Critical Race Theory: How Policy Language Differentially Engages Symbolic Racism and Partisanship

Abstract

Recent years have seen a marked shift in the salience and politicization of any incorporation of race into teaching at the elementary and secondary levels. “Critical race theory” (CRT) has become a prominent feature of the current debate, even as there is a good deal of misunderstanding about what CRT actually is. Drawing on a pre-registered survey experiment, we consider the impact of the phrase “critical race theory” in activating both racial biases and partisan identity. Our expectation was that CRT would tend to activate partisanship independent of symbolic racism. Results suggest otherwise: where support for culturally relevant pedagogy is concerned, CRT appears to engage partisanship particularly amongst those who exhibit high levels of symbolic racism.

The past twenty years have seen the development of a rich literature on culturally relevant pedagogy in elementary and secondary education. This literature emphasizes the importance of teaching children to “accept and affirm their cultural identity while developing critical perspectives that challenge inequities that schools (and other institutions) perpetuate” (Ladson-Billings 1995a, 469). Key aspects of culturally relevant pedagogy include motivating students to choose academic success, teaching with culturally relevant language and examples, and engaging students in the critique of norms and dated knowledge (Ladson-Billings 1995b). The idea is related to, but also distinguishable from, critical race theory (CRT). CRT emerged as a legal theory in the 1970s; it is a disruptive critique of the typical liberal agenda focused on civil rights legislation (for details on CRT, see, e.g., Bell 1995). Critical race theory tends to highlight the structural aspects of racism, and the ways in which racism continues to permeate the everyday life of racialized minorities despite the removal of explicitly racist laws and less explicitly racist beliefs held by the public (for a review, see Crenshaw 2011).

Until recently, discussion related to culturally relevant pedagogy and CRT took place primarily in journals of education and law, respectively. Recent years, however, have seen a marked shift in the salience and politicization of any incorporation of race into teaching at the elementary and secondary levels (for an overview, see Kaplan and Owings 2021). And although CRT itself began as a legal theory, the phrase “critical race theory” has played a central, mobilizing role in the current political debate. For example, Florida governor Ron DeSantis, who has announced a bid for the 2024 Republican presidential nomination, signed into law in 2022 the so-called Stop WOKE Act whose explicit aim is to prevent the teaching of CRT in schools. DeSantis is quoted as saying “in Florida we are taking a stand against the state-sanctioned racism that is critical race theory. We won’t allow Florida tax dollars to be spent teaching kids to hate our country or to hate each other.” ¹ Florida is just one of over 30 states that have proposed or passed legislation or executive orders that seeks to ban or limit the teaching of CRT in schools (Green 2022).

It is in this context that our current paper seeks to understand the power of the words “critical race theory” in current popular debate and in primary election campaigns to mobilize and divide public opinion. Our starting point is that the argument surrounding the teaching of CRT in American schools is based in part on a misunderstanding of what CRT actually is. The role that CRT plays in the debate is, we suspect, primarily symbolic—CRT is a cue, signaling a vaguely-defined (mostly, incorrectly-defined) liberal agenda on race. We regard this as one instance in which language (or “issue framing”) may be especially impactful on attitude measurement, and perhaps on actual attitudes as well. Our study accordingly sets out to explore the impact that different descriptions of culturally relevant pedagogy, including CRT, may have on public support for teaching about race in public schools.

We report results here from a pre-registered survey experiment in which we describe versions of culturally relevant pedagogy differently and observe differences in policy support moderated by symbolic racism (SR), partisan identification (PID), and white identity (WID). Our logic was initially motivated by the suspicion that the phrase “critical race theory” is especially powerful in activating not only racial biases, but also partisan identity, independent of racial biases. Our interest is accordingly not so much on the direct impact of issue framing, but rather on the ways in which partisanship and symbolic racism are activated using “CRT” versus other possible descriptions of teaching about race and racism in U.S. schools.

As we shall see, results are not perfectly in line with our expectations. It is nevertheless clear that the “CRT” language produces a marked decrease in support for teaching about race; and that CRT may engage a combination of partisanship and symbolic racism more powerfully than some other descriptions of—to most audiences, at least—roughly equivalent phenomena.

Critical Race Theory, Partisanship and White Identity

CRT became a highly salient phrase, particularly amongst right-wing commentators, following the publication and subsequent controversies surrounding the 1619 Project, a journalistic documentary series published by the New York Times Magazine in August 2019 and accompanied by educational programs developed by the Pulitzer Center for K–12 classrooms. The Black Lives Matters protests in 2020 further heightened Americans’ attention to issues of race and civil rights. Relatedly, in June and July of 2021, Fox News mentioned the phrase “critical race theory” 1,914 times (Barr 2021). This amounts to more than 30 mentions per day.

This heightened attention has also made evident large partisan divides on issues of race and racism. For example, a 2021 YouGov poll found that only 48% of Republicans regarded racism as somewhat of a problem or a big problem, compared to 95% of Democrats (Frankovic 2021). A 2021 Reuters/Ipsos Poll found that 85% of Democrats supported teaching high school students about the impacts of racism, while only 58% of Republicans felt the same (Duran and Jackson 2021). Such partisan divides should not be surprising, given the increasing polarization on issues related to race in the United States that emerged during the Obama administration and solidified under Trump (Knuckley 2011; Tesler 2016; Abramowitz and McCoy 2019). Such partisan divides on the racial issues make for fertile ground for elites to mobilize on the issue.

Indeed, public discussion and media coverage have been both affected by and reflected in policymaking. Shortly after CRT emerged as a topic of conversation in public debate, President Trump issued an executive order in 2020 ceasing the teaching of CRT in training in federal agencies, describing it as a “divisive concept” (Trump 2020). Many conservatives then mobilized to prevent the teaching of CRT in schools (Borter 2021). In the latter half of 2021, several state legislatures passed bills prohibiting the teaching of critical race theory in schools, including those in South Carolina and New Hampshire. Only a few of those bills directly mention the words “critical race theory.” Even so, most indicate that any discussion or training surrounding ideas of privilege, implicit and explicit bias, and oppression are not allowed, with consequences such as fines and loss of state funding (Ray and Gibbons 2021). In early 2022, bills directed at critical race theory became a prominent element of several ongoing primary campaigns. In the Wisconsin gubernatorial election, for instance, Republican candidates have mobilized around Governor Evers’ veto of a bill that banned the teaching of (what they consider) CRT (Edelman 2022). And most recently, the Florida Department of Education rejected an AP African American studies course because of the “woke indoctrination” that violates the law signed into legislation by Republican Governor DeSantis that regulates teaching about race in the classroom (Fawcett and Hartocollis 2023)

Bills that do not explicitly mention CRT—and even those that do—often contain language pertaining to feelings of remedying white guilt and white blame. Indeed, bills in Michigan, Missouri, and New Jersey, for example, all mention that schools and teachers should avoid imposing any “psychological distress” or “guilt” on students by discussing the actions of people of certain races or sexes in the past (H.B. 5097 2021 for Michigan; S B. 22 2022 for Missouri; S.B. 2685 2022 for New Jersey). In Virginia, H.B. 787 (2022) states that it would be unlawful for teachers to promote or teach that any member of a race or sex “bears responsibility for actions committed in the past by other members of the same race or sex.” Although not all these bills have passed, their introduction has been a topic of coverage in local and national news.

We regard these recent developments as a clear example of the ways in which language can be central to both the distribution and nature of public attitudes and the construction of policy. There is of course a rich literature exploring the impact of “framing” on public attitudes (e.g., Chong and Druckman 2007; Druckman 2004; Iyengar 1991; Nelson, Oxley, and Clawson 1997; Scheufele and Tewksbury 2007); and a corresponding literature demonstrating the influence of framing and “problem definition” on policy development (e.g., Hulst and Yanow 2016; Rochefot and Cobb 1993; Stone 1989). Definitions of “framing” have varied considerably, to be sure (e.g., Entman 1993; Scheufele and Iyengar 2012). But much of the existing literature shares the belief that the language used to describe political issues can have significant effects on the ways which publics and governments respond—a belief that is roughly in line with Edelman’s (1985) provocative argument about the centrality of language in politics.

Our objective here is thus to explore the way in which the now-highly-politicized language surrounding culturally relevant pedagogy matters for public attitudes. And we regard our analysis as being related to the literatures on language and framing in politics, but especially also to work that has more narrowly focused on the role of elite cues in shifting or mobilizing public opinion (e.g., Hopkins 2018; Amsalem and Ziozner 2022). In this paper, our interest is in the ways in which attitudes shift with the use of the term “CRT” relative to other possibilities.

Discussions of race in education are likely to prime racial considerations (for a recent review of racial priming, see Stephens-Dougan, 2021). Yet we suspect that the term CRT will be more likely to also activate partisan considerations. Indeed, as we have noted, partisanship has played a key role in current debates around culturally relevant pedagogy in education, with Republican elected officials, candidates, and media personalities railing against teaching of this kind in schools. Existing literature suggests that partisanship can play a key role in the reception of information and attitude formation (Taber and Lodge 2006), particularly when elites are clearly polarized on the issue (Druckman et al. 2013). This is in part because people are psychologically motivated by directional goals to confirm existing attitudes and social identities (Kunda 1990). Partisanship should be more important in a CRT frame given the polarized nature of the debate, and the clear partisan cues that its use may invoke.

As noted, racial attitudes have become increasingly intertwined with partisanship in the U.S. context. While it makes sense that racial attitudes would affect how people react to issues related to the teaching of race in schools, our argument is that partisan framing of the issue should make partisanship matter to a greater extent. There is, of course, debate about whether racial attitudes in the current American context are independent from partisan preferences. Some have argued that partisanship has become so intertwined with racial attitudes that they are “inseparable” (Westwood and Peterson 2022). In a recent study of framing effects on support for Black Lives Matter (BLM), Drakulich and Denver (2022) found that partisans were largely immune from framing effects because the issue itself had become so partisan. We suspect that while teaching about race will activate racial attitudes regardless of framing, the term CRT has become so politicized that it will make partisanship more salient.

Yet partisanship is not the only identity at play in the debate about culturally relevant pedagogy. The argument that CRT is an ideology that is racially divisive, particularly when it is understood to place blame on white people, may lead to the activation of white identities. Jardina’s work (2019, 2021) has shown that white identity is related to but distinct from hostility toward racial outgroups. This view is consistent with research in social psychology that distinguishes ingroup identities from outgroup animosities (Brewer 1999). Appeals to white identity are clearly present in the current framing of CRT. Fox News personality Tucker Carlson, who has notably spoken out against CRT on his television program numerous times, explicitly frames CRT as threatening to whiteness and an attack on those who are white. He has stated that one must be “brave” to speak out against CRT, and that “if you’re a straight white American, even if you’re a very small child, you’re guilty. It’s your fault. You’re a bad person” (Carlson 2021a). Carlson has also stated that CRT is about “abolishing whiteness” and “racism” (Carlson 2021b).

The notion of “white identity” is roughly as follows. In an increasingly diverse America, some whites feel the need to protect their status and group interests (Jardina 2019). Some white Americans may also adopt more politically conservative viewpoints (Craig and Richeson 2014). Reactions to CRT reflect this “fear mongering about the rise of otherness and the displacement of whiteness, the white patriarchy and a dominant white narrative” (Blow 2021). Such framing of CRT by political and media elites creates a context that might be particularly resonant with those who feel threatened by changes in American society and identify with the threatened white majority. Indeed, past work suggests that both Republicans and strong white identifiers are more concerned about teaching CRT in schools because it is perceived as “an attack on their party, racial group, and nation” (Kaufmann 2022, 787). We consequently suspect that language describing culturally relevant pedagogy that focuses on white privilege may be more likely to activate not only racial bias, but also white racial ingroup identity.

Treatments and Hypotheses

The teaching of race in American schools has become politicized, and the nature of this politicization is likely to engage different attitudes and social identities. To examine this possibility, we developed four different frames around how race might be taught in schools. Respondents were randomly assigned to receive one of the following four treatments:

• History: Some people think we should teach young people about the history of race in this country, while others disagree.

• CRT: Some people think we should teach young people about Critical Race Theory, while others disagree.

• Discrimination: Some people think we should teach young people about how American institutions perpetuate racial discrimination, while others disagree.

• Privilege: Some people think we should teach young people about how White people have been privileged and Black people discriminated against, while others disagree.

The History treatment is intended to be the most neutral phrasing and serves as our baseline category. The CRT treatment explicitly mentions critical race theory, which has become politicized in current debates, and so we expect it to activate people’s partisan identities. The Discrimination treatment is intended to describe a common understanding of CRT without using the word. And finally, the Privilege treatment is an alternative phrasing that explicitly labels an opposition between white and Black people likely to elicit a defence of white social identity.

Note that our expectations were not focused on the (likely) possibility that “CRT” will generate less support than some of the other language used in our treatments. Our primary interest, rather, was the potentially varying impact of partisanship, symbolic racism, and white identity across treatments. We hypothesized that different frames would engage these three attitudes to varying degrees. The hypotheses that follow were pre-registered. ² First, we had some straightforward expectations about the relationship between each of our three correlates and policy support:

H1. Support for all policies, regardless of framing, will be lower amongst those high in symbolic racism.

H2. Support for all policies, regardless of framing, will be lower amongst those who identify as Republican.

H3. Support for all policies, regardless of framing, will be lower amongst white respondents who are high in white identity compared to white respondents who are low on white identity.

We then had expectations about the impact of treatments on the relative impact of white identity and partisan identification, as follows:

H4. The estimated negative association between Republican partisan identification and policy support will be strongest for the CRT treatment (H4a) especially in comparison with the History treatment, where the negative association will be weakest (H4b).

H5. The estimated negative association between white identity and policy support will be strongest for Privilege treatment (H5a), especially in comparison with the History treatment, where the negative association will be the weakest (H5b).

Finally, we had expectations regarding the impact of CRT in particular:

H6. The estimated negative effect of the CRT treatment on policy support will be strongest for individuals who identify as Republican and who indicate comparatively low levels of symbolic racism (H6a), especially in comparison to the generic History treatment, for which the estimated effect will be weakest (H6b).

H7. The estimated negative effect of the CRT treatment on policy support will be strongest for individuals who identify as Republican and who indicate comparatively low levels of white identity (H7a), especially in comparison to the generic History treatment, for which the estimated effect will be weakest (H7b). ³

In sum, we expected that CRT would be particularly effective at engaging partisan identification, which will be especially effective for respondents who might otherwise be supportive of similar policies given their comparatively liberal positions on measures of symbolic racism and white identity.

Methods and Measures

To explore the impact of these frames on attitudes about schooling, we fielded a survey experiment amongst a representative sample of U.S. citizens. The survey was fielded from December 8–21, 2021, with 2,020 respondents recruited by Cint, an online panel provider. ⁴ Quotas were used for region, age, gender, and education. Sample descriptives are included in online appendix table 1. The raw data are also available in Soroka et al. (2024).

As noted, respondents were randomly assigned to see one of the four treatments described in the previous section. After each prime, respondents were asked, “Would you say that we should teach history this way in 1) Elementary schools, 2) High schools, and 3) Colleges and Universities.” All responses were captured using a five-point scale ranging from Strongly Agree to Strongly Disagree, and we rescaled these (dis)agreement measures from 0 (high disagreement) to 1 (high agreement) and take the average to form an index we refer to later as policy “support.” There was an explicit “don’t know” option, so the 126 respondents selecting “don’t know” are excluded from our analyses. We expected that there may be increased support for culturally relevant pedagogy (regardless of treatment) at more senior levels of schooling. Results confirm this suspicion: mean levels of support across all treatments are 0.55 for elementary school, 0.65 for high school, and 0.68 for university. Even so, there are relatively small differences across the three measures, which have an average inter-item correlation of 0.82. Results do not shift significantly when we focus on just one level of schooling; we accordingly rely on a simple scale averaging policy support across all three levels of education.

We are focused on three potential correlates of agreement with treatments: 1) partisanship (PID), 2) symbolic racism (SR), and 3) white identity (WID). Partisanship is measured using the standard 7-point party identification variable. We rely on questions from a well-established measure of symbolic racism as an indication of one form of racial bias. Symbolic racism, also sometimes called modern racism or racial resentment, is most associated with the work of Kinder and Sears (1981). We acknowledge that there is some debate about the degree to which the measure captures anti-Black racism per se—rather than some combination of racism, ideology, or policy positions, for instance (for a more complete discussion of potential weaknesses of the SR measure, see, e.g., Carmines, Sniderman, and Easter 2011; Feldman and Huddy 2005; Wilson and Davis 2011; Sniderman and Carmines 1997). Recent work suggests that it continues to measure important components of racial bias in U.S. politics, however (e.g., Simmons and Bobo 2018); and that it is correlated with other measures of anti-Black racism (e.g., Valentino, Neuner, and Vanderbroek 2018). This informs our interpretation in the results section.

White identity is captured using questions developed by Jardina (2019), for white respondents. (Sample sizes in the models that follow are reduced when WID is included, based on this limitation.) All three measures are rescaled from 0 to 1, as with our index of policy support, ⁵ although we rely below on a 3-category variant of partisan identity as well (Democrat, Independent, and Republican, in which “leaners” are categorized as either Democrats or Republicans). The wording of these questions can be found in the online appendix, and distributions of each variable are included in appendix figure 1.

Results

Before diving into models of policy support, we consider the relationships between our independent variables. We capture these relationships in figure 1, where each data point is a respondent plotted by their level of symbolic racism (y-axis), white identity (x-axis), and partisanship, and where different symbols indicate Republicans, Democrats, and Independents. Data points are jittered so that the figure more easily reflects the number of respondents within each category. Dashed lines show the midpoints of the x-and y-axes.

Figure 1 The relationship between symbolic racism, white identity, and partisanship

Data in the bottom left quadrant of figure 1 indicate that there are very few (if any) Republican identifiers who are low in both symbolic racism and WID. Data in the top right quadrant suggest that there are Democratic identifiers who are high in both, although Republicans do clearly report higher levels of WID and symbolic racism than Democrats. WID is correlated with racial resentment at r=0.48; WID and party identification are correlated at r=0.22; symbolic racism and party identification are correlated at r=.45 (N=2,020; all Pearson coefficients are significant at p < 0.01). Clearly, then, these variables are related, but each has a good amount of unique variation as well.

Our basic findings regarding policy support across treatments are captured in table 1, which shows an ordinary least squares (OLS) model of policy support as a function of experimental treatment. The reference category, here and throughout our analyses, is the History treatment (“Some people think we should teach young people about the history of race in this country while others disagree.”). Coefficients thus capture differences in support for policy for the Discrimination, Privilege, and CRT treatments versus this History treatment. Negative coefficients suggest that support is highest amongst those who received the residual History treatment. The largest negative coefficient is, as we expected, for the CRT treatment, but note that this coefficient is not statistically distinguishable from the Privilege treatment. ⁶ Point estimates based on the model in table 1 are shown (with 95% confidence intervals) in figure 2.

Table 1 Policy support across treatments

Notes: + p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001

Figure 2 Policy support, by treatment

Our pre-analysis plan indicates that we will test H1–H3 by extending the OLS model in table 1—by regressing policy support on treatments, adding SR, PID, and WID individually. We do this in table 2. In Models 1 through 3, each of the attitudinal variables has a significant negative association with policy support, though based on a comparison of R-squareds WID has far less explanatory power (R²=.039) than does SR (R²=.234) or PID (R²=.229). It appears to be the case that attitudes about culturally relevant pedagogy are linked to a somewhat different set of concerns than WID. The strictest test of H1–H3 is Model 4 in table 2, in which all three attitudinal variables are included simultaneously. The estimated effect of SR is unaffected by the inclusion of the other variables, and the effect of PID drops by just under 40%, but both continue to have significant negative effects on policy support. The effect of WID is now positive, a consequence of multicollinearity between the three attitudinal measures. Our results thus clearly support rejecting the null hypothesis for H1 and H2, but there is less support for rejecting the null hypothesis for H3. WID itself is negativity associated with policy support, but this may largely be a function of its positive relationship with SR and PID. On its own, WID explains much less variation in policy than do these other variables; and it has no independent negative influence once these other variables are included in the model.

Table 2 The effects of SR, PID, and WID on policy support

Notes: + p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001

H4 and H5 are tested using OLS models shown in table 3. In this instance, we allow PID (Model 1) or WID (Model 2) to interact with the categorical treatment variables. The estimated direct effect of PID or WID thus reflects the impact of each for the reference (History) treatment. Interaction coefficients then capture the extent to which the impact of PID or WID varies across treatments.

Table 3 The varying impact of PID and WID across treatments

Notes: + p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001

The direct effects of treatments are scarcely evident when we allow for an interaction with PID (Model 1). Only the coefficient for the CRT treatment is statistically different from zero, suggesting a reduction in support relative to the History treatment. That said, F-tests indicate that the estimated coefficient for the CRT treatment is not statistically distinguishable from the Discrimination or Privilege treatments. ⁷ Treatment coefficients for Model 1 capture the magnitude of the effect only amongst strong Democrats (where PID=0), of course, since interaction coefficients capture shifts in the impact of treatments amongst respondents with more moderate/Republican identifications. The significant negative coefficient for PID suggests that partisanship matters for policy support, independent of treatments; but all interactions are negative and significant, suggesting that the impact of partisanship is greater in each of the Discrimination, Privilege, and CRT treatments, relative to the History treatment.

The combined effects of treatments and partisanship are difficult to distinguish from coefficients alone, so figure 3 shows the estimated levels of support for Democrats and Republicans across all four treatments. There is clear evidence that the Discrimination, Privilege, and CRT treatments produce significant drops in policy support for Republicans. The estimated gap in partisan support is widest for the CRT treatment, but F-tests indicate that the difference between this and the Discrimination and (especially) the Privilege treatments is not statistically significant.

Figure 3 Policy support and PID across treatments

Model 2 in table 3 suggests no significant effect of WID on policy support in the History treatment, no significant interactions coefficients, and—most importantly given our hypothesis—no significant differences in the interaction coefficients. (The estimated interactive effects in table 3 are plotted in online appendix figure 2.) There thus appears to be only a very limited association between WID and policy support—evident only in table 2 when neither PID nor SR are included in the model. We thus fail to reject the null hypothesis for H4a, but find support for H4b: partisanship matters all the time, but least in the History framing. ⁸ We fail to reject the null hypotheses for H5a and H5b: in the interactive model in table 3, WID is not systematically associated with policy support, regardless of treatment.

H6 is tested using three OLS models, included in online appendix table 2, that interact treatments with SR—one model for Democrats (Model 1), Independents (Model 2) and Republicans (Model 3). ⁹ The direct effect of the CRT treatment, capturing the impact of that treatment amongst those lowest in SR, is not more negative amongst Republicans. (The coefficient itself is smaller in magnitude than for Democrats, although F-tests suggest no significant difference between coefficients across models.) We consequently fail to reject the null hypothesis for H6. Indeed, results suggest a rather different dynamic, namely, a stronger impact of SR in the CRT treatment amongst Republicans relative to Democrats. Figure 4 shows estimated policy support across levels of SR for Democrats and Republicans, within each of the four treatments. ¹⁰ At low levels of SR, policy support is relatively high and roughly equal amongst Democrats and Republicans. Increasing levels of SR are then associated with a downward shift in policy support, particularly for Republicans.

Figure 4 Policy support and SR amongst Democrats and Republicans in the CRT treatment

These results are not in line with our expectations, but they do suggest that the relative strength of the CRT treatment may be that it engages symbolic racism especially well for Republicans. Indeed, the negative impact of symbolic racism appears to be especially high for Republicans in the CRT treatment in figure 4. This is indeed the only treatment for which the interaction with SR is negative and statistically significant (in online appendix table 2); that said, even as the direct and interactive coefficients for the Privilege treatment are insignificant, F-tests suggest that they are statistically indistinguishable from the coefficients related to the CRT treatment. Evidence for the special influence of the CRT treatment amongst Republicans is therefore limited.

Our test of H7 relies on a similar set of models as H6, although this time we include WID rather than SR. Results are included in online appendix table 3. Republican party identification has a negative effect on policy support, evident in the constants in online appendix table 4 (which capture levels of policy support for the residual History treatment when WID is equal to 0). But pairwise comparisons of coefficients across models do not suggest any heterogeneity in the impact of treatments across levels of WID. We accordingly do not reject the null hypothesis for H7.

Discussion

Racial priming has a long history in American politics, and in the study of American public opinion as well. While norms around explicit racial rhetoric seem to have declined (Valentino, Neuner, and Vandenbroek 2018), the salience of race and racism in recent political debates raises important questions about its potency for shaping partisan opinion. Our results suggest that the debate about culturally relevant pedagogy (and CRT in particular) in the United States is one example of the ways in which explicit discussions of race and racism can be mobilized by political actors, with potentially significant impacts on public opinion.

This study was designed to explore how language surrounding culturally relevant pedagogy affects the expression of political attitudes toward teaching about race in schools. We suspected that given the current climate, support would be lowest when the term “critical race theory” is used, relative to more neutral language, or the use of less politicized terms. Our results partly support this hypothesis. On one hand, compared to a neutral frame about the teaching of race, people were significantly less supportive when a CRT frame was used. On the other hand, even as the impact of the CRT frame is more robust (i.e., significantly different from the History frame across all specifications) than other frames, the differences between CRT, Discrimination, and Privilege frames are in many cases negligible. Each of these frames appears to drive support downwards.

Our core interest, however, was in whether the use of different policy language could differentially prime other attitudes and social identities. While we suspected that any discussion of teaching about race in school would prime racial attitudes, we hypothesized that the term CRT would be especially politically charged to activate partisan identities, (almost) regardless of racial attitudes. Similarly, we suspected that language focused on white privilege would be more likely to engage white identity. Again, our findings only partially support these expectations. CRT language is associated with decreased support, particularly amongst Republican identifiers, compared to the neutral history condition. Yet this may be in part because CRT effectively engages symbolic racism—though only marginally more so than the Discrimination and Privilege treatments.

Contemporary American politics are both racially and politically polarized, and as Tesler (2016) notes, these two processes are in lock step. This study demonstrates one way in which polarization can play out in a policy domain where elite messaging is both clearly partisan and explicitly engages an (increasingly partisan) divide on racial issues. Our study is of course not without its limits. We have a single measure of racial prejudice—symbolic racism—that captures just one dimension of racial attitudes. We also have a single snapshot of the effects of framing on attitudes taken in 2021. We thus can say little about how this effect emerges over time, and if it eventually erodes when the issue falls off the public radar. We suspect, however, that issues like CRT that combine partisan and racial issues with a defense of a more traditional view of the American nation and its history are likely to be—when used by political and media elites in the current polarized U.S. context—particularly effective in dividing public opinion into identity-based camps. And note that this is readily evident in our results, even as we do not have explicit partisan cues in the experiment.

As these debates unfold in the public sphere, the partisan identity of actors mobilizing for and against more culturally relevant pedagogy in the classroom may mute the effects we see here. When a well-known Republican or Democrat speaks about the teaching of race in schools, for instance, the partisan affiliation of the speaker may make more neutral phrasing inconsequential. The case of the public debate around culturally relevant pedagogy is nevertheless both of contemporary relevance and of serious importance for the nature of education that Americans receive.