On May 7, 2021, Texas legislators in the state’s House of Representatives debated and passed Senate Bill 7, an omnibus bill restricting voting in various ways. The bill would reduce access to mail voting, ban drive-through and 24-hour voting, and require large counties to redistribute their polling places away from Black and Latino neighborhoods (Ura Reference Ura2021 b; Ura, Essig, and Dong Reference Ura, Essig and Dong2021). Although this particular bill would fail after Democrats in the State Senate broke quorum before the final vote by leaving the state, many of these provisions would ultimately become law as part of Senate Bill 1 during a special session called by the governor (Ura Reference Ura2021 a).
The debate in the House, however, was marked by an argument about a single phrase, used in the opening text of the bill. Senate Bill 7’s self-described purpose was to “detect and punish fraud and preserve the purity of the ballot box” (
${\rm{\S}}$
1.02). This phrase—“the purity of the ballot box”—has a long history in Texas, enshrined in the state’s 1876 Constitution and used to defend the state’s white primary that effectively shut nonwhites out of the political process for decades (Knowles Reference Knowles2021; Morris and Pérez Reference Morris and Pérez2021). Democratic representative Rafael Anchía questioned the bill author’s use of this “specific set of words that has a lot of meaning in state history,” (quoted in Knowles Reference Knowles2021) saying the constitutional provision “was drafted specifically to disenfranchise Black people.” The implication was clear: Texas legislators in 2021 were tapping into long-standing legal racism to pass new legislation that would disproportionately impact voters of color. The phrase was dropped from the final version of Senate Bill 1, passed in August 2021.
The twin features of the introduction to Texas’ Senate Bill 7—protection against fraud and appeals to purity—typified Republicans’ rhetoric during the 2021 legislative session around the country. After losing his re-election bid the November before, then-president Donald Trump claimed repeatedly that the election had been stolen (Dale Reference Dale2020), a claim he has continued to maintain and that some 70% of self-identified Republicans believed by early 2022 (Cuthbert and Theodoridis Reference Cuthbert and Theodoridis2022). Many state legislators also justified their support for restrictive legislation in terms similar to Oklahoma State Representative Sean Roberts (sponsor of the restrictive HB 2842 and HB 2847), who told reporters that “[I]t was very clear that the election was stolen from President Trump. We must do everything we can to close those loopholes” (quoted in May Reference May2022). Concerns about election security were not limited to state legislators: 147 Congressional Republicans voted against the certification of the 2020 presidential election, including a supermajority of Republicans in the House (Yourish, Buchanan, and Lu Reference Yourish, Buchanan and Lu2021).
Despite these widespread claims, no evidence of fraud arose following the 2020 election. The New York Times explained: “After bringing some 60 lawsuits, and even offering financial incentive for information about fraud, Mr. Trump and his allies have failed to prove definitively any case of illegal voting on behalf of their opponent in court—not a single case of an undocumented immigrant casting a ballot, a citizen double voting, nor any credible evidence that legions of the voting dead gave Mr. Biden a victory that wasn’t his” (Rutenberg, Corasaniti, and Feuer Reference Rutenberg, Corasaniti and Feuer2020). The lack of evidence of fraud led many to believe that the backlash was due to another force: namely, racial animus (e.g., Bacon Reference Bacon2022). These arguments are bolstered by incidents such as the Texas anecdote opening this article, in the targeting of Black cities such as Detroit, Milwaukee, and Philadelphia for “racialized charges of corruption” (Badger Reference Badger2020), and by none-too-subtle comments like that from one Arizona lawmaker about the need to “to get back to 1958-style voting” (quoted in Fischer Reference Fischer2022).
This paper asks whether racial context and white racial backlash can explain the patterns of restrictive voting bills following the 2020 presidential election, above and beyond any partisan influences. I begin by looking, as past work has done, at state-level counts of introduced and passed bills. Like past work, I find that racial threat continues to play an important role in the introduction and passage of these bills: while states with unified Republican control were more likely than other states to introduce and pass these bills, restrictive legislative activity is highly concentrated in the least-white, Republican-dominated states.
I then move to the legislative district level, asking whether district characteristics predict lawmakers’ sponsorships of these bills. Here, racial backlash is operationalized in two ways. I begin by showing that sociological theories of geographical threat explain the patterns in bill sponsorship: lawmakers from the whitest parts of the least-white states were by far the most likely to serve as sponsors on restrictive legislation. Secondly, I use two waves from the Cooperative Election Study to show that legislators from districts where white respondents have the highest racial resentment scores were considerably more likely to sponsor one of these bills.
This paper contributes important insights into how race informs restrictive voting legislation in the post-2020 presidential election world. I uncover strong evidence that, despite claims about fraud and fraud alone, theories of racial threat clearly predict where these bills are sponsored, introduced, and passed. Given that the scholarly literature firmly establishes that restrictive voting laws disproportionately impact voters of color (e.g., Herron and Smith Reference Herron and Smith2014; Kuk, Hajnal, and Lajevardi Reference Kuk, Hajnal and Lajevardi2022; Weaver Reference Weaver2015), this trend is concerning. The project makes several other contributions as well. This project allows us to understand the sub-state factors influencing legislative behavior with respect to voting policy, moving beyond the state level that previous studies have looked at. Finally, this project is the first to establish a link between district-level racial resentment scores and legislative behavior, raising important future questions about the extent to which lawmakers are introducing bills reflective of their constituents’ racism and anti-Black attitudes.
Recent Work on Restrictive Voting Laws
Over the past 15 years, scholars have explored the introduction and passage of restrictive voting laws across the country. This work has almost exclusively focused on state-level factors, with a general consensus that these laws find the most fertile ground in states with large demographic change and a growing nonwhite electorate (Bentele and O’Brien Reference Bentele and O’Brien2013), where large numbers of Black Americans reside (Behrens, Uggen, and Manza Reference Behrens, Uggen and Manza2003; Biggers and Hanmer Reference Biggers and Hanmer2017), and in electorally competitive states where Republicans hold a slight edge (Hicks et al. Reference Hicks, McKee, Sellers and Smith2015).
Behrens, Uggen, and Manza (Reference Behrens, Uggen and Manza2003) use a historical approach to understand the passage of laws disenfranchising citizens convicted of felony offenses. As they note, all but two American states restrict voting rights for at least some incarcerated citizens; the two that do not—Maine and Vermont—are also the two whitest states in the nation. Behrens and colleagues document the rise of these restrictive laws in the aftermath of the passage of the 14th and 15th Amendments which expanded formal citizenship and granted voting rights to Black men. Drawing on Blumer (Reference Blumer1958) and other scholars of group threat, they argue that white (male) Americans were threatened by the prospect that their sole control over the political domain was no longer so secure. Of course, their claims to racial political dominance were threatened proportionate to the number of nonwhite potential voters; as such, states with larger nonwhite populations had political incentives to develop new ways to disenfranchise Black men. They find strong support for the theory that the widespread adoption of felony disenfranchisement rules rose from this threat. “Our key finding can be summarized concisely and forcefully,” they write. “States with greater nonwhite prison populations have been more likely to ban convicted felons from voting than states with proportionally fewer nonwhites in the criminal justice system” (Behrens, Uggen, and Manza, Reference Behrens, Uggen and Manza2003, 596). Their conclusions have been corroborated more recently. Eubank and Fresh (Reference Eubank and Fresh2022) find that states subject to strict federal oversight under the 1965 Voting Rights Act’s Section 5 selectively increased the incarceration of Black Americans, providing further evidence that increased political opportunity for racial minorities leads white majorities to seek other ways of restricting their effective power.
Of course, the incarceration of citizens and subsequent legal disenfranchisement is perhaps only the most drastic example of curtailing access to the ballot.Footnote 1 Might less extreme attempts to limit the pool of eligible voters follow a similar pattern? And do such considerations structure legislative behavior in the modern era? Bentele and O’Brien (Reference Bentele and O’Brien2013) consider the introduction and passage of 5 types of restrictive voting legislation over the 2006–2011 period. They conclude that the strongest predictor of the introduction and passage of restrictive voting laws is the political power demonstrated by racial and ethnic minorities, arguing that “legislative developments in this policy area remain heavily shaped by racial considerations” (Bentele and O’Brien Reference Bentele and O’Brien2013, 1089). At the same time, they find no evidence that prevalence of voter fraud impacted the introduction of restrictive provisions and that it was “only a minor contributing factor” to the passage of these laws in 2011 (1103). Biggers and Hanmer (Reference Biggers and Hanmer2017) similarly find that states with larger Black populations are more likely to adopt voter ID laws—especially when a Republican resides in the governor’s mansion.
A further insight from Bentele and O’Brien (Reference Bentele and O’Brien2013)—that restrictive provisions are passed most frequently in electorally competitive states—is corroborated by Hicks et al. (Reference Hicks, McKee, Sellers and Smith2015). Looking specifically at the introduction and passage of restrictive voter identification laws in the early 2000s, they find that states with more Republican legislators were considerably more likely to enact these provisions—but that “Republicans have not pursued this scorched-earth policy in all states, nor have they done so consistently over time” (29–30). Instead, Republicans were more likely to pass these bills where their electoral majorities were slim. Hicks et al. (Reference Hicks, McKee, Sellers and Smith2015, 18) thus conclude that “where elections are competitive, the furtherance of restrictive voter ID laws is a means of maintaining Republican support while curtailing Democratic electoral gains.” Other work (e.g., Wang Reference Wang2012) also indicates that restrictive voting laws are passed by Republican-dominated legislatures to shore up flagging electoral majorities.
This scholarship sheds important light on where restrictive voting laws are the most likely to go into effect, and the results are not encouraging. There is strong evidence that racial threat predicts the passage of these restrictive bills across the country, even as legislators proclaim that the changes are needed to combat widespread fraud (see, for instance, Minnite Reference Minnite2010; Piven, Minnite, and Groarke Reference Piven, Minnite and Groarke2009). Important as this research has been, however, it fails to explain the full set of dynamics between demographic composition and bill introduction. The explosion in the introduction of restrictive voting laws in 2021 makes this clear: according to the data from the Brennan Center for Justice used throughout this project, just one state (Vermont) introduced no voting bills in 2021 containing a restrictive provision. Moreover, the number of restrictive provisions introduced and passed in 2021 has little historical precedent: according to the Brennan Center, 880 restrictive provisions were introduced and 93 were passed. By way of comparison, Bentele and O’Brien (Reference Bentele and O’Brien2013)—which also used data from the Brennan Center—call the roughly 20 passed provisions in 2011 a “dramatic increase” (1088; see their Figure 1).
State and local characteristics and the sponsorship of restrictive bills
Clearly, something more complex than state-level factors are at play in the contemporary push to restrict voting rights. A pair of papers from Seth McKee and colleagues sheds some light on this. They look specifically at roll-call votes on voter ID bills, focusing only on states where they were passed between 2010 and 2013 and between 2005 and 2013 (Hicks, McKee, and Smith Reference Hicks, McKee and Smith2016; McKee Reference McKee2015, respectively). They find that Democratic legislators representing Black constituents were the most likely to vote against these bills, while Republican legislators with sizeable Black constituents voted for them at higher rates. Unfortunately, without looking at legislator behavior with respect to other sorts of restrictions, and without looking at states where restrictive bills fail to pass, understanding the full dynamics of sub-state legislator behavior is difficult. Further, the overwhelming polarization on this issue (more than 97% of Republicans voted in favor of the bills, while fewer than 12% of Democrats did (Hicks, McKee, and Smith Reference Hicks, McKee and Smith2016, 424)) might limit how much information can be gleaned from roll-call votes.
By considering not only state-level factors but also examining the demographics of the districts represented by legislators who sponsor these restrictive bills, this project marks a significant step forward in understanding how racial threat’s influence on the policy-making process is mediated by factors at multiple geopolitical levels. The following section steps back to engage with the (racial) threat literature and, more specifically, consider how spatially situated theories of threat help us to formulate expectations about the roles played by state and local factors in the introduction and passage of restrictive voting laws.
Patterns in Support at the Sub-State Level
Rather than look to roll-call support for restrictive voting laws, I consider which legislators sponsored restrictive voting bills in 2021. Roll-call votes are unlikely to yield substantively interesting results: in some Democratic states, restrictive bills that were introduced never made it to the floor; on the other hand, where these bills were considered, the roll-call votes broke along highly partisan lines: bills passed into law were supported by north of 98% of Republicans and fewer than 25% of Democrats (Morris and Wilder Reference Morris and Wilder2022). When it comes to who voted in favor of these bills, partisan considerations seem to override other factors.
Given the polarization in roll-call votes, the choice to sponsor a bill with a restrictive voting provision is likely to be at least partly an expressive act. Existing literature argues that legislators take non-roll-call actions like bill sponsorship to send messages to interested groups like potential donors (e.g., Rocca and Gordon Reference Rocca and Gordon2010) and to their base constituents (Highton and Rocca Reference Highton and Rocca2005, but see Waggoner Reference Waggoner2019). In the sections that follow, I first explore what might lead constituents to support restrictive voting policy. I then consider whether legislators are likely to reflect constituent preferences on this salient, racialized policy.
A Changing Electorate and Threat
Scholars across the social sciences have long noted the importance of threat to the policy-making process; indeed, each of the studies discussed in the previous section implicitly or explicitly draws on this literature. Tilly (Reference Tilly1978, 73) separates collective action into three categories: defensive, offensive, and preparatory. Of these, two—defensive and preparatory—are explicitly linked to threats, where political actors pool their resources to fend off challenges to their interests, or to regain what has already been lost. Beck (Reference Beck2000) extends this theory to note that defensive actors need only perceive that their interests have been compromised to mobilize in a reactionary way; the reality of any worsened station is perhaps less important. These threats can take multiple forms, be they economic, political, or demographic (Van Dyke and Soule Reference Van Dyke and Soule2002). Much of this work draws from what is known as social identity theory (Tajfel Reference Tajfel1978, Reference Tajfel1982) which argues that group members—in this case, whites—react conservatively to maintain and consolidate the benefits accrued to their own group (Bobo and Hutchings Reference Bobo and Hutchings1996).
Sociological work pays increasing attention to how spatial organization and threat can interact with one another (e.g., Zhang and Zhao Reference Zhang and Zhao2018). Tilly and Tarrow (Reference Tilly and Tarrow2015) explain how social movements can undergo what they call an “upward scale shift,” which they say “moves contention beyond its local origins, touches on the interests and values of new actors, involves a shift of venue to sites where contention may be more or less successful, and can threaten other actors or entire regimes” (125). In other words, political actors may move beyond the local context to make use of institutional tools available only at higher levels of government.
Despite this growing literature, relatively little attention has been paid to the policy implications of scale shift and geographically mediated threat. A notable exception to this is Andrews and Seguin (Reference Andrews and Seguin2015), which explores how racial threat, group contact, and differential levels of government structured legislative activity in the early 20th century. They argue that “threat arises primarily from interactions between spatially proximate units at the local level… and therefore higher-level policy change at the state level is not reducible to the variables driving local policy” (476). In short, examining local and state characteristics alone is not sufficient to understand legislator support for racially conservative policy changes; instead, responses to racial threat arise from the interaction of these circumstances.
Voting rights are likely a policy issue where geographically mediated threat plays an especially large role. It is clear how both local and state factors would influence the racial threat experienced by white Americans. States like Texas and Georgia continue to have large pockets of homogeneous white communities, despite large nonwhite populations in their urban cores. Although white residents of these areas might not compete with nonwhite residents at the local level, they nevertheless are in competition with nonwhite voters at the state level over things like education policy and U.S. Senate seats. This contrasts with white communities in more homogeneous states like Montana, where white voters compete with nonwhite voters at neither the local nor state level. In other words, state-level factors might differentially structure racial threat even among districts that—when examined in isolation—look quite similar. This should lead to different legislative responses on a racialized policy like voting rights.
Racial Resentment and Support for Democracy
In the past section, we considered how threat and, specifically, geographically mediated racial threat might structure voters’ support for restrictive voting bills. I turn now to a discussion of how constituents’ racial resentment might have the same effect.
Racial resentment has a long history in political science, dating back to the late 1980s and 1990s (Kinder and Sanders Reference Kinder and Sanders1996). The racial resentment scale was developed in the aftermath of the Civil Rights Era of the 20th century. As the expression of biological or “old-fashioned” racism became socially unacceptable, politicians and others took up what is often referred to as “symbolic” racism. Rather than speak directly about racial groups as being inherently and biologically inferior, symbolic racism relies on tying violations of the Protestant work ethic to nonwhite—and particularly Black—Americans. This reflects a cultural shift away from the acceptability of justifying worse life outcomes because of racial differences, tapping instead into the meritocratic mythos of the United States. In short, by associating the violation of cultural norms with racial and ethnic minorities (whether those norms are in fact violated or not) provides cover for continued discrimination. Although the racial resentment scale has faced criticism over the years for overlapping with other ideological commitments (e.g., Feldman and Huddy Reference Feldman and Huddy2005), the scale continues to explain whites’ racial attitudes better than simple measurements of liberal or conservative ideology (Enders Reference Enders2021). As Ashley Jardina explains it, racial resentment “is, quite clearly, a central component to the way in which whites interpret the political and social world” (Jardina Reference Jardina2019, 14).
Over the past few years, scholars have taken a serious look at how white Americans’ racial attitudes structure support for democracy and democratic norms. Miller and Davis (Reference Miller and Davis2021), for instance, use four waves of the World Values Survey to look at the racial attitudes of white Americans over the 1995–2011 period. Although they do not use the traditional racial resentment scale, their survey items similarly tap into whites’ racial attitudes. Miller and Davis argue that democracy requires at least a minimal “commitment to tolerance and engaging in neighborly actions toward others” (336). In other words, in a pluralistic society, one cannot claim to support democratic ideals without being willing to allow members of other races and groups to make claims on the state and to share in decision-making power.
But what happens when people are unwilling to “engag[e] in neighborly actions” towards members of different racial groups? How might whites who express hostility to racial outgroups reconcile these attitudes with residence in a country that places a high value on democratic ideals? It seems, Miller and Davis argue, that there really is a tension between what they call “social prejudice” and democratic commitments: they find that whites who score higher on their prejudice scale are significantly more likely to oppose democratic rule (as well as exhibit higher support for strong leaders and army rule).
Building on the theoretical construction of white racial identity developed by Jardina (Reference Jardina2019), Jardina and Mickey (Reference Jardina and Mickey2022) come at this question from a slightly different angle. Rather than ask whether animus directed against racial outgroups undermines support for democracy, they explore whether an affirmative white identity—which is “not highly correlated with standard measures of racial prejudice” (81)—undermines democratic commitments. Here, they argue that whites may recognize that populist white leaders like Donald Trump accrue benefits to them as whites and that democratic norms might undermine their preferred place on the racial hierarchy. Despite the different theoretical mechanisms, Jardina and Mickey find that when white Americans strongly identify with their whiteness, they are less concerned with federal checks-and-balances or presidents who are constrained by rules.
In much of the same vein, Enders and Thornton (Reference Enders and Thornton2022) find that racial resentment among white Americans substantially moderates satisfaction with democracy and democratic outcomes. Looking at presidential elections from recent years, they argue that white Democrats high in racial resentment were more satisfied after their preferred candidate lost, presumably because the Republican president would better represent their racial preferences. Conversely, white Republicans low in racial resentment were more satisfied with democracy in the aftermath of Obama’s victory than their more racially resentful copartisans.
Constituents’ Preferences and Legislative Behavior
But do these attitudes spill over into legislative behavior? Do lawmakers who represent racially resentful white Americans sponsor restrictive bills at higher rates? While we know much about how race and racial resentment influences the policy preferences of white Americans, very little is known about how these racialized preferences influence legislator behavior and the actual policy-making process (but see Garcia and Stout Reference Garcia and Stout2020).
The political science literature does, however, have a good sense of how the views of a representative’s constituency shape how they legislate more generally. The “retrospective voting” model (e.g., Fiorina Reference Fiorina1981; Key Reference Key1968) is central to our understanding of how politicians act in office; if elites fail to reflect the policy demands of their constituents, voters will elect a competitor in the next election. This looming spectre of being ousted from office provides a powerful incentive for the representative to reflect the preferences of her voters (Canes-Wrone, Brady, and Cogan Reference Canes-Wrone, Brady and Cogan2002; Mayhew Reference Mayhew2004). Over time, Caughey and Warshaw (Reference Caughey and Warshaw2022) show legislators in state politics have been increasingly good at reflecting the preferences of their voters, and other work underlines the important role of public opinion in shaping policy-making (see, for instance, Burstein Reference Burstein2003). Yet as Lax and Phillips (Reference Lax and Phillips2012) point out there are still major “democratic deficits” in the states: policies are only reflective of the majority’s will about half the time. Nevertheless, scholars argue that these increasingly strong links between policy preferences and legislative behavior are probably a net positive. In most domains, it seems uncontroversial to prefer lawmakers who reflect their constituents well.
The normative value of legislators acting upon their constituents’ desires becomes less obvious when it comes to racialized policy preferences. Voting rights, of course, fall into this category. When policy preferences are reflective of racial resentment and racial threat, and when they would restrict the liberties of nonwhite Americans, we would hope that legislators would push back against their constituents. However, it seems that voting restrictions are the sort of policy on which lawmakers are most likely to reflect their constituents’ desires.
The more salient an issue is, the more likely a lawmaker is to act on it in accordance with the public’s will (e.g., Adler, Cayton, and Griffin, Reference Adler, Cayton and Griffin2018; Burstein Reference Burstein2003; Canes-Wrone and Shotts Reference Canes-Wrone and Shotts2004; Lax and Phillips Reference Lax and Phillips2012). In the aftermath of the 2020 election, the health of American democracy was top-of-mind for many Americans. Although their views differ substantially, Democrats and Republicans alike feel that democratic systems of governance are under unprecedented threat. Some feel the threat comes from a stolen election, others from curtailed access to the ballot box. But for most Americans, the health of elections is salient today in ways it hasn’t been in generations. The literature indicates that this salience should make lawmakers especially responsive to public opinion.
Legislators are thus most likely to be in step with their constituents on a highly salient issue like voting rights. But might preferences on this topic flow from the lawmaker to the voters, and not the other way around? It seems unlikely for a few reasons. First, a recent paper from Barberá et al. (Reference Barberá, Casas, Nagler, Egan, Bonneau, Jost and Tucker2019) explored this question of directionality, finding that members of Congress follow, rather than lead, the conversations initiated by their supporters. Moreover, members of Congress are especially likely to follow their core supporters’ opinions on topics of high salience. In other words, legislators act as delegates of their constituents’ opinions most often on high-salience issues and are also more likely to parrot—and not guide—their core supporters on precisely these issues.
The second reason that state-level legislators are likely learning from, and not instructing, their constituents on the topic of voting rights centers on Americans’ low knowledge about their representatives (e.g., Jaeger, Lyons, and Wolak Reference Jaeger, Lyons and Wolak2017). The majority of Americans cannot correctly name their Congressional representative; the share who can do so for their state legislative representatives is surely far lower. Finally, and as discussed below, district-level racial resentment measures pre-date the 2021 legislative session, when the rhetoric about voting laws was particularly heated. In short, the relationship between constituents’ racialized views of voter fraud and legislative activity almost surely flows from the constituent to the legislator.
Data, Methods, and Expectations
Throughout my analyses, I primarily rely on the Voting Laws Roundup, a project of the Brennan Center for Justice at NYU School of Law. The Brennan Center systematically reviews all laws introduced around the country that relate to voting and the administration of elections in each state, coding each provision in each bill as “restrictive,” “neutral,” or “expansive.” I use these records from the 2021 legislative session, and the number of restrictive provisions introduced and passed in each state that year forms the dependent variables of my state-level analyses.
I account for the partisan control of each state in two ways. Following Hicks et al. (Reference Hicks, McKee, Sellers and Smith2015), it seems possible that electorally competitive states where Republicans hold unified power would be most likely to introduce and pass restrictive provisions. I thus include 2 dummies: one measuring whether the state was competitive in 2020 (that is, Biden received between 45% and 55% of the two-way vote share) and one measuring whether Republicans held unified control in 2021. Data on electoral competitiveness come from the MIT Election Data and Science Lab (MIT Election Data and Science Lab 2021), and data on partisan control come from the National Conference of State Legislatures.Footnote 2 Although Nebraska’s unicameral state legislature is formally nonpartisan, they are considered to be under unified Republican control for the purposes of this study.
Standard socioeconomic district-level characteristics (income, age, collegiate education, and population density) come from the 5-year ACS estimates ending with 2021. I also incorporate information about how difficult voting was prior to 2021 using the Cost of Voting Index (COVI) (Schraufnagel, Pomante II, and Li Reference Schraufnagel, Pomante and Li2020). Finally, I use the Squire Index (Squire Reference Squire2017) to control the professionalism of legislatures in the 50 states.
These data allow me to test H1: Other things being equal, racially diverse states with unified Republican control will introduce and pass more restrictive voting legislation than whiter, Republican-dominated states, and than states with split or Democratic control.
In the second set of analyses, legislative districts form the unit of analysis. The empirical framework is the same, though the dependent variable changes from the count of restrictive provisions to a dummy variable indicating whether a legislator sponsored restrictive provisions, due to the smaller scale. The analyses conducted at the level of the legislative district primarily leverage data from LegiScan, an organization that tracks state-level bills around the nation. Importantly, LegiScan records the legislators who sponsor each of these bills. By merging the restrictive bills identified by the Brennan Center with the LegiScan data, I identify all districts represented by a lawmaker who sponsored at least one law with a restrictive voting provision in 2021.
The primary independent variables for the first set of district-level analyses are the white share of the district and the white share of the state, along with their interaction. These are included to test whether the influence of the whiteness of a district on the probability that a lawmaker sponsors a restrictive bill is influenced by state-level factors. Of course, administrative and demographic data cannot give us insight into the political disposition of district residents. As such, I also incorporate survey data from the 2018 and 2020 waves of the Cooperative Election Study (CES) to test whether districts’ racial resentment scores are associated with the sponsorship of restrictive bills.
The CES asks white voters how strongly they agree (on a scale of 1 to 5) with two statements related to racial resentment: Irish, Italians, Jewish, and many other minorities overcame prejudice and worked their way up. Blacks should do the same without any special favors and Generations of slavery and discrimination have created conditions that make it difficult for blacks to work their way out of the lower class. I reverse code agreement with the first statement, such that higher scores for both statements are associated with higher levels of racial resentment. Respondents’ racial resentment scores are calculated as the mean of their response to these two questions. I retain only the responses of white respondents, resulting in just under 90,000 respondents (N = 45,011 in 2018; N = 44,128 in 2020).
While the CES data do not include respondents’ home legislative districts, the survey makes home ZIP codes available. To calculate district resentment scores, I begin by assigning every Census block in the country the mean resentment score of the ZIP code in which its centroid falls. District resentment scores are then calculated as the population-weighted average racial resentment score of all blocks in the district. In the Appendix, I show that the results are generally consistent if instead I assign each CES respondent to the legislative district in which a plurality of the population in their home ZIP code lives and aggregate using raked survey weights, an approach that correctly identifies the home legislative district of more than 80% of Americans.
In addition to these primary independent variables, I also control for a variety of other legislator-specific characteristics (partisanship, gender, and race) obtained from LegislatorDiversity.com.Footnote 3 District-level characteristics come again from the 5-year ACS estimates ending in 2021 and mirror those used in the state analyses. The district-level analyses also include state-level controls for competitiveness, unified Republican control, and the Squire Index.
Formally stated, I test the following hypotheses in the district-level models:
H2: Whiter districts in less-white states were the most likely to be represented by legislators that sponsored restrictive voting provisions, other things being equal.
H3: Districts where white CES respondents had higher levels of racial resentment were more likely to be represented by legislators that sponsored restrictive voting provisions, other things being equal.
Results
Legislative Activity at the State Level
Table 1 presents the results of an econometric model testing the relationship between state demographics, partisan control, and the number of restrictive provisions introduced and passed in each state. I use what is called a robust regression to keep outliers like Texas from skewing the results too dramatically; the results are substantively similar using OLS. The OLS regressions can be found in the Appendix, along with scatter plots of the underlying data. Further, I here interact the linear, squared, and cubed state share white with the partisan control of the state; in the Appendix, I show that this results in the best model fit, but that the story does not meaningfully change using a first or second order polynomial.
State-level restrictive provisions, 2021
Non-Hispanic White, Non-Hispanic White2, and Non-Hispanic White3 computed using orthogonal polynomials.
95% confidence intervals shown below estimates.
Interpreting the interactions between the state share white and partisan control of the state is difficult looking only at Table 1; as such, Figure 2 presents these relationships graphically. Throughout the Results section, these predicted probabilities plots show the predicted value of the dependent variable for any value of the key independent variable, inclusive of any transformations made to that independent variable. All other covariates are held constant, at their mean. This allows us to see the relationship between the dependent and key independent variable net of those other covariates.
Partisanship, race, and restrictive provisions
Table 1 and Figure 2 make a number of things clear. Firstly, there is no statistically significant relationship between racial demographics and the introduction or passage of restrictive voting provisions in the 26 states where Republicans do not hold all three branches of government. The median state without unified Republican control introduced 8 or 9 restrictive provisions and passed zero. Simply put, there was not a lot of restrictive activity happening in these states.
Taken as a whole, the 24 Republican-controlled states did not look all that different. The average state only introduced and passed a few more provisions than the split or Democratic states. What is different, however, is the role that race plays in these states. In contrast with the states without unified Republican control, there is a strong relationship between racial characteristics and restrictive activity in states where Republicans hold all the levers of power. Lawmakers in less-white states where Republicans can unilaterally pass bills into law along party lines introduced and passed far more restrictive provisions.
This relationship remains after I control for other characteristics in the states with total Republican control. How can this be? It turns out that even uncompetitive racially diverse Republican states saw significant legislative activity, especially when it came to the introduction of these bills. While the four whitest uncompetitive Republican states (Wyoming, North Dakota, Montana, and West Virginia) collectively introduced 28 restrictive provisions in 2021, the four least-white uncompetitive Republican states (Mississippi, Alaska, South Carolina, and Oklahoma) introduced 63 restrictive provisions—more than twice as many. Thus, race seems to be a driving factor for voting rights backlash in Republican-dominated states even when those states are not electorally competitive. While Table 1 also shows that competition in Republican states is associated with a higher number of restrictive provisions passed (but not in other states, and not when it comes to the number of restrictive provisions introduced), racial demographics are clearly playing an important role, too.
These results corroborate past work highlighting the importance of race to restrictive voting laws and provide strong evidence in support of H1: in states where the Republican party has total control, the presence of racial and ethnic minorities is strongly predictive of how many restrictive provisions were introduced and passed.
State and Local Racial Composition
I turn now to the relationship between a district’s racial composition, the racial makeup of its home state, and whether the lawmaker signed on as a sponsor to at least one bill with a restrictive provision. I use racial estimates from the U.S Census Bureau’s American Community Survey.
Table 2 presents the results of OLS regressions run at the district level. Models 1 and 2 present these results for the lower chambers, while models 3 and 4 present the upper-chamber models. In models 1 and 3, I present the relationships between district and state demographics, and sponsorship, alone; in models 2 and 4, I include other relevant covariates. In the Appendix, I show that using the squared term of the share white results in a slightly better model fit than the linear term, but that the results do not differ meaningfully when using only the linear district share white term. In Table 2, the first set of covariates are the primary independent variables (i.e., the district and state share white). These are followed by legislator-level covariates, district-level covariates, and finally state-level ones. We can see that Republican legislators were more likely to introduce restrictive bills, but that white ones were not meaningfully more likely to do so than nonwhite ones.
District-level sponsored provisions, 2021
* p < .05, ** p < .01, *** p < .001
The dependent variable ranges from 0 to 100.
Non-Hispanic White and Non-Hispanic White 2 computed using orthogonal polynomials.
95% confidence intervals shown below estimates and computed with robust standard errors.
As before, interpreting interacted coefficients from a regression table is difficult. Figure 1 thus once again plots the predicted probabilities of bill sponsorship for legislators from different types of districts and states. In particular, it shows that the relationship between sponsorship and racial characteristics is highly moderated by the demographics of the state: lawmakers from the whitest districts of the least-white states sponsored restrictive bills at the highest rates. This is especially true of upper-chamber legislators. Importantly, these relationships are net of the other covariates.
So far, we have only looked at simple racial and ethnic demographics, but what we see is exactly what geographic theories of racial threat would predict: white enclaves in competition within their state with racial and ethnic minorities are represented by legislators restricting voting. This provides strong support for H2.
Racial Resentment and Legislative Sponsorship
Finally, I turn to the relationship between districts’ racial resentment and the likelihood that their representative sponsored a bill with at least one restrictive provision in it. These models are similar to the ones presented in the past section interested in geographically mediated racial threat; here, instead of looking at the interplay of local and state circumstances, I use the scores estimated from the two CES waves. Table 3 shows that the results are broadly similar to the previous section: Republican legislators are the most likely to sponsor these bills, as are legislators from competitive states and with unified Republican control. Of particular note, with the exception of population density in the lower-chamber model, the only district-level characteristic associated with bill sponsorship is racial resentment. As before, I present these relationships graphically in Figure 3.
District-level sponsored provisions, 2021
* p < .05, ** p < .01, *** p < .001.
The dependent variable ranges from 0 to 100.
95% confidence intervals shown below estimates and computed with robust standard errors.
District racial resentment and the sponsorship of restrictive bills
Even after accounting for these other potential explanations, district-level estimates of racial resentment have a strong, statistically significant relationship with the likelihood of being represented by a lawmaker who sponsored a restrictive bill.Footnote 4 Just as in the previous section where theories of spatial racial threat explained why lawmakers from white areas in diverse states sponsor racially restrictive legislation, I find evidence that lawmakers from more racially resentful districts are more actively supporting bills restricting access to the franchise. This is strong evidence in support of H3.
Conclusion
I opened with a look at a debate on the floor of the Texas House of Representatives, where lawmakers argued in favor of restrictive voting laws using language directly tied to the Lone Star State’s long history of disenfranchising voters of color. While the bill’s proponents claimed ignorance of this fact, and the phrasing was dropped from the eventual omnibus bill, the question remained: is there any evidence that ties legislative support for restrictive voting regimes to racial animosity?
The answer is a resounding yes. Here I used two different ways of getting at this question. I started by considering what theories of spatial racial threat have to tell us about where racially regressive policies find support. The restrictive legislation introduced in the 2021 legislative sessions around the nation was not sponsored by representatives from racially diverse districts, where white voters are directly competing with voters of color over whom they send to the statehouse. They also are not being sponsored by lawmakers in overwhelmingly white states, where voters of color are too few in number to pose immediate political, economic, or social threat. Instead, the lawmakers most likely to sponsor these bills come from the white bastions of diverse and diversifying states. In these states—states like Georgia and Texas and Arizona—white areas might not face internal competition from racial and ethnic minorities, but do compete with them in statehouses, Senate races, and Electoral College votes. These electoral threats need not be real to motivate behavior: take Arkansas, for instance. Despite high racial diversity (only about two-thirds of residents are non-Hispanic white), the state was not competitive in 2020: Trump won by nearly 30 points. Yet only three states—Texas, Iowa, and Georgia—passed more restrictive provisions into law in 2021.
Of course, demographic data do not tell us anything about how residents of different areas actually think about racial and ethnic minorities. To get a sense of whether constituents’ racial animus was related to whether lawmakers supported these bills, I turned to the racial resentment scale and two waves of the Cooperative Election Study. While recent scholarship has detailed how racial resentment and other measures of discrimination are tied to antidemocratic beliefs (Enders and Thornton Reference Enders and Thornton2022; Jardina and Mickey Reference Jardina and Mickey2022; Miller and Davis Reference Miller and Davis2021), there has been no work exploring how or whether this spills over into legislative behavior. Though the results are not causal, I do find evidence that lawmakers from more resentful districts sponsor restrictive voting legislation at considerably higher rates—and that these relationships cannot be explained away by partisanship, competition, or district-level demographics like race and education.
The implications of this research extend, of course, from who attempts to make voting more difficult to those that are actually disenfranchised. And here the scholarly literature is clear: whether we consider the observed (Enders and Thornton Reference Enders and Thornton2022; Jardina and Mickey Reference Jardina and Mickey2022; Miller and Davis Reference Miller and Davis2021), or potential (Barreto et al. Reference Barreto, Nuño, Sanchez and Walker2019) impacts of voter ID laws; cuts to early voting (Herron and Smith Reference Herron and Smith2014; Weaver Reference Weaver2015); the shuttering of polling places (Cantoni Reference Cantoni2020; Morris and Miller Reference Morris and Miller2022); or more onerous mail ballots rules (Morris and Grange Reference Morris and Grange2022), there is consistent and convincing evidence that the impacts of these bills fall hardest on voters of color. In short, these bills appear to have racial motivations, and they will almost certainly have racially discriminatory effects.
While I lean heavily on classical racial resentment in this project, I take the critiques of the racial resentment index levied by Davis and Wilson (Reference Davis and Wilson2021) and others seriously. Although the measure is still widely used, it is probably best not interpreted as indicative of overt racial prejudice. Without better measures of resentment like those proposed by Davis and Wilson, and without measures of old-fashioned racism, the interpretation of the role of racial resentment in bill sponsorship is somewhat murky. Nevertheless, we do know that higher scores on the racial resentment scale are associated with stronger beliefs in stereotypes about Black Americans and higher investment in status quo maintenance. Thus, despite the murkiness surrounding racial resentment, these results shed light on the relationship between beliefs about the undeservingness of Black Americans and the symbolic power of sponsoring regressive legislation. Future work must further interrogate whether this expressive desire to restrict the franchise loads more fully on outright racial animus, on a desire for hierarchy maintenance, or direct political resentment. Similarly, scholars should investigate whether measures of constituents’ racial orientations inform how legislators act in other racialized policy areas. One potentially fruitful avenue for such research would be exploring how relationships between resentment and legislative sponsorship of restrictive voting laws have changed over the past two decades, as the issue has become progressively higher-salience and more polarized.
Future work should also interrogate just how much these new restrictions increased barriers to effective representation. Unfortunately, standard tools like the Cost of Voting Index (Schraufnagel, Pomante II, and Li Reference Schraufnagel, Pomante and Li2020) are not perfectly comparable from year to year; restrictions on passing food and water out to voters, for instance, were not captured prior to 2022. Understanding how these laws directly impact turnout through raising the cost of voting, indirectly through influencing internal and external efficacy, or otherwise influencing participation is of similar import.
The threats to American democracy are real and growing. As Edelman (Reference Edelman2022) shows, more than half of states had election deniers on their ballot in the 2022 general election. While most state-wide election deniers lost in the 2022 midterms, potential dangers to the security of the 2024 presidential are clear. These threats surely extend beyond race, but this project demonstrates that they cannot be understood without taking racial threat and racial resentment into serious account. As America continues its generational march toward an inclusive, multiracial democracy, protecting the voting rights of racial and ethnic minorities is of signal importance.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/rep.2023.26
Competing interests
The author declares none.
Open access
