Divisive or Descriptive?: How Americans Understand Critical Race Theory

Critical Race Theory (CRT) has become a flashpoint of elite political discord, yet how Americans actually perceive CRT is unclear. We theorize that Republican elites utilized a strong framing strategy to re-define CRT as an “ empty signifier ” representing broader racial and cultural grievances. Using a survey and a pre-registered experiment among U


Introduction
In a few short years, the scholarly approach known as Critical Race Theory (CRT) went from a relatively obscure academic framework to a flashpoint in the American culture wars.CRT has been featured in newspapers, primetime specials on cable news, Presidential executive orders, and a slate of laws and regulations concerning how history can be taught in public schools.However, despite the prominence of CRT in contemporary politics, there is limited research examining public attitudes toward CRT.
The rise in anti-CRT discourse is largely traced to right-wing backlash starting in 2019 against the New York Times' "The 1619 Project" (New York Times 2021).Conservative political media and elites expressed grievances about how American history is understood and taught under the term CRT (Bailey et al. 2021).They claim that the CRT framework, and especially views highlighted by "The 1619 Project," is an inaccurate representation of American history.Issues surrounding race further rose to prominence in the summer of 2020, when an uptick in protests occurred to curb police violence against Black citizens.Simultaneously, a conservative backlash against both the Black Lives Matter movement (Reny and Newman 2021) and efforts to educate the public around issues of racism triggered heightened focus on CRT.Indeed, as is evident from even a cursory glance at Republican campaign rhetoric and anti-CRT laws in Republican led states, CRT has become a crucial part of the Republican agenda.They signal opposition to CRT, tying it to an array of grievances (e.g., divisiveness, anti-White discrimination).
The extent to which Republican voters have followed suit remains unclear, as do Democratic reactions.Do rank-and-file Republicans strongly oppose CRT, and why?What do they think about CRT as compared to teaching about the impact of racism more generally?Finally, have Democrats done the reverse and come to support CRT in similar ways as they do for teaching about racism in general, thus reacting to the opposing party's stance (Nicholson 2012)?
A substantial literature on framing exists that can help answer these questions (Chong and Druckman 2007;Gubitz et al. 2018).We add to this literature by using the CRT case to demonstrate an elite strategy that increases the strength of a frame.Specifically, Republican elites have invoked CRT as a targeted "empty signifier" that activates a wide range of grievances (beyond race) among rank-and-file Republicans.This negativity leads them to oppose curricula that some Republicans may otherwise support if it was instead described as teaching about the legacies of racism.
We demonstrate the impact of this framing strategy on public opinion, among Republican and Democratic identifiers and among those with high and low levels of negative affect toward Black people.We do so with a large original survey conducted in Fall 2021 among 19,060 American adults.The survey included a pre-registered framing experiment that asked about support for educational curricula that did or did not mention CRT. 1 We augment our experimental approach with an open-ended item asking respondents to articulate what the term "CRT" means to them to demonstrate the valence and breadth of considerations that the frame (CRT) stimulates.These openended responses provide fascinating insight into how terminology with vague meanings (to the individuals) can activate wide-ranging, valenced responses.
We find that Americans overall are much more supportive of teaching how racism continues to impact American society today than teaching CRT in public schools.CRT does indeed work as a strong, negative frame that evokes a broad set of grievances.There are two main reasons for this.First, the modal respondent was not familiar with CRT and was neutral or uncertain regarding whether it should be taught in schools.The lack of knowledge led Democrats, in particular, to equivocate (rather than just do the opposite of what Republican elites signaled).Second, Republican respondents were on average more familiar with the concept, relative to Democrats, and held particularly negative (and wide ranging) attitudes toward the concept.This is clear from the open-ended reactions where Republicans follow elite discourse in connecting CRT to perceived anti-White racism across society and other cultural grievances.Moreover, we find that negative affect toward Black people condition our results: Those with higher anti-Black affect express more opposition to CRT relative to merely teaching about the continued legacy of racism.This greater opposition demonstrates that they view CRT as substantively different from and farther-reaching than teaching about racism's continued impact.
Taken together, these results suggest that backlash to CRT, and its prominence in contemporary political discourse, is not a reflection of where most Americans stand on the issue.Instead, it is likely an elite-driven phenomenon where conservative elites have re-defined CRT to stand in for a variety of cultural and racial grievances for conservative citizens exposed to their cues.This implies that conservative elites have been successful in inducing a minority of the population-e.g., some on the political right-to oppose something about which they ostensibly know little.More generally, the results reveal an effective (strong) framing strategy where one side employs a targeted empty signifier to push their constituents in a preferred direction.When elites employ a frame with vague but clearly valenced meanings, their constituencies seem to follow.

CRT and Its Effect on Education Policy
CRT is an academic movement of scholars who investigate and seek to change the existing power dynamic between race and racism in society (Bell 1980;Crenshaw et al. 1995;Delgado and Stefancic 2011).While proponents exhibit substantial theoretical variation, they agree on a few central premises.Among these are the notion that race is socially constructed (there is no biological basis for what we think of as race), the idea that racism is normalized as part of everyday society (it is systemically entrenched in modern institutions and policies and can be difficult to combat), and the idea that society's dominant groups have little incentive to eliminate racism because the current racial hierarchy serves important material and psychological needs for the majority (Bell 1980(Bell , 1995;;Harris 1995).Other themes in CRT include the idea of intersectionality, or that belonging to multiple oppressed groups is a distinctive experience that is more than just the sum of its parts (Crenshaw 1989).A core theme in CRT is that racism is not just a thing of the past but continues to have an impact on society today.The theory thus emphasizes the existence of systemic or institutional racism (e.g., Bonilla-Silva 2001, 2006).
As noted, the current anti-CRT backlash began with the New York Times' "The 1619 Project," 2 which commemorated the 400th anniversary of the first slaves arriving on American soil in a series of essays reframing U.S. history around slavery and emphasizing the contributions of African Americans (Bailey et al. 2021;Hannah-Jones et al. 2021;Kaplan and Owings 2021).Over time, conservative backlash to "The 1619 Project," anti-bias training, and expanded efforts to teach the legacy of racism in modern American society sparked outcry against CRT, including calls to end CRT teaching in K-12 schools.
Concurrent with CRT's new prominence in political discourse, officials introduced a wave of state and federal attempts to restrict how racism and history are taught in schools and workplaces.Among the most well-known of these efforts was then-President Trump's September 2020 memo and executive order banning certain types of diversity and inclusion training for federal employees and government contractors, including those using or mentioning CRT (characterized as "propaganda"), White privilege, or suggesting that America is a racist country (Executive Order No. 13950 85 FR 60683 2022; Memorandum for the Heads of Executive Departments and Agencies, M-20-34, 2020).Although the order was revoked by the Biden administration (Revocation of Executive Order No. 13950, 2021), many states took action to restrict how history and race are taught in public schools.Between January 2021 and early February 2022, 37 states introduced such bills or took other steps to restrict teaching on these subjects (Schwartz 2022).Politicians such as Florida Governor Ron DeSantis (R) and Virginia Governor Glenn Youngkin (R) attracted national attention by making their opposition to CRT and "woke" ideology central to their policy agendas and electoral campaigns (Barakat and Rankin 2022;Izaguirre 2023).
Much of this anti-CRT backlash mischaracterizes CRT by conflating institutional and individual racism and mis-attributing its prevalence in K-12 classrooms.This has been fueled by conservative media and right-wing political elites (Nelson 2021).For instance, during the summer of 2020, conservative activist Christopher Rufo attended an anti-racial bias training and shared the training materials publicly.In doing so, he argued that under the guise of anti-racism, the training promoted segregation, group-based guilt, and race essentialism against Whites (Wallace-Wells 2021).This led to numerous Rufo appearances in conservative media such as Fox News, as well as endorsement from many Republican elected officials including former president Donald Trump (Kiernan 2020), Virginia governor Glenn Youngkin (Wulfsohn 2021), and Florida governor Ron DeSantis-the latter of whom appointed Rufo to the New College of Florida's Board of Trustees (Strauss 2023).A search of congressional e-newsletters demonstrates that messaging about CRT goes well beyond a single conservative activist and points to party asymmetry in the framing of CRT.Between April 20, 2021 and September 5, 2023, Republican congressional e-newsletters mentioning "CRT" outnumbered Democratic enewsletters that included the term 453-3 (see Figure 1).This is consistent with Deshpande et al. (2023), who find that Republican legislators and conservative media used the term far more frequently than their Democratic and liberal counterparts.CRT thus not only became a prominent talking point on the political right, but it was also a topic that the left generally avoided discussing explicitly.

Expectations for American Views on CRT
To understand the impact of CRT rhetoric, we turn to work on framing effects.A framing effect occurs when a speaker's emphasis on a subset of potentially relevant considerations in describing an issue or event causes individuals to focus on those considerations when constructing their opinions (Druckman 2001: 226-231).Framing effects come in at least two varieties.Equivalency framing effects occur when the use of different, but logically equivalent, words or phrases (e.g., 90% fatfree versus 10% fat) lead people to attend to those words/phrases and this alters their preferences.Emphasis framing effects occur when a speaker's accentuation of relevant, substantively distinct aspects of an issue/event (e.g., a hate group rally as free speech versus public safety; Nelson, Clawson, and Oxley 1997) leads the audience to base their evaluation on those aspects.Emphasis framing is similar to Riker's (1986) heresthetics, which refers (in part) to changing the dimensions on which decisions are made.
The distinction between types of framing effects is not always clear, particularly when it comes to using similar words. 2 As Druckman (2001: 239) explains, "many treat question wording effects as examples of equivalency framing : : : [However, this requires that] the alternative wordings are in fact logically or objectively equivalent : : : While some examples clearly satisfy this requirement, others are less clear.For instance, some point to a wording effect where people favor increased government assistance to 'poor people,' but then oppose it for 'people on welfare' : : : Yet, it seems quite reasonable to argue that 'poor people' include many others besides 'people on welfare.'"As with this example, identifying a logical match or perfectly synonymous term for CRT, given it encompasses a wide range of ideas, is difficult if not impossible.As such, invoking CRT as opposed to similar terminology-"legacies of racism"-is best considered a type of emphasis framing where the terms are not completely synonymous (i.e., they highlight distinct outlooks or curricula). 3Our acute interest lies in how support for teaching students about race differs when this education is framed as CRT versus framed as teaching about legacies of racism.
A central question for work on emphasis framing effects concerns what makes a particular frame strong, with the considerations being seen as highly applicable in opinion construction (Chong and Druckman 2007, 111;Arceneaux 2012;Gubitz et al. 2018, 43).The case of CRT exemplifies an intriguing strategy for doing so.Specifically, we suspect that "CRT" acts as an "empty signifier" (Chandler and Munday 2011;Laclau 2005): a signifier divorced from its original meaning and replaced with a set of vaguer, more complex meanings.This rhetorical approach can generate frame strength when it is aimed at a constituency for whom the signifier is valenced to encompass meanings that are "pro" or "con" on a given topic.We think of this as using a "targeted empty signifier" to increase frame strength.In the case of CRT, Republican and conservative speakers invoke the term when talking about race, particularly in education, to generate opposition to race-based curricula.The frame becomes stronger because the signifier, CRT, activates broader cultural grievances (beyond race).
Indeed, Rufo described exactly how "CRT" would come to function as an empty signifier in conservative discourse.Discussing the concept with James Lindsay, a conservative pundit, on Twitter in March of 2021, Rufo stated that "We will eventually turn [Critical Race Theory] toxic, as we put all of the various cultural insanities under that brand category.The goal is to have the public read something crazy in the newspaper and immediately think 'critical race theory.'We have decodified the term and will recodify it to annex the entire range of cultural constructions that are unpopular with Americans" (emphasis added; Rufo 2021a, 2021b). 4Republican elites have sent clear, if imprecise, messages about CRT: Signaling that it is divisive and discriminatory against Whites.Consequently, the audience, here Republicans, thinks of multiple negative considerations when it comes to the curriculum and opposes it.In essence, the strategy strengthens the frame such that there will be a stronger negative reaction among Republicans to "CRT" than "legacies of racism" when it comes to classroom instruction.This leads to a set of hypotheses.
First, research shows that individuals with more hostile views toward African Americans are also more likely to doubt that racism, of any sort, remains a potent force in American society (e.g., Todd et al. 2012;Van Dijk 1992) and thus may view education about the effects of racism as unnecessary (e.g., Bobocel et al. 1998).This may be due to a lack of importance these individuals place on racial differences, or the belief that racial disparities are driven by a lack of effort rather than the effects of discrimination (Pager 2007).Consequently, those with racial attitudes more hostile toward African Americans will likely express greater opposition to teaching about legacies of racism in schools.Such low baseline support may make differences in support for teaching CRT relative to teaching about racism more muted.Put another way, because those with higher anti-Black affect will oppose both teaching about legacies of racism and CRT, the two frames then will exhibit similar effects.We thus expect small differences between the frames among those with high anti-Black affect, relative to the effects among those with lower anti-Black affect where the frame differences will be more salient.
H1: We expect smaller differences in support for teaching CRT relative to teaching about the legacy of racism in public schools among those higher in anti-Black affect (compared to those lower in anti-Black affect), all else constant.
As mentioned, the CRT frame (signifier) should resonate with Republicans.Theories of learning and persuasion make clear that Republicans will thus follow the credible frame from their party, consider wider grievances, and oppose CRT (Lupia andMcCubbins 1998, Druckman 2001).We should therefore see a large framing effect.What about Democrats?Interestingly, as is made clear by Figure 1, Democratic elites have been much quieter on the issue, not offering a decisive position about CRT specifically.As Deshpande et al. (2023, 1) state "Republican legislators and conservative media's use of the term 'critical race theory' dwarfed that of Democratic legislators and liberal media, respectively."Some work suggests that, in asymmetric cases, partisans simply do the opposite of what the other party suggests (follow an "anti-cue") (Nicholson 2012, Druckman et al. 2024).Yet here we do not expect that to occur.The "anti-cue" dynamic focuses on established issues where individuals can likely infer the implications of the frame.Yet, we suspect that most individuals, including Democrats, do not know what CRT is (Safarpour et al. 2021) and thus lack any information to otherwise take a position.They thus will equivocate, given the lack of a clear Democratic frame/position.Consequently, we expect Republicans to demonstrate a large difference depending on whether they are queried about CRT or the legacies of racism (given elite stances) whereas Democrats will show less since they will be unclear on where to stand regarding CRT. 5 H2: The difference in support for teaching CRT relative to teaching the impact of racism will be greatest among Republicans (relative to Democrats), all else constant.
Underlying H2 is the presumption that Republicans receive elite rhetoric concerning CRT-more so than Democrats do.Given evidence of media echo chambers, particularly on conservative cable television where much CRT coverage has occurred (Broockman and Kalla 2023), it follows that Republicans will be more familiar with CRT than Democrats.Indeed, CRT has been consistently presented in what Boydstun (2013) calls "alarm model" offering a skewed portrait of its relevance.Moreover, the basic point of the targeted empty signifier framing tactic is to invoke valenced considerations, in this case, negative.Republicans thus should express more negative definitions of CRT.
H3: Republicans will report being more familiar with CRT than will Democrats, all else constant (since it is a prevalent part of Republican rhetoric).
H4: Republicans will express broadly negative views of CRT when asked to define the concept in their own words, all else constant (relative to Democrats).
Finally, since we expect Republicans to be very opposed to teaching CRT relative to legacies of racism, and Democrats to be ambivalent between the two, the aggregate effect will be greater support for legacies of racism versus CRT.
H5: Individuals will express greater support for teaching the continued impact of racism than teaching CRT in public schools, all else constant (since CRT is bundled to encompass more).
Lastly, though not pre-registered, we also examine racial differences in attitudes, comparing results among Whites, Blacks, Asians, and Hispanic respondents.

Data and Methods
To determine how Americans understand CRT, we recruited 19,060 adult respondents to complete a survey.Data were collected between November 3, 2021 and December 2, 2021 by PureSpectrum via an online, nonprobability sample, with state-level representative quotas for race/ethnicity, age, and gender.Although nonprobability samples have well-documented limitations, research shows that similar online nonprobability platforms perform well relative to probability samples and yield valid inferences, particularly for experimental research (e.g., Mullinix et al. 2015;Coppock et al. 2018).Our pre-registered exclusion criteria were followed, eliminating respondents who failed standard survey attention check questions, and quality checks (e.g., straight lining and speeding).Overall, roughly 28% of those who started the survey were eliminated for these reasons.This is a standard exclusion rate for this vendor and for similar nonprobability sample suppliers. 6 The final, unweighted sample was diverse: 17% were 18-29, 39% 30-49, 23% ages 50-64, and 21% were 65 or older.Roughly half (49%) identified as Democrats or Democratic-leaning independents, 31% as Republicans/Republican-leaning independents, and 19% as pure independents.The sample was 71% White, 12% Black, 9% Hispanic, 6% Asian, and 2% some other race.About three-quarters (74%) attended at least some college or more and 26% held a high school degree or less.The sample skewed female (61%).Regionally, 39% of respondents were from the South, 24% from the Midwest, 21% from the West, and 17% from the Northeast.A summary of unweighted and weighted sample descriptives is in Table 1.The weights are constructed using demographic characteristics to match the national U.S. population with respect to race/ethnicity, age, gender, education, and residence in urban, suburban, or rural areas.
The CRT survey was embedded in a broader survey focused on COVID-19 attitudes and behaviors.The relevant items proceeded as follows: All respondents first answered a series of questions about their attitudes and behaviors relating to COVID-19, and their attitudes toward various groups in American society, their opinions about how history is currently being taught in public schools, and a series of questions on their attitudes toward CRT.After rating their level of concern with how history is taught in public schools, respondents were randomly assigned to one of two conditions.In the "definition" condition, 9,535 respondents rated whether they favor, oppose, or neither favor nor oppose teaching how "racism continues to impact American society today" in public schools (a statement consistent but more limited than the CRT movement).In the "CRT" condition, 9,525 respondents rated whether they favor, oppose, or neither favor nor oppose teaching "CRT" in public schools.As mentioned, teaching CRT is not explicitly the same thing as merely teaching the impact of racism, as we expect respondents to view teaching the impact of racism much more narrowly than teaching CRT and, indeed, may differ in what they define "racism" to mean.After rating their support for teaching CRT or the legacy of racism, all respondents rated how familiar they were with CRT and described what CRT means to them in an open-ended question.Exact question wording is in Appendix A. Results reported in the main text of the paper are unweighted, and weighted estimates are in the Appendix.There are no significant differences between unweighted and weighted estimates.Regression tables are in the Appendix.

Experimental Results
We start by evaluating H1 that those with more anti-Black affect will exhibit less of a differential reaction to teaching about CRT relative to teaching about legacies of racism.We do so by regressing our dependent variable on a binary indicator for assignment to the treatment condition (CRT), respondent's anti-Black affect, and the interaction of the two.The dependent variable is coded 1 if the respondent supported teaching the concept in schools (either CRT of the legacy of racism), and 0 if they opposed or neither supported nor opposed teaching the concept.Anti-Black affect is computed as the difference in feeling thermometer ratings given to Whites versus Blacks and scaled between 0 and 1, with scores near 0 indicating a preference toward Blacks over Whites, and scores near 1 indicating preference for Whites over Blacks, with 0.5 indicating no preference.As shown in Figure 2-and consistent with H1-treatment effects are smallest among those with higher anti-Black affect (Average Marginal Effect (AME) = −0.14among those with scores of 1) and greatest among those with lower anti-Black affect (AME = −0.35among those with scores of 0).Results computed with an ordinal dependent variable (using all the response categories) are consistent and shown in Appendix Table B1 and Figure B1.
To better illustrate where the shifts occur, Figure 3 displays the percentage point differences in attitude toward teaching CRT relative to teaching the impact of racism by quartiles of negative affect.Negative values indicate respondents rated teaching CRT lower than the impact of racism in public schools.Positive differences indicate CRT was rated higher.As shown in Figure 3, the larger treatment effects among those with the lowest anti-Black affect (scores near 0) are driven by large differences in uncertainty rather than notably greater outright opposition to teaching CRT. 8 Indeed, the gap in opposition between teaching CRT versus legacies of racism is smallest among those with the lowest anti-Black affect.As anticipated, the difference in support for teaching CRT versus legacies of racism is lowest among those with the highest anti-Black affect (scores near 1), due to low support for teaching either concept.Notes: OLS estimates.Error bars are 95% confidence intervals.Dependent variable is support for teaching "critical race theory (CRT)"/"How racism continues to impact American society today" in public schools.Dependent variable is coded 1 = strongly/somewhat support, 0 = strongly/somewhat oppose and neither support nor oppose.
We next evaluate H2, which predicted larger treatment effects among Republicans versus Democrats.We evaluate H2 by regressing teaching support on partisanship, a treatment condition indicator (CRT) and an interaction between the two.Results are shown in Figure 4.Although we predicted in H2 that the treatment would have the greatest impact on Republican identifiers, we instead find the greatest treatment effects among Democrats and Independents, with conditional average treatment effects weakest among Republicans.While surprising, it also is explicable, upon close examination.Specifically, the larger effect stems from greater uncertainty among Democrats, likely due to less familiarity with CRT.
While 24% of Republicans support teaching about the legacies of racism, only 8% support teaching CRT (a 16-percentage point gap).In contrast, 73% of Democrats support teaching about the legacies of racism, yet just 45% support teaching CRT (a 28-percentage point gap).The larger shift among Democrats is the result of a greater share of Democrats who neither support nor oppose teaching CRT relative to Notes: Negative differences indicate respondents rated CRT lower than for teaching the impact of racism in public schools.Positive differences indicate CRT rated higher.
teaching about the legacies of racism (45% vs 18%) rather than increased opposition (9% opposed teaching about racism, and 11% opposed teaching CRT).In contrast, effects among Republicans are primarily driven by greater opposition to CRT (61% oppose) relative to teaching about the legacies of racism (48% oppose).The results using an ordered probit model are consistent (see Appendix Figure B2). 9Thus, the clear cues from Republican elites seem to impact Republicans in the public, as we expected.While we also thought that Democrats would express uncertainty about CRT, we did not forecast that such uncertainty would generate as much lukewarm support as it did (relative to teaching about legacies of racism).In short, the results cohere with our theoretical expectation that Democrats would not simply take an anti-cue, but the extent to which uncertainty generates neutral opinions was unexpected.Regardless, the results accord with Deshpande et al.'s (2023) finding regarding the partisan asymmetries around CRT: Republicans follow their elites while Democrats mostly equivocate.Notes: OLS estimates.Error bars are 95% confidence intervals.Dependent variable is support for teaching "critical race theory (CRT)"/"How racism continues to impact American society today" in public schools.Dependent variable is coded 1 = strongly/somewhat support, 0 = strongly/somewhat oppose and neither support nor oppose.
Notably, deviations by party within each racial/ethnic group (shown in Figure 5) are relatively minor.However, effects among Hispanic Republicans were significantly greater than effects among White Republicans.This is due to significantly greater support for teaching the legacy of racism among Hispanic Republicans relative to White Republicans and similar levels of support for teaching CRT among both groups.

Predictors of Familiarity with CRT
We next turn to H3, which suggested increased Republican familiarity with CRT, relative to that of Democrats.Overall, we find 68% of respondents said they were "not very" or "not at all" familiar with CRT, including 27% who said they were "not at all" familiar with the concept.This is consistent with the responses to the openended question (as we detail below), with over 41% saying they do not know what CRT is.Predictors of familiarity with CRT are shown in Figure 6.As expected in H3 and consistent with the greater uncertainty among Democratic respondents observed in the experimental results, Republicans expressed greater familiarity with CRT than did Democrats net of other factors.This speaks to the mechanism we posited underlying Republicans' reactions to CRT-that is, they are asymmetrically exposed to it in their information ecosystem (e.g., see Figure 1).These results are unchanged when accounting for the possibility that assignment to the CRT treatment condition may have primed familiarity for some respondents but not others, as shown in Figure 6.
In Their Own Words: How Americans Understand CRT  theory, regarding using an empty signifier as a framing strategy, suggests that these negative reactions should encompass a wide range of grievances.
To analyze the answers, we trained undergraduate students who were unaware of the researchers' hypotheses to code the responses based on a set of categories developed by the authors.These categories captured the valence of respondents' answers (e.g., whether their understanding of CRT was positive, negative, or neutral) as well as the content of these responses (e.g., among negative responses, whether the respondent said that CRT was anti-White, divisive, etc.).The intercoder reliability was very high (Cohen's kappa = 0.73, percent agreement = 82%). 10 Responses to the open-ended question were given up to three codes depending on the length of respondents' answers, with the order of codes determined by the order in which respondents mentioned various concepts. 11A breakdown of the three codes is shown in Figure 7. Example responses for some of the major categories include "I don't really know" for the Don't Know category; "I think it is about how blacks were treated in our country over time, from slavery to the present" or "American history accurately reflecting racism and the historical mistreatment of African Americans" for the Neutral/positive category; and "It means my family, who are white, are in league with the devil and they are NOT" or "All I know is that racism only exists when people keep bringing it up" for the Negative category.Other example responses of prominent categories include College ("CRT is NOT taught in grades K-12.It is taught at college level and is a discussion on race issues in America") and Partisan ("its [sic] a cheap ploy republicans use even if they have no idea what it is").See Appendix Table E1 for detailed information about the full coding scheme, including more example responses.Overall, a plurality of respondents (41%) indicated they did not know what CRT was, while smaller shares gave answers that were coded as either neutral or positive interpretations (28%) or negative (14%) interpretations.Democrats were significantly more likely to express uncertainty than Republicans (40% versus 36%, difference=4.3points, p < 0.001), consistent with the familiarity result discussed earlier.It is possible that the experimental conditions influenced open-ended responses.However, we find only a five-percentage point difference in the share of do not know responses between conditions, with those in the CRT condition less likely to report being unfamiliar with the concept.Differences between conditions for the other open-ended categories were statistically and/or substantively insignificant.Of those who provided neutral or positive responses, about onequarter stated that CRT meant teaching about how racism impacts people or society today, 16% stated CRT reflects the historical treatment of minority groups, 14% noted that CRT has to do with institutional or systemic racism, and 14% defined the concept as teaching the truth about history or the way society operates.Smaller shares recognized the CRT concept of intersectionality (1%), that CRT is an activist endeavor that seeks to promote equality (6%), or that it teaches how race is socially constructed (3%).A breakdown of responses with examples in the neutral/positive category is shown in Table 2.Among those with negative interpretations of CRT, about one-third gave generic negative responses, 22% noted that CRT is anti-White, 17% said the concept is racist (without necessarily specifying it as anti-White), and 15% said CRT is divisive.Smaller shares remarked how Black people are complaining or asking for special favors in society (5%), considered CRT propaganda or brain-washing (4%), noted that our society is post-racial or that racism is only real when discussed (3%), or that CRT is a Marxist or communist ideology (3%).A breakdown of negative responses with examples is shown in Table 3.
To evaluate whether Republicans were significantly more likely than Democrats to define CRT as broadly negative (H4), we evaluated the share of partisans who mentioned specific negative concepts when defining CRT.As shown in Figure 8, Republicans and Republican-leaning independents are significantly more likely than Democrats and Democratic-leaning independents to say that CRT is racist or anti-White, divisive, partisan, propaganda, represents Black people asking for special favors or "playing the victim," and that CRT is taught to children.Overall, small shares of both parties stated that CRT was aimed at children, with just 3% of Democrats and 6% of Republicans mentioning children in their responses.Conversely, roughly 17% of Republicans noted that CRT is racist or anti-White.This is consistent with our expectation that opposition to CRT is not solely rooted in misinformation about the concept being taught to children but rather reflects a  much broader grievance that society is becoming anti-White.Importantly, this supports the idea that CRT functions as an empty signifier on the political right: Republicans follow frames and are more likely to oppose CRT, and they express relatively more confidence in their knowledge about it.Yet that knowledge envelopes a range of ideas beyond CRT and reflective of other ideas and concerns about society (e.g., that CRT implies individual anti-White attributions). 12 Although not pre-registered, the prevalence of open-ended responses defining CRT as racist and anti-White prompted us to examine the association between these responses and affinity toward White people.We expected that CRT was viewed as threatening among White respondents with stronger White identities.Among White respondents, those defining CRT as racist or anti-White rated White people significantly higher on a feeling thermometer than did Whites who gave some other interpretation of CRT.The average feeling thermometer rating with 95% confidence intervals for these two groups is shown in Figure 9.We also examined differences in open-ended responses by race.These results (shown in Appendix Figure B5) indicate that non-White respondents are more likely to define CRT in neutral or positive ways.Overall, Black respondents are most likely to define CRT in neutral or positive ways.Hispanic and Asian respondents are also more likely than White and Black respondents to say they do not know what CRT is.The analysis of open-ended responses by race was not pre-registered.

Overall Opinions
We predicted in H5 that the aggregation of the processes documented so far, with Republicans opposed to CRT and Democrats ambivalent, would mean overall less support for teaching CRT relative to legacies of racism.This is exactly what we find: Respondents were significantly less supportive of teaching CRT than about the legacies of racism in public schools (AME = −0.25,p < 0.001).Results using an ordinal coding of the dependent variable are consistent and shown in Appendix Table B1.
Beyond evaluating our pre-registered expectations, we also explore the impact of respondents' self-identified race and ethnicity. 13The share of respondents who support teaching the legacy of racism and teaching CRT overall and among each racial group is shown in Figure 10.
Surprisingly, the effect of the CRT condition is greatest among Black respondents (−0.29) and significantly greater than the effect among Whites (−0.24).This is primarily due to strong support for teaching the legacy of racism in schools among Black people (about 75% support), whereas only about 45% support teaching CRT, which is equivalent to the share of Whites who support teaching the legacy of racism (48%).Effects of the CRT condition among other racial minorities in our sample (Hispanics, Asians, and those who identify as some other race) are roughly equivalent to the effects among Black respondents.However, the differences in the effect of the CRT treatment between Hispanics, Asians, and Whites are not significant at conventional levels.Similar shares of Hispanic and Asians respondents support teaching the legacy of racism (62% and 60%, respectively) and for teaching CRT (34% and 31%, respectively).
These results suggest that, overall, people of color were least likely to receive or believe the negative CRT frames espoused by Republican elites.Given our expectations regarding the Republican CRT frame as capturing broadly racial and cultural grievances (which may be influenced by perceived racial threat), it makes sense that groups for whom these grievances are least likely to appeal (racial minorities) are also impacted by these frames least.However, as discussed above, once results are broken down by partisanship, Hispanic Republicans appear most impacted by the anti-CRT messaging.This is due to significantly greater support for teaching about the legacies of racism among Hispanic Republicans and greater uncertainty regarding CRT, relative to non-Hispanic White Republicans for whom support for teaching legacies of racism was lower.

Discussion
Our results reveal not only the dynamics behind attitudes about CRT but also about an underappreciated framing strategy.In terms of the former, we find that respondents overall are more willing to express greater support for "teaching the impact of racism" in public schools than for "CRT."However, the extent to which those different frames matter depends on racial predispositions and partisanship.Those with high levels of anti-Black affect exhibit the least sensitivity to the competing characterizations.This is because they generally oppose teaching about the legacy of racism in public schools regardless of how it is framed whereas those with the lowest negative affect express significantly greater uncertainty about teaching CRT.Republicans, unexpectedly, are less reactive to the framing effect than Democrats.Republicans express scant support for both teaching about legacies of racism and even more so to CRT.Democrats, by contrast, overwhelmingly support teaching about the legacies of racism but are highly uncertain and ambivalent about CRT (creating a larger gap between the two curricula).Underlying these partisan differences are greater familiarity with and strong negative feelings regarding CRT among Republicans.
Taken together, the findings demonstrate that opposition to CRT is more than simply the denial of racism, nor is it synonymous with general opposition to teaching about the continued impact of racism.Rather, for a portion of the public The Journal of Race, Ethnicity, and Politics (Republicans) the specific phrase "CRT" taps broader concerns and anxieties related to race and culture.The use of CRT to express these broader grievances likely reflects rhetoric by Republican and right-wing elites.A key question going forward is whether Democrats will gain greater familiarity.For this reason, future studies should delve into the direction and content of CRT attitudes in the public over time.Are Americans in the process of developing more substantive opinions on CRT, and what is the valence of these opinions?Beyond demonstrating the opinion dynamics of CRT specifically, the results suggest that empty signifiers are underappreciated as a targeted framing strategy.Not only does it lead the targeted constituency to hold a position in line with the elites' preferences (opposition to race based curricula), but it may effectively allow them to "talk past" the other side, which is engaging with the issue on fundamentally different, narrower terms.As such, an empty signifier constitutes a strong framing strategy.An open question is what works best as a counter-response.
Beyond CRT, other issues may be particularly susceptible to similar framing strategies.It is possible that racialized and cultural threat topics are particularly susceptible to becoming empty signifiers, as treatment of one minority group may be used to evoke broad concerns over social status and group-based hierarchy among majority group members.Democratic politicians may also be reluctant to discuss such issues openly, out of fear of alienating swing voters for whom such topics constitute "wedge issues," thereby leaving the Republican party with a messaging monopoly in the public discourse.
Supplementary material.The supplementary material for this article can be found at https://doi.org/10.1017/rep.2023.39 2 Various examples of similar wordings that lead to distinct opinions include "Obamacare" versus "the Affordable Care Act" (Holl, Niederdeppe, and Schuldt 2018), rights for "gays and lesbians" versus "homosexuals" (Smith et al. 2018), beliefs regarding "global warming" versus "climate change" (Zaval et al. 2014;Schuldt et al. 2017), and "poor people" versus " people on welfare" (Green and Kern 2012).While some of these words could be presumed to be equivalent, others such as the last example are clearly not (e.g., there are poor people who are not on welfare).CRT, given its complexity, falls into this latter category (Druckman 2001).3 Importantly, the legacies of racism does not reference systemic racism and thus could be read as individual racial attitudes.It does, however, capture the key CRT tenet on the modern persistence of racism.The use of "structural racism" would be more ambiguous as it is an inherent part of CRT. 4 Of course, this is not to attribute the entirety of public discourse surrounding a given political concept to one specific actor.Rather, we see Rufo as succinctly articulating the process by which the concept we are interested in, "critical race theory," has come to take on a wider-ranging and more negative meaning among those receptive to this conservative framing.5 Deshpande et al. (2023) show that CRT has an asymmetric effect in mobilizing only Republicans (and not Democrats to defend it).Also, while some suggest that information party endorsements operate regardless of information content, recent work makes clear that content matters (Tappin et al. 2023).6 Personal communication with authors of Mullinix et al. (2015).7 As pre-registered, our main analysis presents results among all respondents.Results for only White respondents are consistent and shown in Appendix C. 8 The descriptive analysis shown in Figure 3 was not pre-registered.It is included along with the preregistered analysis (Figure 2) to demonstrate in greater detail where the shifts in attitudes occur.9 We did not pre-register the ordered probit model, opting in our pre-registration to use the simpler linear probability model.However, given the pattern of results seen in the cross tabulations and the results of the pre-registered analysis inconsistent with theoretical expectations, we ran an ordered probit model as well.10 We did not pre-register that we would report inter-coder reliability statistics.11 Pre-registered quantitative analyses of these open-ended responses using unsupervised machine learning are reported in Appendix D and show substantively similar results as those reported here based on hand-labeled responses.12 We find in supplemental analyses that those who view Fox News more often exhibit greater familiarity with and more negative responses to CRT.This is consistent with our argument that elite frames shape reactions.This supplemental analysis was not pre-registered.13 We thank an anonymous reviewer for this suggestion.These analyses were not pre-registered.

Figure 2 .
Figure 2. Conditional average treatment effect of "CRT condition" by negative affect.

Figure 3 .
Figure 3. Difference in attitude toward teaching CRT versus impact of racism by negative affect quartiles.

Figure 4 .
Figure 4. Conditional average treatment effect of "CRT condition" by party identification (binary DV).

Figure 5 .
Figure 5. Conditional average treatment effects by respondent's race/ethnicity (overall and by party).Note: Error bars are 95% confidence intervals.

Figure 6 .
Figure 6.Predictors of familiarity with CRT.Notes: OLS estimates.Error bars are 95% confidence intervals.The robustness check model includes a control for experimental treatment and was not pre-registered.

Figure 9 .
Figure 9. Average White feeling thermometer by racist/anti-White CRT meaning.Notes: Unweighted estimates among White respondents only.Error bars are 95% confidence intervals.

Table 1 .
Sample characteristics

Table 3 .
Breakdown of responses coded "negative" (code 1) Support for teaching about the legacy of racism and CRT in public schools overall and by race.