Fear, Political Legitimization, and Racism: Examining Anti-Asian Xenophobia During the COVID-19 Pandemic

The early impacts of the COVID-19 pandemic on anti-Asian hate and bias are well-documented (Gover et al., 2020; Lantz & Wenger, 2022; Tessler et al., 2020; Wong, 2020).¹ According to the Stop AAPI Hate organization, which tracks such incidents, more than 10,000 hate incidents against Asian American and Pacific Islander people were reported between March 2020 and September 2021 (Yellow Horse et al., 2021). Other data sources, including law enforcement data have documented similar increases. One study, for example, found that officially recorded anti-Asian hate crimes in 16 major U.S. cities had increased by 145% in 2020 (CSU Center for the Study of Hate & Extremism, 2021). There are also several anecdotal examples of such incidents in the national media. In New York, for example, which established a task force specifically to deal with COVID-19-related increases, there were several anti-Asian incidents, including one in which a man poured acid on an Asian woman (Lee & Yadav, 2020). Some media reports have indicated that nearly one-third of Americans—and as much as 60% of Asian Americans—witnessed someone blaming Asian people in the early months of the COVID-19 pandemic (Ellerbeck, 2020).

Early empirical research has also found significant evidence of anti-Asian xenophobia and bias during the pandemic. Nguyen et al. (2020), for example, found that the proportion of tweets expressing negative attitudes toward Asian people increased significantly and concurrently with the spread of the disease across the U.S. Another study, conducted by Gray and Hansen (2021) used a difference-in-difference approach to assess the impact of the pandemic on anti-Asian hate crime in London and found that, in the first months following the identification of COVID-19, there was a significant increase in hate crimes against Chinese people, yet no such increases in other hate crimes or crime more generally. They argued, like many others have, that these patterns highlight an “intrinsic link” between COVID-19 and anti-Asian racism. Lu and Sheng (2020) similarly found that tweets containing anti-Asian slurs significantly increased in geographic areas immediately following local COVID-19 diagnoses.

Prior research has suggested at least two primary explanations for this demonstrable anti-Asian prejudice and xenophobia in the context of the COVID-19 pandemic. On the one hand, some research has suggested that increases in xenophobic attitudes may be driven by individual-level increases in fear, wherein individuals respond to the threat of infectious disease by “othering” that threat (Reny & Barreto, 2020; Silva et al., 2022). Indeed, there is an extensive body of research linking disease outbreak to prejudice and xenophobia (Clissold et al., 2020; Lee, 2019; Taylor, 2019). The disease-related stigmatization of Asian people, in particular, is situated within a longstanding history of racist tropes linking Asian people to stereotypes of disease, like the “Yellow Peril” myth propagated during the nineteenth century (see Chun, 2020; Molina, 2006; Mudambi, 2019; Power, 1995).

Within this context, Faulkner and colleagues (2004) drew on research in evolutionary psychology to present a theoretical model in which they posited that individual concerns regarding feelings of vulnerability to disease motivate negative reactions toward foreign people (see also, Christensen et al., 2020; Kurzban & Leary, 2001; Park et al., 2003). Others have also emphasized that individuals are more likely to activate stereotypical beliefs and attitudes when they feel threatened (e.g., Bodenhausen, 1993; Smith, 1993). According to this model, people have a basic tendency to avoid the risk of disease infection, and those who may be “likely” carriers of disease. In assigning “likelihood” in this scenario, individuals draw upon visibly identifiable features. In some cases, these features may be actual—albeit largely imperfect—indicators of individual ailments (e.g., lesions). In other cases, however, these features are grounded in prejudicial stereotypes and racist tropes, and include basic features like skin color. Racial outgroup members are conceptualized as importing or directly carrying disease pathogens (Faulkner et al., 2004). Critical to this model, however, is an individual sense of personal vulnerability; if one does not feel personally vulnerable, they are less likely to draw upon these xenophobic stereotypes, while those who feel more personally vulnerable, are more likely to draw upon these stereotypes (other factors aside, which we discuss below). In testing these propositions, Faulkner and colleagues found that those individuals with more concerns regarding chronic disease were more likely to perceive foreign outgroups as dangerous, and to have less positive attitudes toward immigrants.

While research directly examining these processes during the COVID-19 pandemic has been limited given the comparatively recent nature of the disease, research on other global health crises—like the bubonic plague (Markel, 1999), the 2014 Ebola outbreak (see Prati & Pietrantoni, 2016), and a host of other disease outbreaks (Cohn, 2012; Gover et al., 2020; McKiven, 2007)— has provided support for this conceptual link between personal fear and concern regarding the disease and outgroup prejudice. In one of the few empirical examinations of these processes during the pandemic, Mandalaywala et al. (2021) found evidence that personal perceptions of the intensity of COVID-19 were associated with less willingness to interact with racial outgroups, including both Asian and Black people; these patterns were also associated with increased indicators of prejudicial attitudes (see also, Tabri et al., 2020). Taken together, existing research has provided considerable evidence for the role of fear and racialized disease threat in exacerbating prejudicial outgroup attitudes during significant disease outbreaks.

A separate explanation, on the other hand, posits that these relationships are not just situated within the context of the disease itself, but in public discourse surrounding the disease. Unfortunately, narratives placing the onus for COVID-19 on Asian people are pervasive in the U.S., and this was particularly true at the beginning of the pandemic (Noel, 2020). Put simply, anti-Asian prejudice has been blatantly expressed by several public office holders, including the former President (Zhou, 2020), who disparagingly referred to the disease as the “Wuhan virus,” the “Chinese virus,” and even the “Kung Flu.” As Lee and Yadav (2020, p. 17) poignantly stated, “in one fell swoop, the coronavirus—and Trump's blithe description of it—reanimated a century-old racist trope that Asian Americans are vectors of filth and disease.”

This state- and nation-level xenophobic political rhetoric does not exist in a vacuum, however, and inevitably filters down to the local and individual level (see Piatkowska & Lantz, 2021). Specifically, the political legitimization perspective, or the related idea of “emboldenment” (see Dugan & Chenoweth, 2020), posits that negative government attention toward specific population groups can facilitate increases in biased behavior toward that group. Put another way, when government elites and political leaders signal supremacy over other groups, that signal serves to embolden members of the dominant group and to send the message that negative attitudes and behavior—including xenophobic attitudes, and even hate crimes—toward that group can be acted upon or committed without consequence (see Gagnon, 1995; Kalmoe, 2014; Kteily et al., 2015; Müller & Scwarz, 2020). As Dugan and Chenoweth (2020, p. 716) stated, some would-be bigots “perceive that the government is giving them a license to act on their anger.” Indeed, Kalmoe (2014) found that those with aggressive traits who were exposed to political advertisements with violent metaphors were more likely to support political violence, compared to those exposed to nonviolent ads.

This theoretical perspective is specifically applicable to the rhetoric of former President Trump and other high-profile political figures. As the highest elected political official in the US, the president has substantial power to impact and control societal discourse. Within the parlance of this legitimization framework, a number of scholars have pointed to former President Trump's rhetoric toward specific population groups—not exclusive to Asian people during the pandemic—as emboldening individual bias and hate-motivated violence (Dugan & Chenoweth, 2020; Müller & Scwarz, 2020). Indeed, his election campaign relied heavily on provoking antagonism toward Muslim people, Hispanic people, immigrants, and other minority groups. In this context, Müller and Scwarz (2020) found that Trump's Islamic-related tweets—which were frequently hostile or antagonistic toward the Muslim population—were associated with increases in anti-Muslim hate crimes. As a whole, this research provides support for the core notion of the legitimization perspective, which is that negative sentiment from the dominant group—and highly revered, political elites, in particular—has the potential to activate and embolden negative out-group sentiment among the general population. When coupled with disease threat, this emboldening discourse has the potential to exacerbate the negative consequences of fear, stoking individual racism. Put simply, when a group is labeled as dangerous or harmful by those in power, individuals are more likely to respond to them as such, and to perceive them as dangerous or harmful.

Taken together, while both mechanisms described thus far—responding to fear with “othering,” and anti-Asian political rhetoric—are different in process, they are similar in their goal: to place collective blame for the COVID-19 pandemic on a racial outgroup. “Othering” is inherently facilitated by the placement of blame and responsibility on an outgroup which, in the context of the COVID-19 pandemic, is Asian people (Nelkin & Gilman, 1988). The implicit goal of negative race-centered COVID-19 discourse is also to place—or displace—blame; indeed, some scholars have posited that the incitement of negative outgroup attitudes is intentional, with the goal of displacing blame. Gagnon (1995) argued that elite political figures are aware of their power and influence, such that they engage in negative outgroup discourse so as to intentionally incite violent intergroup conflict, because this conflict tends to serve a distractive function which draws attention away from other contentious or politically charged topics. Tolnay and Beck (1995) similarly noted that White political elites in the South benefited from intergroup conflict between the Black population and poor White population. In the context of the COVID-19 pandemic, the displacement of blame onto the Asian population—and Chinese people especially—may very well have served the dual function of assigning responsibility to an “other,” while simultaneously drawing attention away from the governmental response to the pandemic itself, which was perceived by many to be ineffective, or inefficient (Kim & Kreps, 2020; Miller et al., 2020; Rosenberg, 2021).

When viewed within the context of research on collective blame and retribution, this displacement of blame may have serious consequences. Lickel and colleagues (2006) argued that vicarious retribution, which is when a member of a group commits an act of aggression toward an outgroup member in response to a provocation or perceived provocation against other ingroup members, is often directed at outgroup members who have no direct influence on the perceived grievance. The first stage of this vicarious retribution framework, in particular, focuses on the initial event categorization and “act construal,” wherein individuals identify who caused the event, the “malevolence of their intentions,” and the degree of harm inflicted by the event. Applied to the COVID-19 pandemic, former President Trump and other major political figures simplified this part of the process, by explicitly telling their followers who they believed was responsible, and overtly blaming Asian people for the pandemic (see Fisher et al., 2019; Holland & Fermor, 2017). In this context, Lickel and colleagues (2006) also note that social influence from prominent ingroup members can have an important influence on this interpretation.

aken together, existing research has suggested that both disease threat-related fear and political legitimization may be important facilitators of anti-Asian prejudice during the COVID-19 pandemic. There is likely significant variation, however, in individual susceptibility to these mechanisms, a notion that has received far less attention. Recent research by Mandalaywala and colleagues (2021), for example, has demonstrated that some factors—like intergroup contact (see Pettigrew et al., 2010)—can mitigate the relationship between fear of COVID-19 and anti-Asian prejudice. As outlined previously, the social comprehension of disease is often rooted in prejudice, stigma, and the ostracization of an outgroup (Clissold et al., 2020); the “strength” of these roots varies, however, and one factor that is likely to be particularly important in predicting variation in individual susceptibility is the consumption of COVID-19-related information, which might reasonably be conceptualized as a proxy for individual access to alternative explanations and narratives.

Following this, the current research suggests that the consumption of information regarding the pandemic—which is likely to include information about the origins of the disease, as well as safe practices for reducing risk—might be a critical factor impacting differences in anti-Asian prejudice. There is some evidence for the importance of information in mitigating the effect of fear on negative outgroup attitudes during disease outbreak. Prati and Pietrantoni (2016), for example, found that increased knowledge and access to factual information during the 2014 Ebola outbreak was associated with decreased prejudice toward African immigrants, while low information, false information, and rumors were associated with increased prejudice. There is good reason, however, to suspect an important impact of information consumption on the legitimizing effect of racialized political discourse. As McVeigh (2004) noted in his seminal theory of structured ignorance, there are certain contexts in which individuals are more likely to be receptive or accepting of racist white supremacist discourse. While McVeigh was referring to neighborhood or contextual structure (e.g., inequality) as an important predictor of susceptibility to racist messaging, it is reasonable to assume that the consumption of factual information, or lack thereof, functions much the same. Specifically, McVeigh argued that contextual features impact receptivity to racist messaging by providing evidence consistent with white supremacist claims, without providing alternative explanations.

Applying a similar theoretical process to racist messaging and political discourse during the COVID-19 pandemic, it follows that (a) those who are exposed to racist political messaging, but are consuming less information regarding the disease—and are thus less exposed to alternative information—are likely to exhibit greater prejudicial attitudes; while (b) those who are exposed to racist messaging, but are consuming more information regarding the disease and are exposed to more alternative information, are likely to exhibit less prejudicial attitudes. In other words, anti-Asian racist rhetoric related to the COVID-19 pandemic is likely to be particularly appealing to those who lack alternative explanations and information sources. Likewise, the impacts of anti-Asian rhetoric are likely to be less potent among those who have access to information that can be used to reject the racist messaging (Nguyen et al., 2020). Put simply, those with more information are likely to be less inclined to diagnose the problem within a racial framework, and to attribute blame for the pandemic to an entire racial outgroup.

Our primary dependent measure is anti-Asian prejudice, which we measured using an indicator of anti-Asian xenophobia adapted from Van Der Veer et al. (2013). To construct this measure, respondents were asked to report their level of agreement with nine statements aimed at revealing anti-Asian sentiment: “Asian immigration in this country is out of control;” “Asian immigrants cause increase in crimes;” “Asian immigrants take jobs from people who are here already;” “Interacting with Asian immigrants makes me uneasy;” “I worry that Asian immigrants may spread unusual diseases;” “I am afraid that in case of war or political tension, Asian immigrants will be loyal to their own country of origin;” “With increased Asian immigration I fear that our way of life will change for the worse;” “I doubt that Asian immigrants will put the interest of this country first;” and “I am afraid that our own culture will be lost with increases in Asian immigration.” For each statement, respondents were given a 5-point Likert scale ranging from strongly disagree to strongly agree; we then averaged the 5-point scales across each of the indicators to create a summary measure of anti-Asian xenophobia (α = 0.95).

The current research uses three primary independent measures. First, we include a measure intended to assess individual-level fear of COVID-19. Specifically, we account for individual assessments of the personal risk respondents believe they face from COVID-19 using a question asking each respondent: “What level of threat do you think COVID-19 poses to you or your family?” This measure was operationalized as a continuous measure, such that higher values indicate higher assessments of risk. Political attitudes, which serve as a proxy for the potential influence of political legitimization in the current study, were measured using an indicator of voting behavior; specifically, respondents were asked the following: “Did you vote in the 2016 Presidential Election?” Respondents then chose from the following options: “yes,” “no,” “don't remember,” or “prefer not to say.” If the respondent answered “yes,” they were then asked: “If so, who did you vote for?” Respondents chose from “Donald Trump,” “Hillary Clinton,” “other candidate,” or “don't know.” This measure was then operationalized as a dummy measure equal to 1 if the respondent indicated they voted for “Donald Trump” and equal to 0 if the respondent indicated they voted for “Hillary Clinton” or “other candidate.” Those who did not vote or refused to share voting information were excluded from the analyses (see Malcom et al., 2022 for a similar approach). Third, we assess the role of information consumption. Specifically, we account for the amount of information about COVID-19 that the respondent has been exposed to using a question asking: “How much information have you seen, read, or heard about COVID-19?” Respondents were then given the following four options: “nothing at all,” “not very much,” “a fair amount,” or “a great deal.” This measure was then operationalized as a continuous measure, wherein higher values indicate that respondents report seeing, reading, or hearing more about COVID-19 and lower values indicate seeing, reading, or hearing less about COVID-19.

We also included a number of control measures which might be related to our research questions. Respondent race/ethnicity was measured using three questions from the survey, which asked respondents whether they are of Spanish, Latino, or Hispanic origin (yes/no/I don't know); which racial category describes their race (with the option to select more than one); and then which racial category they identify with most (if they indicated more than one racial category in the previous question). Response options for this question included White, Black, Asian, Native Hawaiian or other Pacific Islander, American Indian or Alaska Native, and other race; all respondents who indicated that they identified as Asian were excluded from the study, given our focus on anti-Asian attitudes. All remaining respondents who answered yes to the first question were coded as Hispanic. Respondents were coded as non-Hispanic Black if they indicated Black as one of their racial identities and the racial category they identify with most, coded as non-Hispanic white if they indicated white as one of their racial identities and the racial category they identify with most; and coded as other race if they did not fall into one of these group. White was used as the reference category in all multivariate regression models.

Several additional respondent characteristics are included as control measures as well, including age, whether the respondent has a high school education or less (1 = yes), whether the respondent was employed (1 = yes), and whether the respondent identified as Protestant or Catholic (1 = yes).² Respondents were also asked to indicate their gender/gender identity and were given the option of selecting male; female; transgender male; transgender female; gender variant/non-conforming; other; and prefer not to answer. Gender was then coded such that respondents who identified as “male” or “transgender male” were coded as male (=0), while those who identified as “female” or “transgender female” were coded as female (=1). Because of the small number of observations, all other respondents were coded as missing (n = 13). We controlled for foreign-born status by including a dichotomous measure indicating whether respondents were born in the United States (1 = yes). Next, we control for right-wing authoritarianism (RWA), given the demonstrated relationship between RWA and prejudicial attitudes (e.g., Altemeyer & Altemeyer, 1996). Specifically, we measure this construct using a six-item scale adapted from Bizumic and Duckitt (2018), in which respondents are asked to what extent they agree with each of the following statements: “it's great the many young people today are prepared to defy authority,” (reverse coded), “what our country needs most is discipline, with everyone following our leaders in unity,” “God's laws about abortion, pornography, and marriage must be strictly followed before it's too late,” “there is nothing wrong with premarital sexual intercourse (reverse coded),” “our society does NOT need tough government and stricter laws (reverse coded),” and “the facts on crime and the recent public disorders show we have to crack down harder on troublemakers, if we are going to preserve law and order.” Responses were averaged to create an index ranging from 0 to 4, with higher values signifying more right-wing authoritarian attitudes (α = 0.64). Finally, because perceptions of risk and related attitudes might be related to such behavior, we account for the extent of socializing engaged in at the time of the survey using a continuous measure where higher values indicate more socializing (0 = complete social distancing).

The current research proceeds in four steps. First, we conduct an OLS regression model in which anti-Asian xenophobic attitudes is regressed on our independent measures, in order to assess the main effects of each of our primary independent measures. Second, we include all independent variables and an interaction between perceptions of COVID-19 risk and political attitudes. Third, we include all independent variables and an interaction between information consumption and political attitudes, respectively. Finally, we present each of these interactions graphically.