Understanding anti-Asian sentiment and political behavior in the wake of COVID-19

ABSTRACT

Given media reports of racism and hate crimes against Asians since the start of the COVID-19 pandemic, we examine whether explicit attitudes toward Asians in the United States changed more generally following the start of the pandemic. We compare two national samples of Americans before and after the onset of the pandemic, as well as replicating our findings in a panel dataset that spans the onset of the pandemic. We find that Americans’ feelings toward Asian Americans – but not toward other racially minoritized groups – became more negative after the onset of the pandemic. This heightened negative sentiment toward Asians is observed regardless of political ideology. Moreover, we find that the degree of exposure to the coronavirus is associated with anti-Asian attitudes, suggesting that the circumstances of the pandemic are related to increased anti-Asian attitudes. Finally, across model specifications, anti-Asian attitudes robustly increase the probability of voting for Donald Trump in the 2020 presidential election. Democrats with highly anti-Asian attitudes are nearly as likely as Republicans to vote for Trump, weakening the effect of partisanship on vote choice. This research suggests that the pandemic is exacerbating social inequalities in the U.S. in part through shifting racial attitudes – and with political consequences.

KEYWORDS:

• racism
• anti-Asian attitudes
• attitude change‌
• COVID-19
• pandemic‌
• voting

Acknowledgements

We are grateful to Andrew Engelhardt, Stanley Feldman, Michelle Io-Low, Samuel Jens, Yanna Krupnikov, Neil Malhotra, Katherine McCabe, Jennifer Merolla, Joseph Sandor, Payel Sen, and Joseph Vitriol for helpful comments on this manuscript.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Ethics approval

The studies in this paper were approved by the Stony Brook University Institutional Review Board for Human Subject Research, under IRB #s 1167088 and 2020-00308. All participants provided written online consent prior to participating in the study.

Data availability statement

Data, materials, and analysis code are available at https://osf.io/f9xev/.

Notes

1 Centers for Disease Control and Prevention, COVID-19 Tracker. Accessed online September 28, 2022: https://covid.cdc.gov/covid-data-tracker/#datatracker-home.

2 Such rhetoric also increased Democratic Party favorability among Asian Americans, as social exclusion turned the group away from the Republican Party (Chan, Kim, and Leung 2022). Social exclusion, as well as the role taken on by ethnicity-based organizations to provide assistance and services to AAPI communities facing harassment, discrimination, and violence, can empower political participation in these communities moving forward (Sadhwani and Kulkarni 2021).

3 https://www.washingtonpost.com/nation/2020/03/19/coronavirus-china-cornyn-blame/.

4 Whereas such racialized rhetoric about COVID-19 tends to be specific to China, we suspect that pandemic-related racial attitudes are likely applied to (East and Southeast) Asians more broadly, in part due to perceptions of racial outgroup homogeneity (e.g., Judd, Ryan, and Park 1991; Ostrom and Sedikides 1992), whereby members of an outgroup are seen as less distinguishable from one another than members of one’s ingroup.

5 So-called “positive” stereotypes about Asians as a monolithic “model minority” that achieves academic and economic success (Chou and Feagin 2008; Chao et al. 2013) bolster the perception that Asians do not experience racial discrimination or need governmental assistance, thereby justifying broader dismissal of Asians’ concerns (Ancheta 2006; Gee et al. 2009; Saito 1997).

6 Although not the primary focus of our research questions, our data also allow us to explore questions of variability in anti-Asian sentiment across different demographic groups. To our knowledge, there is very little – if any – existing work that has compared how different racial groups express attitudes toward Asians, much less in the context of the COVID-19 pandemic (but see Tan, Lee, and Ruppanner 2021 for an examination of anti-Asian sentiment across socioeconomic and partisan identities). Indeed, most research on anti-Asian attitudes focus on white respondents or typically treat racial group as a control variable (e.g., Mandalaywala, Gonzalez, and Tropp 2021; Reny and Barreto 2022). In our data, we explore racial subgroup analyses on anti-Asian sentiment (see Appendix Tables 7–9).

7 We also suspect that racist rhetoric about the novel coronavirus set the stage for pathogen anxiety to merge with anti-Asian sentiment – especially given the salience of such rhetoric (as we note above) – but it should be noted that we do not have direct measures for elite rhetoric in our data.

8 It may be worth noting that prior observations of links between pathogen concern and political outcomes have suggested an asymmetric ideological or partisan effect – that is, an association between pathogen concerns and socially conservative preferences (e.g., Aarøe, Petersen, and Arceneaux 2017, 2020). One aim of the current research is to explore whether liberals and Democrats are attitudinally immune to the influences of pathogen exposure, or if the impact of COVID-19 on sociopolitical attitudes lands across the political spectrum.

9 Our pre-registration document can be reviewed at https://aspredicted.org/f95na.pdf. It should be noted that the pre-registration also includes an experiment, which is a separate study that is outside the scope of this article. Data and analysis code are available at https://osf.io/f9xev/.

10 Although one might expect that Asian Americans would not report increased negative perceptions of their own racial group following the onset of the pandemic, to reflect the agency and attitudes of all participants in this study we retain participant responses from all people of color in our primary analyses reported in-text. However, in additional robustness checks, we restrict our analyses to all groups except Asian Americans (see Appendix Table 4) and to white Americans only (Appendix Table 5; for results compared to changes in other subgroup feeling thermometers, see Appendix Table 6). In additional exploratory analyses, we further restrict our sample to Native Americans only (Appendix Table 7), Black Americans only (Appendix Table 8), and Latinx Americans only (Appendix Table 9). These exploratory analyses suggest that while Latinx Americans, similar to non-Latinx white Americans, report an increase in negative feelings toward Asians post-pandemic onset, not all subgroups similarly report increased negative feelings toward Asians. However, given the small sample size, particularly of the Native American subgroup, we caution against overgeneralizing from these analyses.

11 In our initial round of data collection (pre-pandemic), we collected data on political ideology but not partisanship. However, in our second round of data collection (post-pandemic onset), we measured both ideology and partisanship. As our theory relates more to partisanship, we report the results of our main analyses in Tables 2 and 3 using our measure of partisanship. However, for consistency across analytic approaches, in the Appendix, we additionally explore the effect of political ideology on anti-Asian attitudes and probability of voting for Trump (e.g., see Appendix Table 14 and Appendix Table 23).

12 Parallel models examining changes in feeling thermometers toward Latinx and Arab Americans during the same period do not suggest significant negative shifts in sentiment (p > 0.10); however, results using these data suggest a statistically significant negative shift in sentiment toward Black Americans (p < .01), perhaps a function of the second survey occurring some months after the murder of George Floyd and shifting sentiments about the resulting widespread Black Lives Matter activism.

13 Feeling thermometers have frequently been used to measure a range of intergroup attitudes since being added to the 1964 American National Election Study, but the measure has been critiqued with concerns about intraindividual response validity and interindividual response consistency (e.g., Regenwetter, Hsu, and Kuklinski 2019; Wilcox, Sigelman, and Cook 1989). Although no measure is perfect, we therefore sought to go beyond feeling thermometers in assessing sentiment toward Asians in the next analysis.

14 The intent of this adapted measure is similar to the Asian American Resentment Scale that was introduced by Kim (2021) insofar as we are attempting to assess attitudes toward Asians in the American context that hew closely to but are distinct from attitudes toward Blacks (e.g., an item from our measure states “Asians have more influence upon school admissions than they ought to have”). When we fielded our survey in June 2020, the Asian American Resentment Scale (Kim 2021) was not, to our knowledge, publicly available.

15 Individuals who record a personal diagnosis of COVID-19 or COVID-19 consistent symptoms in addition to experience with family, friend(s), coworker(s), or acquaintance(s) with COVID-19 or COVID-19 consistent symptoms are coded as 2.

16 As a robustness check, we ran the analyses reported in Table 2 (and the analogous robustness checks) with two categorical variables of COVID-19 exposure: personal experience with COVID-19 and other experience with COVID-19 (with no exposure to COVID-19 as the baseline). The results lend further support to the pathogen exposure hypothesis: more personal, direct exposure is associated with increased anti-Asian attitudes while more indirect exposure does not have a similar effect. See Appendix Tables 12, 13, and 14.

17 Partisanship is a self-reported measure of party identification from 1 to 3; 1 = Democrat, 2 = Independent, 3 = Republican. As an additional robustness check, we report the results using dichotomous variables for Republican and Democrat (with Independent as the baseline reference group). The results are robust across model specifications. Please see Appendix Table 16.

18 In addition, we report the results using dichotomous variables for Republican and Democrat (with Independent as the baseline). The results are robust across model specifications. Please see Appendix Table 20.

19 Parallel models examining an interaction between political ideology and anti-Asian attitudes (as opposed to partisanship) show consistent results, such that at high levels of anti-Asian attitudes, liberals are nearly as likely as conservatives to vote for Trump (see Appendix Table 23 and Appendix Figure 1). The relationship between anti-Asian attitudes and voting for Trump holds across various model specifications, including accounting for more general outgroup sentiments (Appendix Table 24), exposure to COVID-19 (Appendix Table 25), as well as the interaction between exposure to COVID-19 and anti-Asian attitudes (Appendix Table 28).

20 We believe that this partisanship interaction is not a model artifact, as our sample includes Republicans, Democrats, and Independents who express highly anti-Asian attitudes and are voting for Trump. See Appendix Tables 26 and 27 for proportions of Republicans, Democrats, and independents expressing high vs. low anti-Asian attitudes by self-reported plan to vote for Trump.