The Yale Law Journal


Building an Umbrella in a Rainstorm: The New Vote Denial Litigation Since Shelby County

08 Feb 2018

abstract. Since the Supreme Courts 2013 decision in Shelby County v. Holder, a two-part test for vote denial liability under Section 2 of the Voting Rights Act (VRA) has emerged. This Essay traces the development of that test and assesses its application in the courts. All five courts of appeals that have considered this issue nominally agree on the same two-part test under Section 2. But their approaches differ in some respects. This Essay describes tension among the courts of appeals as to the necessity of (1) statistical evidence regarding the effect of a voting practice on voter turnout, and (2) evidence concerning discriminatory intent on the part of a state actor. In particular, the Seventh Circuit has suggested that evidence of both factors is required to establish a Section 2 violation for vote denial, making it an outlier. The Essay ends by concluding that the majority of circuits are correct not to have adopted these dual evidentiary requirements for Section 2 liability.


In the run-up to the 2012 presidential election, nineteen states passed laws making it more difficult to register to vote or cast a ballot.1 Litigation under Section 5 of the Voting Rights Act2 blocked or mitigated the effects of several of these laws, including voter ID laws in Texas3 and South Carolina,4 as well as early voting cutbacks in Florida.5 The Supreme Court’s 2013 decision in Shelby County v. Holder,6 however, largely immobilized the federal preclearance protections of Section 5, which required certain states and local jurisdictions to obtain federal approval before making any changes to their voting laws. Justice Ginsburg famously warned that “[t]hrowing out preclearance when it has worked and is continuing to work to stop discriminatory changes [to voting laws] is like throwing away your umbrella in a rainstorm because you are not getting wet.”7 And, sure enough, the rain came, as fifteen states passed or implemented new restrictions on voting leading up to the 2014 midterms.8 Voting rights litigators were thus confronted with the challenge of developing a new legal strategy under a different VRA provision, Section 2.9

The voting rights community was writing on something of a blank slate due to the absence of clear case law under Section 2 concerning restrictions on voting. After the passage of the VRA in 1965, outright disfranchisement through devices like literacy tests became far less common. But while African-American registration rates in the South skyrocketed immediately, many jurisdictions continued to deny African Americans a meaningful voice by adopting electoral arrangements that canceled out or minimized African-American voting strength. One particularly prevalent practice was at-large elections, which give the majority of a jurisdiction the ability to elect all of the jurisdiction’s representatives. More familiar single-member districting arrangements, on the other hand, can afford a geographically-concentrated minority group the opportunity to elect their preferred candidates from a particular district.10 As the Supreme Court explained in Shaw v. Reno, “it soon became apparent that guaranteeing equal access to the polls would not suffice to root out other racially discriminatory voting practices.”11

Thus, the VRA—which was originally conceived primarily as a tool to address “vote denial” schemes that restrict a person’s ability to “vot[e] or hav[e] their votes counted”—was largely repurposed to combat “vote dilution,” i.e. “practices that diminish minorities’ political influence in places where they are allowed to vote.”12 In 1982, Section 2 was amended to incorporate a discriminatory results standard, with vote dilution primarily in mind,13 and over the next two-plus decades, the vast majority of Section 2 litigation occurred in the vote dilution context.14 The bulk of Section 2 case law, therefore, is largely concerned with issues involving the drawing of district lines—which affects the weight of votes that are cast—rather than an individual’s ability to register or cast a vote in the first place.

In order to contend with the resurgence of registration and ballot restrictions sweeping the country after Shelby County was decided, voting rights litigators were faced with the formidable task of establishing a clear and robust test for vote denial liability under Section 2, and litigated a flurry of new vote denial cases under Section 2 in the 2014 and 2016 election cycles.15 Ultimately, five courts of appeals rendered decisions in Section 2 vote denial cases, involving restrictions such as early voting cutbacks in Ohio;16 voter identification requirements in Wisconsin,17 Texas,18 and Virginia;19 restrictions on absentee ballots in Arizona;20 the elimination of straight-ticket voting in Michigan;21 and a sweeping law in North Carolina that imposed a broad range of restrictions on early voting, registration, and identification requirements.22

Although outcomes have been mixed—some restrictions, like Texas’s voter ID law, were struck down23 while others, like Wisconsin’s voter ID law, have thus far been upheld24—and litigation in several of these cases is ongoing, plaintiffs have thus far been largely successful in establishing a consensus around the appropriate legal standard. As I explain below, all five courts have (at least nominally) embraced the same two-part standard for Section 2 vote denial liability. There are, however, some differences in how courts have applied that test. In particular, the Seventh Circuit has suggested that evidence of reduced turnout25 and discriminatory intent on the part of state actors26 may be necessary to establish Section 2 vote denial liability. As this Essay explains, most circuits that have considered this issue have rightly eschewed such requirements, which are inconsistent with the text and purpose of Section 2.

I. the development and adoption of the two-part test for vote denial liability under section 2 of the vra

In a 2014 Article building on the work of Daniel Tokaji27 and Janai Nelson,28 I described various possible tests for vote denial liability under Section 2’s results standard.29Among them was a two-part test, “requir[ing] that a plaintiff show [1] a disparate impact [of a voting restriction on voters of color] plus [2] some of the Senate Factors.”30 These factors are a set of nine nonexclusive factors set forth in the Senate Report to the 1982 Amendments to the VRA, which were intended to guide courts applying the then-new Section 2 results standard,31 and which the Supreme Court recognized in Thornburg v. Gingles32 as relevant to assessing a discriminatory results claim under Section 2. The Senate Factors are:

(1)   the history of voting-related discrimination in the State or political subdivision;

(2)   the extent to which voting in the elections of the State or political subdivision is racially polarized;

(3)   the extent to which the State or political subdivision has used voting practices or procedures that tend to enhance the opportunity for discrimination against the minority group;

(4)   the exclusion of members of the minority group from candidate slating processes;

(5)   the extent to which minority group members bear the effects of past discrimination in areas such as education, employment, and health, which hinder their ability to participate effectively in the political process;

(6)  the use of overt or subtle racial appeals in political campaigns;

(7)   the extent to which members of the minority group have been elected to public office in the jurisdiction;

(8)   whether elected officials are unresponsive to the particularized needs of the members of the minority group; and

(9)  whether the policy underlying the State’s or the political subdivision’s use of the contested practice or structure is tenuous.33

Each of the prongs from this proposed two-part test for Section 2 vote denial liability was borrowed from familiar contexts. The first part of the test—disparate impact—would “draw[] on an existing framework with which the courts have substantial experience in other contexts,”34 such as employment discrimination under Title VII of the 1964 Civil Rights Act35 or housing discrimination under the 1968 Fair Housing Act.36 Under this framework, plaintiffs challenging a facially neutral practice can state a prima facie case for unlawful discrimination based on the practice’s disparate impact on people of color. The second part of the test—the presence of a racialized political context, such as a history of official voting discrimination, racially polarized voting, and the effects of discrimination in socioeconomic areas of life that inhibit participation by voters of color—would draw directly on the factors for vote dilution liability, identified in the legislative history of Section 2’s results standard and canonized by the Supreme Court in Gingles.37

To be sure, a test built on the first prong alone would have been simpler, and would have more closely mirrored the discriminatory effects standards for employment and housing discrimination, which generally do not require plaintiffs to present evidence about the broader socio-historic context in order to state a prima facie case. Requiring evidence of at least some of these factors would thus complicate the litigation by requiring plaintiffs to adduce evidence beyond the direct impact of the challenged practice.

But I view that added burden as a feature rather than a bug. By requiring vote denial plaintiffs to demonstrate the presence of factors “function[ing] as headwinds that prevent minority voters from participating equally in the political process,” the second prong “limit[s] liability only to claims where a challenged law has a particularly burdensome racial effect.”38 In this sense, a voting restriction violates Section 2 when its disparate impact is not a mere statistical happenstance unrelated to race, but rather is intimately connected to and perpetuates a broader context of racial political exclusion in the jurisdiction. The second prong would thus

prevent Section 2’s results standard from morphing into a general prohibition on any election law that happens to have a racially disparate result, which is how some have caricatured it. Rather, Section 2 would prohibit only those laws that disproportionately burden minority voters where there is a preexisting backdrop of inequality in the political process. The disparate result is truly “discriminatory,” then, in the sense that it enhances and exacerbates existing patterns of inequality.39

In the first round of post-Shelby County vote denial litigation leading up to the 2014 midterms, the ACLU represented clients arguing for the adoption of this two-part test for vote denial liability under Section 2 in three cases challenging Wisconsin’s voter ID law;40 the elimination of same-day registration and other provisions in North Carolina;41 and the elimination of weekend, evening, and same-day registration early voting opportunities in Ohio.42 Other organizations—including the NAACP Legal Defense Fund, the Southern Coalition for Social Justice, the Lawyers Committee for Civil Rights under Law, the Brennan Center, and the Advancement Project—also represented plaintiffs in these and other Section 2 vote denial cases.

The first court of appeals to render a decision in one of these cases was the Sixth Circuit, in the Ohio early voting case.43 In its decision, the Sixth Circuit articulated a two-part test for vote denial liability as follows:

First, as the text of Section 2(b) indicates, the challenged “standard, practice, or procedure” must impose a discriminatory burden on members of a protected class, meaning that members of the protected class “have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice.” . . .

Second, the Supreme Court has indicated that that burden must in part be caused by or linked to “social and historical conditions” that have or currently produce discrimination against members of the protected class.44

With respect to the first prong, the Sixth Circuit affirmed the district court’s reliance on expert statistical evidence “demonstrating that African Americans vote [early in-person] at higher rates than other groups, including on the eliminated [early in-person] voting days,”45 and that some voters would be “significantly burdened” by the elimination of these particular early voting opportunities.46

With respect to the second prong, the Sixth Circuit turned to the Senate Factors, and found that the greater reliance by African American voters in Ohio on the eliminated early voting opportunities stemmed from the fact that “African Americans are more likely to be of lower-socioeconomic status in Ohio.”47 More specifically, the court pointed to “stark and persistent racial inequalities . . . [in] work, housing, education and health,” which in turn stem from “both historical and contemporary discriminatory practices.”48

The Sixth Circuit thus affirmed the district court’s preliminary injunction against Ohio’s planned early voting cutbacks. We had successfully obtained the first court of appeals decision articulating a clear standard for vote denial liability under Section 2.

Unfortunately, the Supreme Court, in a 5-4 ruling issued eighteen hours before early voting in Ohio was set to begin, stayed the Sixth Circuit’s preliminary injunction ruling.49 Because the stay lasted ninety days, and the preliminary injunction applied only to the then-imminent 2014 general election,50 the Sixth Circuit’s decision became moot at the moment that the Supreme Court issued the stay order.51 The Sixth Circuit then vacated its own decision as such.52 The case ultimately settled, with Ohio restoring some but not all of the eliminated early voting opportunities.53

But the two-part test for Section 2 vote denial liability lived on. Later that year, the Fourth Circuit “agree[d]” with the Sixth Circuit’s two-part test, and adopted it word-for-word, in a decision enjoining various voting restrictions in the North Carolina litigation54 Meanwhile, the Seventh Circuit issued a decision upholding Wisconsin’s voter ID law, but also noted its “agree[ment]” with the Sixth Circuit’s two-part test test.55 Two years later, in the run-up to the 2016 general election, the en banc Fifth Circuit adopted the two-part test in striking down Texas’s strict voter ID law.56 The Sixth Circuit also readopted the test in a series of decisions—first, a motions panel decision that declined to stay an injunction against Michigan’s elimination of straight-ticket voting,57 and later, a merits panel decision upholding the remaining early voting cutbacks in Ohio.58 And finally, the en banc Ninth Circuit adopted the same test in a decision granting an injunction pending appeal against an Arizona law that “criminalize[d] the collection, by persons other than the voter, of legitimately cast [absentee] ballots.”59

As the summary above indicates, plaintiffs have had mixed outcomes in these cases. Courts have struck down some voting restrictions, upheld others, and issued stay decisions that cut both ways. But with respect to the legal standard, there is general agreement among the courts of appeals as to the broad strokes of the two-part test.

The Seventh Circuit, however, stands as something of an outlier. Although the Seventh Circuit’s decision upholding Wisconsin’s voter ID law noted the court’s “agree[ment]” with the two-part test, the court also expressed “skeptic[ism]” about it,60 and has interpreted and applied this test in ways inconsistent with the rulings of other circuits.

II. two open questions about the standard for vote denial liability under section 2

Despite apparent unanimity as to the applicable legal test, the circuits differ somewhat as to the evidentiary showing needed to satisfy it. In this Part, I address two areas of disagreement: first, whether evidence concerning the effect of a voting restriction on voter turnout is necessary to establish a discriminatory burden (the first prong); and second, whether evidence of intentional state-sponsored discrimination is one of the “social and historical conditions” that a plaintiff must establish (the second prong). A majority of circuits has answered these questions in the negative. But the Seventh Circuit is an outlier on these points, and it appears to have misconstrued the text and purpose of Section 2.

A. Is Turnout Evidence Necessary To Establish a Discriminatory Burden?

The first area of tension among the courts of appeals is whether, for the purposes of the first prong of the Section 2 vote denial standard (a “discriminatory burden”), a plaintiff must demonstrate that a challenged practice has measurably reduced total levels of minority turnout (either in an absolute sense or relative to white turnout). The Fourth,61 Fifth,62 and Ninth63 Circuits have all expressly held that such turnout evidence is not necessary.

For example, in the Section 2 Texas voter ID case,64 the plaintiffs satisfied the first prong with two types of evidence: (1) that African American and Hispanic voters are more frequently affected by the challenged practice, because they are “more likely than their Anglo peers to lack [one of the forms of] ID” required for voting;65 and (2) that they are disproportionately represented among voters who will be particularly burdened by a strict ID requirement, i.e., poorer voters who will have substantial difficulty contending with the cost and logistics associated with obtaining an ID card in Texas. In particular, the court cited “the cost of underlying documents necessary to obtain an [ID card]” (for one plaintiff $81 for his Louisiana birth certificate); the “difficulties with delayed, nonexistent, out-of-state, or amended birth certificates;” and the logistical hurdles associated with obtaining an ID card, as some voters had to travel up to one hundred miles each way to get to the nearest ID-issuing office.66 Together, these facts showed that the Texas voter ID law made voting more difficult (even if not altogether impossible), and this burden was felt disproportionately by voters of color, which the Fifth Circuit deemed sufficient to satisfy the “discriminatory burden” prong of the two-part test for Section 2 vote denial liability.

The Sixth and Seventh Circuits, however, appear to require something more: namely, evidence concerning the effect of the challenged practice on voter turnout. For example, the Sixth Circuit has suggested that evidence of turnout disparities along racial lines is necessary for a Section 2 violation. On that basis, it rejected a challenge to early voting reductions where, after those cutbacks went into effect, “African Americans’ participation was at least equal to that of white voters.”67 Similarly, the Seventh Circuit pointed to the purported absence of evidence that voter ID laws in other states had caused a reduction in overall turnout levels as fatal to the plaintiffs’ claim in that case.68

For three reasons, the Fourth, Fifth, and Ninth Circuits present the better understanding of Section 2’s requirements. First, an insistence on turnout evidence lacks any basis in the text of Section 2. As the Fifth Circuit noted, in addition to proscribing the “denial” of the right to vote, “Section 2 also explicitly prohibit[s] abridgement of the right to vote,”69 which includes practices that make voting more burdensome, even if not altogether impossible.70 The statute also states that a violation occurs where voters of color have “less opportunity” to participate in the political process—not where they have “no” opportunity.71 As Justice Scalia once noted in an aside describing the reach of Section 2 with respect to vote denial practices, “[i]f, for example, a county permitted voter registration for only three hours one day a week, and that made it more difficult for blacks to register than whites, blacks would have less opportunity ‘to participate in the political process’ than whites, and § 2 would therefore be violated.”72 Thus, a voting restriction that burdens the exercise of the right to vote, without making it altogether impossible for voters to participate, falls within Section 2’s ambit.

Second, as a practical matter, requiring turnout evidence in Section 2 cases would make pre-enforcement challenges to discriminatory voting laws impossible, because evidence demonstrating the effect of a law on turnout is, obviously, only available after an election.73 That is particularly problematic in voting rights cases, where the injury is not compensable after the fact.74 Unlike, for example, an employee who has been denied a promotion due to discrimination, individuals disenfranchised by a discriminatory election regime cannot be compensated after the fact with money and interest; nor, generally speaking, can they get a do-over for an election tainted by an unlawful voting restriction.75 Moreover, the Supreme Court in Shelby County cited the availability of preliminary relief under Section 2 as a basis for finding that the Section 5 preclearance regime was no longer necessary.76 But preliminary relief cannot be available under Section 2 if turnout evidence—which is only available after a voting law has been implemented—is a prerequisite to a Section 2 claim.

Of course, one could concede the above but still maintain that, at least for post-implementation challenges, turnout evidence ought to be probative or even dispositive as to the “discriminatory burden” of a voting law. After all, if a law makes it harder to vote in any material way (i.e., abridges the right to vote even without denying it altogether), one might expect that burden to manifest itself in the form of a reduction in turnout.

This leads to a third problem with a requirement of turnout evidence: showing reduced turnout along racial lines resulting from a particular voting law may be impossible in many if not most cases. As an initial matter, changes in aggregate voting levels by race are difficult to measure. Only some states maintain racial data in their voting files,77 meaning that for most states there is no official record of turnout numbers broken down by race. Social scientists thus frequently attempt to measure racial turnout patterns by “estimat[ing] each voter’s racial or ethnic identity . . . us[ing] ‘surname dictionaries’ which will classify many of the last names in a voter history file to many racial or ethnic groups,”78 or by relying on statistical estimates based on survey data.79

But there are significant challenges to these estimate-based approaches. Survey sample sizes, for example, are often too small to produce reliable comparisons of turnout by race among different states. During oral argument in Shelby County, Chief Justice Roberts famously cited two purported facts about Black turnout rates: that Massachusetts has the worst Black-White turnout gap in the country, whereas Mississippi supposedly has the best.80 But as commentators noted after oral argument, the Chief Justice relied on the Census Bureau’s Current Population Survey (CPS), which—for estimates of turnout by different racial groups in individual states—is sometimes based on quite small survey samples with relatively large margins of error. For example, the estimate of Black turnout for Massachusetts of 39.3% had a margin of error of plus or minus 11.5 percentage points, meaning that it could be as high as 50.8% or as low as 27.8%.81 And thus, “if you factor in the margins of error at their extremes—with Mississippi at the low end and Massachusetts at the high end—Mississippi could have had a black voter turnout rate that was 7.5 percentage points lower than Massachusetts.”82

Notably, the CPS data is generally regarded as “high quality.”83 Courts have frequently relied on this same data in a wide range of cases.84 But Census Bureau officials have cautioned that CPS data are sometimes “not reliable for state-by-state comparisons [of turnout rates for individual racial groups] because of the high margins of error in some states.”85 The difficulties of measuring turnout rates by race in individual states thus are not the result of reliance on “bad” data, but are in some sense endemic to the tools that are available to us.86

This is particularly true with respect to voting practices that affect relatively small numbers of voters. Statistical estimates of turnout will typically have a margin of error of several percentage points—which means that any effect on turnout that is smaller than the margin of error will be impossible to detect. Assume, for example, a state in which Black turnout is estimated to be 55%, plus or minus three percentage points. If a law is thought to cause a one to two percentage point reduction in Black turnout, that reduction will be too small to measure using available statistical data.

This is no mere hypothetical. The Fourth Circuit case from North Carolina involved, among other practices, the state’s elimination of the partial counting of ballots cast by eligible voters at the incorrect precinct, known as “out-of-precinct voting.”87 Fewer than half of one percent of voters in North Carolina each year voted out-of-precinct,88 which, in absolute terms, translates to a few thousand voters. That is not insignificant, but in a state in which millions of people vote, the number of voters affected is certainly too small to detect using statistical estimates of turnout that have margins of error of a few percentage points. That does not mean that the practice has no effect on turnout. To the contrary, its effect is self-evident: in each election, several thousand voters cast their ballots at the wrong precinct—their ballots will either be counted or tossed. No statistical analysis of turnout should be necessary to confirm that rather obvious fact. But a rule to the contrary would immunize suppressive voting practices when the effects of those practices are within the margin of error for available turnout statistics.

Even setting aside these measurement issues, it is extremely difficult to ascribe causal connections between turnout shifts and any one particular factor. Voter turnout is an overdetermined phenomenon. Total turnout levels can rise or fall for many reasons,89 including: the competitiveness of elections, voter mobilization efforts and campaign tactics,90 the voting laws governing participation,91 and even the weather.92 As the Fifth Circuit explained, “[a]n election law may keep some voters from going to the polls, but in the same election, turnout by different voters might increase for some other reason. That does not mean the voters kept away were any less disenfranchised.”93 That is, high levels of public interest in a particular election may cause overall turnout to rise, even while a new voting restriction disenfranchises some voters altogether. Given the many factors that affect turnout rates, finding the signal amidst the noise is often quite difficult, if not impossible. Thus, turnout statistics should not be dispositive, because changes in overall turnout levels after the adoption of a voting law are often attributable to many factors.

At bottom, a myopic focus on turnout statistics would transform a potentially relevant piece of evidence into something dispositive. Turnout evidence may in some circumstances be probative of whether a voting restriction burdens voters—but it is not—and cannot be—the sine qua non for that inquiry.

B. Is Evidence of Discriminatory Intent Necessary To State a Violation?

The courts of appeals agree that disparate impact is only the first part of the test for vote denial liability under Section 2’s results standard: plaintiffs must also show a link between the disparate impact of the challenged law and “‘social and historical conditions’ that have or currently produce discrimination against members of the protected class.”94 But whether intentional state-based discrimination is a required element among the relevant social and historical conditions is something of an open question.

The majority rule is that such evidence of intentional discrimination by the state is unnecessary. The Fourth, Sixth, and Ninth Circuits have found a likelihood of success on the merits of Section 2 claims without relying on evidence of intentional discrimination,95 and the Fifth Circuit has expressly declined to decide whether such evidence is necessary for a Section 2 violation.96

The Fifth Circuit’s decision in the Texas voter ID case is exemplary of the range of “social and historical factors” considered by courts.97 Turning to the trial record, the Fifth Circuit found that the plaintiffs had established a number of the Senate Factors,98 including: a history of official discrimination by the State; racially polarized voting; the continuing effects of past discrimination; a dearth of minority elected officials; a lack of responsiveness by the state legislature to the minority community; and the tenuousness of the state’s rationales for the law.99 The court held that the presence of these factors demonstrated that Texas’s voter ID requirement “acted in concert with current and historical conditions of discrimination to diminish African Americans’ and Hispanics’ ability to participate in the political process.”100

The Seventh Circuit, however, has implied that evidence of state-based intentional discrimination is a necessary condition for Section 2 liability.101 The Seventh Circuit thought it irrelevant that minorities in Wisconsin disproportionately lack photo ID and the means to obtain it due in part to patterns of racial inequality in employment, housing, and transportation. The disparate impact, the court held, must be linked to intentional discrimination by the State of Wisconsin itself.102 The Seventh Circuit expressed the concern that, without a state-sponsored intentional discrimination requirement, there will be no limit to liability under Section 2. Given that racial disparities in registration and turnout rates are common, the court predicted that permitting liability without evidence of discriminatory intent would “dismantle every state’s voting apparatus.”103 The Court hypothesized, for example, that without an intent requirement, any racial disparities in voter registration rates would compel the conclusion that a state’s voter registration requirement itself violates Section 2.104 In an apparent search for a way to limit Section 2 liability, the Seventh Circuit suggested that Section 2 expresses an “equal treatment requirement” and not an “equal-outcome command,”105 apparently limiting Section 2 liability to practices that are either expressly discriminatory or facially neutral but motivated by discriminatory intent.

The Seventh Circuit’s imposition of an intent requirement is inappropriate for three reasons. The first is textual. Section 2 speaks clearly in terms of “results.” The Supreme Court has made clear that a “violation of § 2 could be established by proof of discriminatory results alone,”106 and need not be supported by direct evidence of state-sponsored intentional discrimination. As one judge on the Fifth Circuit noted, the Seventh Circuit’s suggestion that discriminatory results (without intent) are insufficient to state a claim under Section 2 is “puzzling” and “ignores” Section 2’s plain text—which plainly incorporates a discriminatory “results” standard for the precise purpose of alleviating plaintiffs of the burden of showing discriminatory intent.107

Second, the Seventh Circuit’s intent rule also misapprehends that the Senate Factors are themselves probative of discriminatory intent, as the Supreme Court has held.108 Take, for example, racially polarized voting. At first glance, the fact that voters’ preferences line up along racial divisions may seem like nothing more than purely private preferences that have nothing to do with state-sponsored discrimination.109 But the Supreme Court has held that racially polarized voting “bear[s] heavily on the issue of purposeful discrimination.”110 Where voting is racially polarized, elected officials have “an incentive for intentional discrimination in the regulation of elections,” because laws that suppress voting along racial lines help incumbents “entrench themselves.”111 Put another way, intentional racial discrimination in voting outside of a racially-polarized context is highly irrational, because it serves no purpose other than the expression of animus; but where voting is racially polarized, efforts to reduce participation by minority voters start to make more “sense” to incumbents, because they help secure their positions of power.

Again, this is not to say that racial polarization is itself intentional state-based discrimination—obviously, it is not. But it is a condition in which there is a heightened risk of such intentional state-sponsored discrimination—and thus, goes to motive.112 Where official state action has discriminatory effects, racial polarization is thus probative of the possibility that those effects are not accidental, but rather are the product of conscious efforts of incumbents to perpetuate themselves in power by targeting voters of color.113

Factors like racial polarization that are probative of discriminatory intent were included in the Section 2 results inquiry due to Congress’s concern that Section 2’s then-exclusively intent-based standard was “unnecessarily divisive because it involves charges of racism on the part of individual officials or entire communities.”114 Congress therefore adopted a results standard that would capture many of the same intentionally discriminatory practices, while relieving judges from being placed “in the difficult position of labeling their fellow public servants ‘racists.’”115 That concern holds true today—an intent requirement would likely leave voting rights violations uncured due to judicial reluctance to make findings of discriminatory intent.116 While Congress enacted Section 2’s results standard precisely to remove the burden of proving discriminatory intent, many of the factors that are relevant to the Section 2 results inquiry are in fact probative of discriminatory intent.

Third, the Seventh Circuit’s underlying fears about spiraling liability are misplaced and ignore the practical role of the Senate Factors, which, as the Supreme Court explained in Gingles, “limit[] the circumstances under which § 2 violations may be proved.”117 Some background on the Senate Factors will be useful here. The Senate Factors were derived from the Fifth Circuit’s decision in Zimmer v. McKeithen118 and the Supreme Court’s decision in White v. Regester.119 These cases addressed vote dilution claims, and sought to distinguish between those situations where minority voters’ losses at the polls could be ascribed to nonracial factors, and those that instead reflect a long-standing pattern of racial exclusion—i.e., where “the political processes leading to nomination and election were not equally open to participation by the group in question.”120

Congress’s reliance on these factors, and the Supreme Court’s incorporation of them into Section 2 jurisprudence, was intended to strike a balance between Congress’s goal of facilitating minority communities’ ability to elect their preferred candidates, and a concern expressed during hearings on the 1982 Amendments that a results standard would, in practice, translate into a requirement of proportional representation.121 That is, some members of Congress worried that under a discriminatory results standard, a Section 2 violation would be found each time minority voters were unable to elect their preferred candidates in numbers commensurate with their proportion of the population. Opponents caricaturized this as a sort of quota system for politics.122

The text of Section 2 is thus careful to note that “nothing in this section establishes a right to have members of a protected class elected in numbers equal to their proportion in the population.”123 The statute accordingly requires Section 2 plaintiffs to “show[] that the political processes . . . are not equally open to participation by [minority voters],”124 via evidence of the Senate Factors. That is, under Section 2, it is not enough for vote dilution plaintiffs to establish that a state or local jurisdiction could rearrange its electoral districts to facilitate the election of more minority-preferred candidates. Rather, plaintiffs must show that a jurisdiction’s failure to do so would perpetuate a racialized political context—one in which the deck is stacked against minority voters such that they lack an equal opportunity to elect their preferred candidates.125

Evidence of the Senate Factors functions similarly in the vote denial context. Under the two-part test, liability is not found in every instance that a voting restriction has a disparate impact on minority voters. Rather, liability results only where that impact is “link[ed to] the effects of past and current discrimination,”126 which can be demonstrated by evidence that the Senate Factors are present.127

This is a critical point. After Shelby County, it would have been tempting to propose a simple disparate impact standard for vote denial liability under Section 2. A test along those lines would have captured more practices and would have been more straightforward than the two-part test incorporating the Senate Factors. But such a framework could have been vulnerable to both legal and policy critiques. Opponents might raise the legal argument that a bare statistical showing of disparate impact is insufficient to make out a Section 2 claim, in light of the Supreme Court’s holding in Gingles that a link to “social and historical conditions” is necessary for Section 2 liability.128 On policy grounds, opponents might argue that such a test for Section 2 liability runs the risk of limitless liability and endless racial balancing,129 potentially raising constitutional concerns that have been invoked in other disparate impact contexts.130

Plaintiffs’ counsel in recent vote denial litigation thus have been careful to emphasize the link between the Senate Factors and the discriminatory burdens imposed by the challenged laws. In Ohio, we did not simply argue that African-American voters would be harmed by early voting reductions because they disproportionately rely on early voting. Rather, we presented evidence that African Americans in Ohio rely on the particular forms of early voting that were at issue (i.e., during evenings and weekends) because they face a wide range of socioeconomic disparities that make it difficult to vote during the nine-to-five workday, “which are themselves tied to contemporary institutional practices and discrimination.”131 Similarly, in Wisconsin, we did not simply argue that voters of color in Wisconsin are less likely to possess photo IDs (or the documents necessary to obtain it). Instead, we noted that racial disparities in ID possession rates were tied directly to “social and historical circumstances in areas such as poverty, unemployment, education, housing, and transportation.”132 These conditions are directly “traceable to discrimination,” which is “the reason Blacks and Latinos are disproportionately likely to lack an ID” and “generally find it harder” to obtain an ID “than do those with greater resources.”133 And in North Carolina, we noted that “myriad lingering socioeconomic disparities [are] attributable to North Carolina’s history of racial discrimination.”134 These disparities, we argued, are the reason why African-American voters in the state disproportionately rely on means of participation such as same-day registration and out-of-precinct voting.135

In each of these cases, evidence of the Senate Factors made clear that the disparate impact of the challenged practices was not the product of chance, but rather arose from the interaction of the challenged practice with longstanding patterns of discrimination and inequality. By embracing the Senate Factors as an element of Section 2 vote denial liability, we made clear that our clients’ claims were not based on disparate impact alone.

Thus, as the Fifth Circuit understood, the Seventh Circuit’s “gloomy forecast” of limitless liability under Section 2 is “unsound.”136 Disparate impact is a necessary component of Section 2 liability, but not sufficient to state a claim on its own. The Senate Factors inquiry narrows liability, serving as a sufficient and familiar way to limit courts’ interference with ‘neutral’ election laws to those that truly have a discriminatory impact under Section 2 of the Voting Rights Act.”137 In this way, the Section 2 results test does not mandate that lawmakers engage in racial balancing with respect to all election administration decisions. Rather, they must simply refrain from adopting restrictions on voting that reiterate or amplify existing patterns of racial discrimination and exclusion.

A consequence of this framework for Section 2 liability is that the same voting restriction—say, a strict voter identification requirement—could be unlawful in some states (i.e., where the Senate Factors are present) but not in others (where they are absent). But again, that is a feature, not a bug. As the Supreme Court made clear in Gingles, “electoral devices . . . may not be considered per se violative of § 2.”138 Section 2’s results standard does not establish a categorical prohibition on any particular voting practices. Indeed, although it was enacted principally to deal with at-large elections, it does not establish an absolute prohibition on that practice. Nor does Section 2 liability occur in every context where at-large elections produce disproportionately low representation for voters of color. Rather, it only occurs where at-large elections perpetuate a longer-standing pattern of racial exclusion, as demonstrated by the presence of the Senate Factors. There is no reason that this should not be equally true in the vote denial context. The question of Section 2 liability is always a “determination [that] is peculiarly dependent upon the facts of each case, . . . and requires an intensely local appraisal of the design and impact of the contested electoral mechanisms.”139

In sum, the majority of circuits are correct to reject a state-sponsored discriminatory intent requirement for Section 2 vote denial liability. Such a requirement finds no basis in the text of Section 2; misapprehends the nature of the Senate Factors, which were designed to be probative of intent without requiring courts to make explicit findings about the motives of defendants; and ignores the Senate Factors’ liability-limiting function in preventing Section 2 from devolving into a racial proportionality rule.


The history of voting rights in America has never been a simple story of linear, progressive expansion. As Alex Keyssar put it in his definitive account of the history of voting rights in the United States,

The evolution of democracy rarely followed a straight path, and it always has been accompanied by profound antidemocratic countercurrents. The history of suffrage in the United States is a history of both expansion and contraction, of inclusion and exclusion, of shifts in direction and momentum at different places and at different times.140

Throughout our history, democratic progress has always been met with reaction. The Founding witnessed the birth of a Republic in which women and free black men could vote in some states, but the following period saw those gains largely reversed.141 Emancipation and Reconstruction promised the enfranchisement of millions of black men who were property only years earlier. But soon thereafter Black Codes and Jim Crow produced nearly a century of disenfranchisement.142 And the forty-five-year period of progressive expansion of the right to participate in elections from 1965 to 2010, which produced the nation’s first African-American President, has now been met with a ferocious backlash in the form of registration and voting restrictions that disproportionately affect precisely those segments of the electorate that emerged in record numbers in 2008, carrying Barack Obama to victory. Thus far, the new two-part test for vote denial liability under Section 2 has functioned as a bulwark against some of the worst attempts at vote denial.

But the future is uncertain. Chief Justice Roberts, who famously opined that “[t]hings have changed in the South”143 does not appear to be moved by more recent developments. In a relatively uncommon move, the Chief Justice issued statements regarding the denial of certiorari in two of these cases. In the Texas voter ID case, he noted that “the discriminatory purpose claim is in an interlocutory posture, having been remanded for further consideration,” and that, on the Section 2 claim, “the District Court has yet to enter a final remedial order.”144 The Chief Justice pointedly noted that “[t]he issues will be better suited for certiorari review” after final judgment, all but promising that the Supreme Court will eventually take the case.145 And in the North Carolina case, after the newly-elected governor of North Carolina sought to withdraw the petition for certiorari, the Chief Justice lamented the confusion that had arisen from “the blizzard of filings over who is and who is not authorized to seek review in this Court under North Carolina law,” and “recall[ed] our frequent admonition that ‘[t]he denial of a writ of certiorari imports no expression of opinion upon the merits of the case.’”146

Numerous cases remain ongoing. A decision is pending in the Seventh Circuit on the plaintiffs’ as-applied challenge to Wisconsin’s voter ID law. The Texas voter ID case is again in the Fifth Circuit (this time for resolution of the plaintiffs’ discriminatory intent claims). And proceedings are ongoing in the challenge to Arizona’s restrictions on absentee voting assistance. As the lower courts refine the two-part test and the Supreme Court is presented with more opportunities to weigh in, it remains to be seen whether Section 2 will continue to provide an effective remedy for the new vote denial.

Dale Ho is the Director of the Voting Rights Project at the American Civil Liberties Union (ACLU). He wishes to express gratitude to his ACLU colleagues, past and present, and his former colleagues at the NAACP Legal Defense Fund (LDF). The views expressed herein are his own and should not be attributed to the ACLU or LDF.

Preferred Citation: Dale E. Ho, Building an Umbrella in a Rainstorm: The New Vote Denial Litigation Since Shelby County, 127 Yale L.J. F. 799 (2018),