White Americans’ Reactions to Racial Disparities in COVID-19

I fielded a survey experiment on a nationally representative sample of 591 white Americans to test whether exposure to information about the disparate impact of COVID-19 on Black people influenced white Americans’ opinion about COVID-19 policies. I found that racially prejudiced white Americans who were exposed to the treatment diminished the importance of wearing a face mask. They also became more supportive of outdoor activities without social distancing guidelines, more likely to perceive shelter-in-place orders as a threat to their individual rights and freedoms, and less likely to perceive African Americans as following social distancing guidelines. Conversely, white Americans who did not endorse an anti-Black stereotype were less likely to perceive shelter-in-place orders as a threat to their individual rights and more likely to perceive African Americans as following social distancing guidelines. These findings highlight that well-intentioned public health campaigns may inadvertently exacerbate existing race-based health disparities.


INTRODUCTION
O n March 16, 2020, President Trump declared a national emergency and asked Americans to avoid gatherings because of COVID-19. Governors issued shelter-in-place orders requiring Americans to stay home. These restrictions were intended to "flatten the curve" of infections. However, the number of COVID-19 cases and deaths grew exponentially, with reports from the Centers for Disease Control indicating almost 400,000 cases and over 10,000 deaths by April 7, 2020. 1 According to preliminary data released by several states, Black Americans were accounting for the majority or plurality of coronavirus deaths in several states. 2 By mid-April, protestors who were overwhelmingly white were descending on state capitols, pressuring their governors to remove COVID-19 restrictions. These protests were a part of a growing trend of reactionary movements on the Right (Parker 2021). It was not uncommon to see protestors holding Confederate flags, a symbol whose support is associated with racial prejudice (Hutchings, Walton, and Benjamin 2010;Strother, Piston, and Ogorzalek 2017). 3 The protestors routinely held placards focusing on individual rights, with phrases like "Give Me Liberty or Give Me COVID-19." However, the relaxation of COVID-19 restrictions the protestors were demanding, would disproportionately affect African Americans. African Americans are more likely than whites to be essential workers, who would be most exposed to the potentially infected public if restrictions were loosened. 4 Throughout the pandemic, African Americans have been more likely to contract, be hospitalized for, and die from the coronavirus than white Americans-facts that have been publicized by the media. However, little attention has been devoted to what influence publicizing these disparities might have on white Americans' opinions about COVID-19. Because the race-based disparities associated with the coronavirus align with one of the most salient social cleavages in American politics-race-I ask whether drawing attention to COVID-19 racial disparities attenuates support for COVID-19 safety precautions among racially prejudiced whites.
Using a preregistered survey experiment fielded on a nationally representative sample of 591 white Americans, I provide causal evidence that exposure to COVID-19 racial disparities information is associated with a negative response to COVID-19 safety precautions among racially prejudiced whites. 5 1 "Geographic Differences in COVID-19 Cases, Deaths, and Incidence-United States, February 12-April 7, 2020," MMWR Morb Mortal Weekly Report 69 (2020)

THE RACIALIZATION OF PUBLIC OPINION
A large body of research demonstrates that racial attitudes influence the political preferences of white Americans. Issues that are explicitly about race, such as affirmative action, are thought to activate racial predispositions because of a natural associative link between the substance of the policy and feelings toward the groups who benefit from them (Nelson and Kinder 1996;Winter 2008). Research also indicates that ostensibly nonracial policies such as welfare, social security, and crime are also racialized through mass communications that either consciously or unconsciously increase the association between African Americans and these issues (Gilens 1999;Mendelberg 2001;Winter 2008). Yet, in the American context, we know relatively little about the relationship between anti-Black racial attitudes and another ostensibly nonracial policy area-health. Most of the research that explores this relationship has focused on the relationship between anti-Black attitudes and opposition to Obamacare (Banks 2013;Tesler 2012), which means that we know little about the racialization of health policies that is not linked to the nation's first Black president.
However, recent work by Harell and Lieberman (2021) examines the relationship between racial attitudes and COVID-19 opinion. They find evidence that whites with a negative affect toward African Americans are less supportive of an aggressive response to COVID-19. However, their study may not have fully captured the extent to which racial attitudes influence white opinion on COVID-19 because their feeling thermometer rating did not account for how white respondents rated Black people relative to whites. Furthermore, feeling thermometer ratings do not always take into account individual variance in response patterns (Wilcox, Sigelman, and Cook 1989). Similarly, Skinner-Dorkenoo et al. (2022) found that white people who read about racial disparities in COVID-19 showed reduced fear of COVID-19 and less empathy for those most vulnerable to the virus. However, they did not measure anti-Black prejudice.
In my study, I captured prejudicial attitudes toward African Americans, using a widely adopted metric that asks respondents to rate how hardworking or lazy and intelligent or unintelligent they think Black people are relative to white people (Piston 2010). This robust measure of racial animus provides causal evidence that anti-Black attitudes can affect support for public health measures. Anti-Black attitudes also influence whether white Americans think Black people follow social distancing guidelines, which is consistent with a long line of research suggesting that many white Americans do not think that Black people "play by the rules" (Kinder and Sanders 1996).

HYPOTHESES
My hypotheses formally stated are as follows: H1: Exposure to COVID-19 racial disparities information will be associated with racially prejudiced whites becoming a. less supportive of COVID-19 guidelines or b. less likely to believe Black people follow guidelines.
H2: Exposure to COVID-19 racial disparities information will be associated with nonracially prejudiced whites becoming a. more supportive of COVID-19 guidelines or b. more likely to believe Black people follow guidelines.

DATA AND METHODS
I fielded a survey experiment on an online nationally representative sample of 591 whites between May 21 and May 26, 2020. The online sample was collected by the National Opinion Research Center (NORC) at the University of Chicago, using their AmeriSpeak Panel. 6 The sample was 39% self-identified Democrats (including leaners) and 46% self-identified Republicans (including leaners). Fifty-one percent of the sample identified as women, 32% resided in the South, and 38% reported having a bachelor's degree or higher. Prior to fielding this study, I fielded a pilot study on an online convenience sample of whites one month earlier, which helped to inform the hypotheses for the present study. 7 Respondents answered questions about their racial attitudes, followed by distractor questions regarding their sleep and spending habits. The main racial attitude is the extent to which respondents endorse stereotypes about Black people relative to whites.
Respondents were asked to rate the extent to which Black people are lazy rather than hardworking on a seven-point scale. They were also asked to rate the extent to which Black people are unintelligent rather than intelligent on a seven-point scale. Finally, respondents were asked to evaluate whites along those same dimensions, and the order in which the racial groups were presented to the respondents was randomized. Subsequently, I subtracted the scores respondents gave Black people from the scores respondents gave white people to allow me to examine how a given respondent rates Black people relative to white people. Unlike the racial resentment measure, which critics argue conflates racial attitudes with ideology (Feldman and Huddy 2005;Sniderman and Tetlock 1986), endorsing the stereotype that Black people are lazier and less intelligent than white people is an unambiguous measure of racial prejudice that cannot be interpreted as a commitment to an ideological principle. A nontrivial fraction of the nationally representative sample-26%-endorsed either the stereotype that African Americans are less hardworking than whites or that African Americans are less intelligent than whites.
Subsequently, I randomly assigned subjects to one of two conditions. In the control condition, respondents were given information about the number of COVID-19 cases and deaths in several states without any mention of race. In the treatment condition, respondents were given the same information as respondents in the control group, with additional information about racial disparities in COVID-19 deaths, including the most recent racial death estimates (see Appendix).
After exposure to either the control condition or treatment condition, respondents answered questions about their coronavirus opinions, including their attitudes about the importance of wearing facemasks, their perception of whether shelter-inplace orders were a threat to their individual rights, whether parks should be reopened without any restrictions, and whether they thought African Americans followed social distancing guidelines. The question about facemasks was intended to test whether adherence to prosocial behavior was influenced by racial disparities information. Similarly, since much of the narrative regarding COVID-19 guidelines has been framed in some conservative circles as a threat to individual rights, the question about whether shelter-in-place orders threatened individual rights, was asked to see whether exposure to racial disparities information heightened that perception. The question about park restrictions was another metric of whether exposure to racial disparities information was associated with a decline in support for restrictions. Finally, the question regarding whether Black people follow social distancing guidelines was motivated by discourse, in which there is a common sentiment that Black people are most lacking in personal responsibility (Kinder and Sanders 1996; Stephens-Dougan 2020).

RESULTS
I estimated several logistic regression models where the dependent variables were the importance of wearing a face mask, the perception that shelter-in-place orders are a threat to individual rights, support for visiting parks without any restrictions, and the perception that African Americans follow social distancing guidelines. 8 I tested for interaction effects between exposure to the treatment and a variable indicating whether the respondent either endorsed the stereotype that African Americans are less hardworking than whites or the stereotype that African Americans are less intelligent than whites (i.e., negative stereotype endorsement = 1). 9 Because the treatment and control conditions were balanced on demographic variables, controlling for demographic characteristics in the regression models was unnecessary. 10 The estimates from the regression models are displayed in Table 1.
As Table 1 indicates, the coefficient representing the interaction between exposure to the treatment and rating Black people as less hardworking or less intelligent than whites is statistically significant for all of the dependent variables. For ease of interpretation, I plotted the predicted probability of agreeing with each pandemic measure in Figure 1, with 84% confidence intervals, the graphical equivalent to p < 0.05. 11 Given the unidirectional hypotheses, I used one-tailed significance tests. As Figure 1 indicates, among racially Note: Entries are logit coefficients, with standard errors in parentheses. *p < 0.10, **p < 0.05, ***p < 0.01. 8 I treated these items separately rather than as a scale because the Cronbach's alpha for these four items was below the standard of .70. 9 There was no main effect of the treatment. 10 The results of the balance tests are in the Appendix. 11 I combined the "Not At All Important" and "Not Very Important" (facemasks), "Strongly Agree" and "Agree" responses (individual rights threatened) and the "Never, Some, and Half" responses (Black people follow social distancing guidelines). The distributions of the dependent variables are plotted in the Appendix.
prejudiced white Americans, exposure to the racial disparities information attenuated support for efforts to slow the spread of the pandemic.
In the control condition, among racially prejudiced respondents the probability of indicating that it was "not at all important" or "not very important" to wear a face mask was 46%. However, among racially prejudiced respondents in the treatment condition, the likelihood of indicating that mask-wearing was "not at all important" or "not very important," was 75% (p < 0.05 for a one-tailed test). These results indicate that a critical preventative measure in fighting the coronavirus was perceived as less important when racially prejudiced whites were exposed to COVID-19 racial disparities information.
Among racially prejudiced whites, exposure to the racial disparities information also heightened the perception that safety precautions were an infringement on their individual rights. The likelihood of racially prejudiced respondents in the control condition agreeing that shelter-in-place orders threatened their individual rights and freedom was 27%, compared with a likelihood of 55% in the treatment condition (p < 0.05 for a one-tailed test). With respect to visiting parks without restrictions, the probability of racially prejudiced white Americans endorsing this measure was 11% in the control condition. However, the predicted probability of racially prejudiced white Americans supporting visiting parks without restrictions was much higher, at 29% in the treatment condition (p < 0.10 for a one-tailed test). Finally, the likelihood of racially prejudiced white Americans perceiving that African Americans follow social distancing guidelines no more than half of the time ("half," "sometimes," "never") was 77% in the control condition but that likelihood increased to 90% in the treatment condition (p < 0.10 for a one-tailed test). When racially prejudiced white Americans were exposed to the racial disparities information, there was an increase in the predicted probability of indicating that they were less supportive of wearing face masks, more likely to feel their individual rights were being threatened, more likely to support visiting parks without any restrictions, and less likely to think African Americans adhere to social distancing guidelines. Drawing attention to the disparate effects of COVID-19 on African Americans heightened the association between race and opinion on COVID-19 such that racially prejudiced whites brought their negative attitudes about African Americans to bear on their COVID-19 opinions.
Akin to how racially prejudiced whites brought their attitudes about Black people to bear on their COVID-19 opinions, I expect that white Americans who did not endorse the stereotype of African Americans being less hardworking or less intelligent than whites might also bring their attitudes about African Americans to bear on their COVID-19 opinions. I hypothesize that there will be an increased probability that they will become more supportive of prosocial COVID-19 behavior when treated with the racial disparities information and less likely to think that Black people are not following social distancing guidelines. In Figure 2, I have plotted the predicted probability of agreeing with the pandemic measures of interest for white Americans who did not endorse the stereotype that African Note: These estimates, with 84% confidence intervals, were generated from the models in Table 1.
Americans are less hardworking or less intelligent than whites (i.e., negative stereotype endorsement = 0).
As Figure 2 indicates, exposure to the racial disparities information was not associated with a significant increase in the predicted probability of the importance of wearing masks. The likelihood of respondents in the control condition perceiving mask wearing as "not at all important" or "not very important is 41%, as opposed to 37% in the treatment condition. On the other hand, among respondents who did not endorse the negative stereotype of African Americans as less hardworking or less intelligent than whites, exposure to the racial disparities information was associated with a decline in the likelihood of them perceiving shelter-in-place as a threat to their individual rights, which might be indicative of racial sympathy on the part of whites who did not endorse the negative stereotype of Black people. 12 In the control condition, among whites who did not endorse the negative stereotype of Black people, the likelihood of perceiving shelter-in-place orders as a threat to their individual rights was 0.31 but the likelihood decreased to 0.21 in the treatment condition (p < 0.10 for a one-tailed test). Finally, among respondents who did not endorse the negative racial stereotype, the likelihood of agreeing that African Americans follow social distancing guidelines no more than half of the time ("half," "sometimes," "never") was 0.67 but decreased to 0.57 in the treatment condition (p < 0.10 for a one-tailed test). In other words, among respondents who did not endorse the negative racial stereotype of African Americans, exposure to the racial disparities information was associated with a decrease in the likelihood of respondents perceiving African Americans as not "playing by the rules." In short, among white Americans who did not endorse the anti-Black stereotype, exposure to the racial disparities information, on average, was associated with a decreased likelihood of perceiving Black people as not following social distancing guidelines and a decreased likelihood of perceiving their individual rights as threatened by shelter-in-place orders.
One question that these results raise is whether the effect of the treatment among racially prejudiced whites was statistically distinguishable from the effect of the treatment among whites who did not endorse the negative stereotype of Black people. Therefore, I subtracted the average treatment effect for whites who did not endorse the anti-Black stereotype from the average treatment effect for whites who did endorse the anti-Black stereotype. These differences in the average treatment effects are displayed in Figure 3.
As Figure 3 indicates, the differences are substantively significant and statistically distinguishable from zero for some of the variables-namely, regarding the importance of facemasks and the perception of whether African Americans follow social distancing guidelines. This confirms that for these two variables, the racial disparities treatment had a different effect among white Americans who endorsed either of the stereotypes and white Americans who did not endorse the stereotypes. However, regarding the perception of whether shelter FIGURE 2. The Effects of Racial Disparities Information (Non-Racially Prejudiced whites) Note: These estimates, with 84% confidence intervals, were generated from the models in Table 1. 12 The survey did not include a measure of racial sympathy.
-in-place orders were a threat to individual rights and freedom, as well as support for visiting parks without any restrictions, the difference in average treatment effects was not statistically distinguishable from zero for white respondents who endorsed the negative stereotypes relative to whites who did not endorse the negative stereotypes.

DISCUSSION
The results presented in this paper are robust to various model specifications. For example, one alternative explanation is that racially prejudiced white Americans might be more resistant to COVID-19 restrictions because they are less likely to live in areas with African Americans (who are more likely to contract and die from COVID-19). Therefore, I estimated models in which I controlled for the percentage of African Americans living in a respondent's zip code, as a metric of perceived risk. The results were robust to the inclusion of this variable (see Appendix), which suggests that racially prejudiced whites' response to the treatment mattered beyond whether they lived around African Americans. Another potential explanation is that racially prejudiced whites might have been less likely to live in counties where the number of COVID-19 cases was high. However, even after controlling for the number of COVID-19 cases in a respondent's county, exposure to the treatment was still associated with racially prejudiced whites being less willing to support behaviors that would slow the spread of the pandemic (see Appendix). Yet, another potential threat to the results is that the treatment condition mentioned specific locales, which might have made people who lived in those locales more responsive to the treatment relative to people who do not reside in those locations. However, even after controlling for whether respondents lived in one of the states that was mentioned in the treatment, the finding that exposure to racial disparities information was associated with resistance to COVID-19 restrictions among racially prejudiced white Americans, still held. The results were also robust to ordinary least squares model specifications where the dependent variables were not dichotomized. 13

CONCLUSION
The results of an experiment fielded on a nationally representative sample of 591 whites indicate that informing some white Americans about racial disparities in COVID-19 may backfire. When white Americans who endorsed anti-Black stereotypes were exposed to information about COVID-19 racial disparities, they became more resistant to COVID-19 restrictions. In short, the pandemic is racialized. Note: These estimates, with 84% confidence intervals, were generated from the models in Table 1.
Among whites who endorsed the negative stereotype of Black people as less hardworking or intelligent than white people, the likelihood of them placing importance on wearing a face mask, one of the main tools for preventing the virus's spread, diminished when they were exposed to the racial disparities information, relative to their racially prejudiced counterparts in the control condition. The results also indicate that when exposed to COVID-19 racial disparities information, whites who endorsed a negative stereotype of African Americans were more likely than were racially prejudiced whites in the control condition, to state that African Americans did not follow social distancing guidelines. Information about the disparate effects of the coronavirus on African Americans seemed to reinforce the stereotype of African Americans as "not playing by the rules" (Kinder and Sanders 1996). Finally, racially prejudiced white Americans in the treatment condition were more likely to perceive shelter-in-place orders as a threat to their individual rights relative to racially prejudiced whites in the control condition and more likely to support the reopening of parks without any restrictions.
Conversely, exposure to the racial disparities information was associated with more prosocial behavior among white Americans who did not endorse a negative stereotype of African Americans, perhaps because these respondents were more likely to be racially sympathetic. Previous research indicates that racial sympathy motivates support for policies perceived to benefit African Americans (Chudy 2019). Exposure to the treatment was associated with the respondents who did not endorse the anti-Black stereotypes being less likely to perceive shelter-in-place restrictions as a threat to their individual rights and freedom relative to whites who did not endorse the anti-Black stereotypes and were in the control condition. They were also more likely to state that African Americans followed social distancing guidelines. However, among whites who did not endorse the anti-Black stereotypes, exposure to racial disparities information was not associated with a change in the importance of mask wearing or decreased support for reopening parks without any restrictions.
Of course, this study is not without its limitations. One limitation is that it captures results at one snapshot in time. Whether results would be of the same magnitude and significance if the study had been conducted at a different point in the pandemic is unclear. Also, the racial disparities information was delivered in a controlled environment, which may raise concerns about external validity. However, the disparate effects of the virus on Black people was a part of the discourse regarding COVID-19 at the time of the study (May 2020), which makes it plausible that the average American may have been "treated" with more information than what was provided in the treatment in this study, which if anything, would have made it more difficult to detect a treatment effect.
Future research might consider examining the effect of reporting disparities that are not race based, such as age. In addition, it would be valuable to learn whether similar results hold for other racial groups who have also been disproportionately affected, such as Latinos.
Finally, future research should examine the extent to which the negative effect of reporting public health disparities might be mitigated. For example, might providing respondents with the information that African Americans are more likely to be essential workers mitigate the negative effect associated with sharing COVID-19 racial disparities information?
Because racial disparities in COVID-19 outcomes are unlikely to dissipate anytime soon, it is imperative that we learn more about how to communicate to the public about the disparate effects of COVID-19. Unfortunately, the results of the present study suggest that efforts to slow the spread of the pandemic may be hampered by the knowledge that African Americans bear a disproportionate burden of the pandemic. The societal implications are staggering.

SUPPLEMENTARY MATERIALS
To view supplementary material for this article, please visit http://doi.org/10.1017/S000305542200051X.

DATA AVAILABILITY STATEMENT
Research documentation and data that support the findings of this study are openly available at the American Political Science Review Dataverse: https://doi. org/10.7910/DVN/HRUQZJ. Limitations on data availability are discussed in the appendix.