Public Opinion and CRT

Two recent studies have produced empirical evidence of how uncertainty of CRT’s meaning has contributed to public opinion regarding the phrase. The first is a 2023 study that finds that clear messaging about CRT from Republican elites has resulted in strong opposition among rank-and-file party members. In contrast, Democrats tend to be uncertain about their attitudes toward CRT, likely because their party elites have not delivered a coherent message on the issue. In short, differences in elite messaging resulted in differences in intensity of belief and, by extension, differences in actions (Deshpande et al. Reference Deshpande2023).

The conclusion that Republicans are more confident in their understanding of CRT is echoed by a study by Safarpour et al. (Reference Safarpour2024), which describes the phrase “Critical Race Theory” as being an empty signifier that came to represent a broad range of “racial and cultural grievances.” While Deshpande et al. showed the effect of intensity of belief, the second study showed that elites’ broad application of the phrase “Critical Race Theory” has shaped the public’s attitudes toward it.

Both the study by Deshpande et al. and that by Safarpour et al. utilized survey experiments that randomly varied the way in which CRT was presented. In the former study, participants read statements about the government either mandating, prohibiting, or taking no position on the teaching of CRT in schools after being shown a brief definition that provides what could be described as an academic summary of the theory, including race described as a social construct and enduring racial inequality resulting in part from U.S. laws and institutions. This design is especially effective for showing the uncertainty among Democratic respondents; even when provided with a definition focused on two of the most publicized aspects of CRT, Democrats were still uncertain about whether they would oppose a ban on teaching it in schools (Deshpande et al. Reference Deshpande2023). In the latter study, some participants were asked whether “Critical Race Theory” should be taught in schools (their “CRT condition”) while others were asked whether schools should teach “the legacy of racism” without any reference to CRT by name (their “definition condition”).

Our study takes this line of inquiry a step further by considering a third framing, one reflecting the legislation addressing the teaching of CRT in schools as proposed in many states, in addition to an academic meaning and the phrase “Critical Race Theory.” We expect to find that individuals will be more likely to support a ban when shown the phrase “Critical Race Theory” than when they are shown a description of the concept as used historically by academic scholars. Like the previous studies, we expect that Republicans will exhibit a stronger effect from the treatment than Democrats due to the uncertainty and comparatively weak messaging on the part of Democratic elites. Finally, we expect that because anti-CRT legislation tends to ignore tenets of CRT such as race being a social construct and enduring effects of race-based slavery by instead focusing on defending the U.S. as a country that ensures equality between groups, individuals will express more support for a ban on CRT as it is articulated in state laws than it is in academic usage. We expect this study to contribute to the literature by testing whether the intensity of support for CRT bans will vary not only by party but also by whether it is framed in the manner of academia, legislation, or name.

Theoretical Grounding

Research shows that politicians and pundits tend to exploit limitations to the public’s understanding of issues. For example, while debating estate taxes, opponents framed them as “death taxes” so they would sound both absurd and universal (Schaffner, and Atkinson Reference Schaffner, Atkinson, Schaffner and Sellers2009). An analysis of political parties’ messages on social media found that their tone shifted from hopeful to fearful at the start of the COVID-19 pandemic (Widmann Reference Widmann2022). Similar analysis shows how tweets guided the debate over the adoption of Common Core standards through implementation of five different frames (government, propaganda, war, business, and experiment frames) in order to touch on values of both liberals and conservatives (Supovitz, and Reinkordt Reference Supovitz and Reinkordt2017). In each of these cases, uncertainty regarding the meanings of words and issues made them flexible enough to serve as tools for opponents and/or advocates to frame the issues to suit their message. Even a term as familiar as socialism can prompt reactions different from those prompted by its technical definition (Liebertz, and Giersch Reference Liebertz and Giersch2022). Political science research demonstrates that politicians can easily lead the public to adopt a particular attitude and it is difficult to correct resulting misperceptions (Nyhan, and Reifler Reference Nyhan and Reifler2010).

The explosion of attention on a previously little-known topic like CRT created room for elites to shape the public’s understanding. When issues are complex and unfamiliar, individuals tend to adopt understandings of those issues provided to them by trusted experts (Druckman Reference Druckman2001; Zaller Reference Zaller1992). And when those issues receive increased attention, as did CRT in 2021, individuals give them more importance (Iyengar, and Kinder Reference Iyengar and Kinder2010). As explained by Safarpour et al. (Safarpour et al. Reference Safarpour2024), CRT served as an empty signifier that could be framed to serve politician’s and activists’ purposes (Laclau Reference Laclau2005). Thus, we designed the following survey experiment to show empirically that a description of CRT in terms that reflect its academic origins will elicit a different response to a political question than both the phrase “Critical Race Theory” as well as a description of CRT framed in terms of recent state legislation aimed at banning its teaching. We expand on the existing literature by using a more complete academic definition, one that specifies policy, rather than the vague term “racism” as a source of inequality. Furthermore, by employing a legislation frame, we bring this strand of research closer to actual policy proposals.

Experiment Methods

To test whether a broad academic meaning of CRT elicits reactions that are similar to those drawn by the phrase “Critical Race Theory” itself or by key elements of typical anti-CRT laws, we conducted a survey experiment in the spring of 2022 with 562 participants from a national self-selected panel with quotas for race, gender, and party affiliation (see Yeager et al. Reference Yeager2011) provided by Qualtrics. Table 1 describes the sample.

Table 1.

Descriptive characteristics of the sample

Each participant was told that state governments play a role in setting school curricula and then were asked whether they would support a state ban on teaching certain concepts related to race in the United States. The proposed bans were framed in three different ways; each participant was shown one of the three at random. The appendix provides a table comparing the groups on several demographic measures, a Consolidated Standards of Reporting Trials (CONSORT) diagram showing participant attrition, and the text of the items appearing in the survey. We wrote the first frame to summarize the original, academic understanding of CRT:

Note that the frame does not include the phrase “Critical Race Theory” but rather a one-sentence explanation of how academics (for example, Ladson-Billings Reference Ladson-Billings, Lynn and Dixson2013; Taylor Reference Taylor1998), broadly speaking, apply CRT in their work. The frame features the idea that racial inequality has been enshrined at all levels of law and over a long period of time. Because this was the original meaning of CRT, we treat it as the control frame for the purposes of analysis.

A second frame does use the term “Critical Race Theory,” but provides no guidance as to its meaning, leaving interpretation up to the survey participant. It reads:

A third frame makes use of language of the various “anti-CRT” legislation being passed in multiple states. Rather than addressing the tenets of CRT as developed in the scholarship of law and social sciences, these laws reflect the argument made by many conservatives that CRT, its application, and its supporters are likely to upend the social hierarchy and portray the United States as a racist country (Ali Reference Ali2021; Frey Reference Frey2022). The frame reads:

In this frame, we took language directly from legislation passed in states like Florida and Idaho. Again, the phrase “Critical Race Theory” does not appear in the frame. Because so many of the laws being passed or proposed include language about both racism and sexism, we felt it important to include both in this frame. The tradeoff, of course, is that it introduces another element to the participants’ decision-making. Results could be biased if on average individual preferences on teaching bans shift based on whether one or both elements are present. We suspect, however, that this is unlikely to be the case (Sanchez et al. Reference Sanchez2017). The three frames are intended to allow us to compare participants’ reactions to CRT as an academic concept to the meaning as expressed in recent law and to the phrase itself. Similar to the work of Jardina and Traugott (Reference Jardina and Traugott2019), we will leverage partisanship in our analysis.

Experiment Results

We begin with descriptive results. Responses varied widely in response to each of the three frames. Figure 1 shows the distribution of responses for each frame. When shown the academic language frame, few participants (about a third) supported a ban. Note that the frame using academic language received the least support for a ban as well as the fewest refusals to answer. We interpret this pattern to mean that our description of how academic research uses CRT was generally acceptable to our subjects, or at least innocuous enough that very few participants saw a need to ban its teaching. The two other frames were polarizing by comparison, with participants more likely to choose the more extreme options, both strongly in support of and strongly opposed to the proposed ban.

Figure 1.

Support and opposition to each frame in sample.

Note: Error bars represent standard errors.

As the polarization observed in the above table is likely happening along partisan lines, we perform a similar comparison with participants who identify as at least leaning Democrat and those at least leaning Republican. When confining the sample to only those participants, the distribution becomes more lopsided, but in predictably opposite directions. Democrats were more likely to oppose bans in all three frames while Republicans were more likely to support bans in the second and third frames, as seen in Figures 2 and 3. As efforts at anti-CRT legislation have been most successful in Republican-majority state legislatures, it is not surprising that both the “Critical Race Theory” and legislation frames produced significantly more support among Republicans, but it is noteworthy that support for a ban is so low among Republicans when they are shown our description of CRT’s use in academic research. Also noteworthy is the fact that Democrats are more willing to show support when shown text from anti-CRT legislation than the other two frames, suggesting that the laws have been written in a manner that not only generates support among Republicans but draws less skepticism from Democrats.

Figure 2.

Support and opposition to each frame among Democrats in sample.

Note: Error bars represent standard errors.

Figure 3.

Support and opposition to each frame among Republicans in sample.

Note: Error bars represent standard errors.

Our next analysis collapses the dependent variable into a binary form, with support for a ban (either somewhat or strong) represented by the value 1 and opposition to a ban (either somewhat or strong) represented by the value 0. Using a logit regression model, we test whether the treatment frame affected the likelihood that a participant would agree with, rather than oppose, a proposed ban. Figure 4 shows the predicted probabilities of a participant expressing any degree of support for a ban rather than opposing a ban (participants who chose not to answer were excluded from this analysis). Note that participants are likely to oppose a ban when given the academic language frame or the “Critical Race Theory” frame, but when shown the legislation frame, the positions reverse. We interpret this finding to mean that the academic meaning of CRT, at least the way we chose to phrase it, is largely acceptable to people. The gap narrows when the frame features the phrase “Critical Race Theory,” which is not surprising given how politically charged the terms have become. Support for a ban when shown the legal language frame is highest, likely because the legislation in many states has been phrased to claim a position of equality in an effort to portray CRT and its advocates as promoting discrimination (Gonzalez Reference Gonzalez2021).

Figure 4.

Predicted probabilities of supporting a ban in each frame.

Notes: Predicted probabilities result from logistic regression models in which the dependent variable was coded 1 = strongly/support and 0 = strongly/somewhat oppose or prefer not to say. Error bars indicate 95% confidence intervals.

Returning to a more fine-grained version of the dependent variable with values for strongly agree (coded 4), somewhat agree (3), somewhat disagree (2), and strongly disagree (1), we examine the association between treatment frame and support for a ban using ordered logit regression analyses in Table 2. The table shows the odds ratios associated with greater levels of support for a ban when shown the “Critical Race Theory” or the legislative language frames instead of the academic explanation, which serves as the reference group. The first model includes all participants; while the “CRT” treatment did not generate responses that were significantly different from the control, the legislation language did, with the likelihood of showing more agreement increasing by 72 percentage points compared to the frame featuring academic language. The second model restricts the sample to only those participants who identified as White and produced substantively similar results.

Table 2.

Estimates from ordered logistic regression models reported as odds ratios with agreement with a ban as the dependent variable. The academic frame serves as the reference category in all four models

Notes: Standard errors in parentheses. The dependent variable was coded as follows: 4 = strongly support, 3 = somewhat support, 2 = somewhat oppose, and 1 = strongly oppose.

*p < .05, **p < .01, ***p < .01.

The third and fourth models further restrict the model to White Democrats (Model 3) or White Republicans (Model 4). Among White Democrats, the frame yields no measurable difference among participants’ responses; this group tends to oppose any such ban. While the legislative language frame appears most popular among White Democrats, the difference is not statistically significant. Among White Republicans, both the phrase “Critical Race Theory” and the language from legislation produce a statistically significant increase in support for a ban, nearly three-fold and four-fold, respectively. We restrict the results to White participants because the issue of CRT is so tightly intertwined with race; a table showing results for Democrats and Republicans of all races in the sample appears in the appendix. The odds ratios appearing in that set of results are similar to those appearing above.