substantial” numbers of noncitizens registered to vote in Kansas during a relevant time period and thus are not pertinent to the registration of noncitizen voters in Kansas. See Aplee.’s Resp. Br. at 37. This showing thus cannot satisfy the Secretary’s burden. The Secretary falls back on arguing that “[t]he elected representatives of the people of the State of Kansas determined that requiring proof of citizenship as a condition of voter eligibility was a permissible response to the threat posed by voter fraud.” Aplt.’s Opening Br. at 57. But this is just an argument that states should be able to determine what § 20504(c)(2)(B)’s term “necessary” means—an 80 argument that we expressly rejected in Fish I. See 840 F.3d at 743 (noting that “the Supreme Court in Inter Tribal rejected such an understanding of federal election regulation and confirmed that the NVRA’s plain language evinces Congress’s intent to restrain the regulatory discretion of the states over federal elections, not to give them free rein”). Moreover, the Secretary relatedly argues that the attestation requirement does not provide the minimum information necessary if even one noncitizen registers to vote because even a small number of noncitizens registered to vote could be “determinative” in certain close elections. Aplt.’s Opening Br. at 60 (pointing to various close elections in Kansas between 2000 and 2016, including 33 elections decided by fewer than 100 votes and one that was tied (citing Aplt.’s App., Vol. 27, at 6938; id., Vol. 39, at 9765S69)); see Aplt.’s Reply Br. at 33S34. But we rejected this argument in Fish I: “The NVRA does not require the least amount of information necessary to prevent even a single noncitizen from voting.” 840 F.3d at 748. As we explained there, “Congress adopted the NVRA to ensure that whatever else the states do, ‘simple means of registering to vote in federal elections will be available.’” Id. (quoting Inter Tribal, 570 U.S. at 12). “This purpose would be thwarted if a single noncitizen’s registration would be sufficient to cause the rejection of the attestation regime.” Id. The Secretary provides no reason here to deviate from that conclusion, and, for that reason 81 alone, we could reject his argument. And, furthermore, if the Secretary is correct that Kansas’s recent history is sprinkled with some hotly contested, close elections such that he reasonably could have an especially keen interest in ensuring that every proper vote counts, we are hard-pressed to see how that interest is furthered by the DPOC law—a law that undisputedly has disenfranchised approximately 30,000 would-be Kansas voters who presumably would otherwise have been eligible to vote in such close elections. Indeed, the DPOC law would appear to undercut, rather than further, the Secretary’s professed interest in ensuring that every proper vote counts. Finally, the Secretary argues that “even if there were no widespread problem of voter fraud in Kansas, the Supreme Court has recognized that the States can proactively fight against the prospect of fraud.” Aplt.’s Opening Br. at 61. To make this point, he cites to Justice Stevens’s discussion in Crawford of Indiana’s undisputed interest in preventing voter fraud even though “[t]he record contain[ed] no evidence of any such fraud actually occurring in Indiana at any time in its history.” 553 U.S. at 194. But this interest was asserted in an entirely different context: Justice Stevens was analyzing a constitutional challenge in a case where the relevant law only imposed a limited burden on voters. Id. at 202S03. He said nothing about the evidentiary burden required to displace the 82 statutory presumption created by section 5 of the NVRA and thus the citation is inapt. In short, we conclude that the Secretary has failed to show that a substantial number of noncitizens successfully registered to vote. 11 * * * The Secretary has failed to show that a substantial number of noncitizens have successfully registered in Kansas notwithstanding section 5 of the NVRA’s attestation requirement. Thus, the DPOC requirement necessarily requires more information than federal law presumes necessary for state officials to meet their eligibility-assessment and registration duties. And so we conclude that Kansas’s DPOC law is preempted by section 5 of the NVRA. We uphold the district court’s entry of a permanent injunction against the enforcement of the DPOC requirement as to the those voters who sought to register under section 5 of the NVRA. IV. Conclusion In Bednasek, No. 18-3134, we conclude the DPOC requirement unconstitutionally burdens the right to vote and thus AFFIRM the district court’s 11 We do not reach the district court’s separate analysis concerning whether—if the number of noncitizen voters here did count as substantial—the DPOC requirement nevertheless satisfied the minimum-information principle. 83 injunction. Likewise, in Fish, No. 18-3133, we AFFIRM the district court’s injunction because section 5 of the NVRA preempts the DPOC requirement. 84