Unequal burdens: Mental health effects of anti-Asian hate incidents across ecological contexts

Abstract

Reports of anti-Asian hate incidents (AAHIs) increased during the COVID-19 pandemic, yet their mental health consequences and the contextual factors shaping them remain insufficiently understood. Guided by Ecological Systems Theory, this study examines whether changes in AAHI exposure are associated with anxiety and depression symptoms and whether this association varies across ecological contexts. We merged respondent-level data from the Understanding America Study COVID-19 panel (waves 3–34; March 2020–July 2021) with FBI hate-crime records. The analytic sample included 6,552 adults contributing 152,276 person-wave observations. Individual fixed-effects models were estimated to isolate within-person changes in mental health associated with changes in hate incidents occurring in the prior survey wave. Models adjusted for time trends and used multi-way clustered standard errors by respondent, county of residence, and survey wave. Moderators captured microsystem (e.g., victim–respondent gender match), mesosystem (e.g., county COVID-19 deaths), exosystem (e.g., incident characteristics), and macrosystem (e.g., county political affiliation). Increases in AAHIs were associated with statistically significant but modest increases in anxiety and depression symptoms. A one-unit increase in lagged AAHI exposure corresponded to a 0.06-point increase in PHQ-4 scores (0–12 scale), indicating a small effect size relative to within-person variation in mental health. Associations varied modestly by incident characteristics, particularly shared victim–respondent gender and physical injury, while COVID-19 mortality and partisan composition showed limited moderation. Targeted, trauma-informed interventions may help mitigate the mental health burden of hate exposure during periods of social instability.

Highlights

• •AAHIs linked to small increases in PHQ-4 symptoms.
• •Shared gender and Asian identity modestly amplified effects.
• •Injury-related incidents showed higher distress levels.
• •COVID deaths and politics showed little moderation.

1. Introduction

1.1. Anti-Asian hate incidents during the COVID-19 pandemic

The COVID-19 pandemic fueled a surge in anti-Asian hate incidents (AAHIs) in the United States (U.S.) (Han et al., 2023). Yoon et al. (2025) documented substantial increases in AAHIs reported to media and law enforcement beginning shortly after the first confirmed U.S. COVID-19 case, with incidents peaking in March 2020 amid intensifying politicized discourse. A secondary peak occurred in March-April 2021, followed by a gradual decline as public health conditions stabilized (Yoon et al., 2025). During this period, misinformation and xenophobic narratives that associated the virus with Asian populations contributed to marked increases in verbal harassment, physical assaults, and other forms of discrimination targeting Asian individuals (Lantz & Wenger, 2023; Roberto et al., 2020; Vidgen et al., 2020, November). This rise in anti-Asian sentiment was driven by stereotypes and fear-mongering rhetoric, which spread rapidly through social media, political discourse, and mainstream news outlets (Gomera, 2020; John & Guy, 2020; Palmer, 2020; Mackenzie & Smith, 2020; Zheng & Zompetti, 2023).

Consistent with these trends, hate crime offenses in the U.S. increased by approximately 13% from 2019 to 2020, and roughly 62% of incidents were motivated by race, ethnicity, or ancestry in 2020 (Federal Bureau of Investigation [FBI], 2022; FBI, 2023). Particularly, according to the FBI Hate Crime Statistics, AAHIs increased by 77%, from 158 in 2019 to 279 in 2020 (FBI, 2022). Using both FBI UCR hate crime data and media-based reports, Yoon et al. (2025) documented that during 2020-2021, the total number of AAHIs recorded in the FBI data was 1,086, while the total number identified through media courses was 1,288, highlighting both the substantial rise in AAHIs during the pandemic and the potential underrepresentation of incidents in official law enforcement records. A study further reported that about 32.3% of the Asian participants experienced non-criminal bias victimization and about 19.5% experienced criminal bias victimization in the U.S. during the pandemic (Lantz & Wenger, 2023).

1.2. Factors that influence the relationship between AAHI and mental health

Experiences of hate incidents, discrimination, and racism during the pandemic have been linked to increased stress, stigma consciousness, depression, and anxiety (Ertorer, 2024; Harii et al., 2025). In response, individuals reported modifying their daily routines (e.g., avoiding public spaces, staying indoors, or only going out when accompanied by white family members) to minimize the risk of encountering discrimination (Hahm et al., 2021). Even without direct personal contact with a specific incident, individuals may experience psychological distress through indirect exposure pathways. Indirect exposure to hate incidents through media coverage and community communication can contribute to a broader climate of fear and psychological distress, while fostering distrust, skepticism, and fatigue toward mainstream media sources (Adversario, 2024; Yu et al., 2020). Such indirect exposure can heighten perceived vulnerability, anticipatory stress, and vigilance, thereby affecting mental health even when incidents are not directly experienced or personally witnessed.

More broadly, the impact of crime on mental health varies based on individual and contextual factors. Exposure to crime is linked to elevated risks of anxiety and depression, with stronger effects observed among women and those living in areas with low employment rates (Dustmann et al., 2016; Morrall et al., 2010). Financial hardship, experienced even by employed individuals with criminal records, has been associated with heightened stress levels (Pogrebin et al., 2014). Among criminal victims, strong social support and positive perceptions of neighborhood connectedness serve as protective factors against adverse mental health outcomes (Green & Pomeroy, 2007; Kilian et al., 2021). Effective therapeutic interventions have also been shown to reduce psychological distress related to victimization (Ford et al., 2013). Furthermore, exposure to violence can worsen mental health by increasing perceived risk of future victimization, reducing feelings of personal safety, and lowering perceived control over preventing future incidents (Kilian et al., 2021).

1.3. Theoretical framework

This study adopts Ecological Systems Theory (Bronfenbrenner, 1974, 1977) to examine how factors related to AAHIs are associated with mental health outcomes. This framework is particularly relevant as it explores the dynamic interplay between individuals and their environments across multiple ecological levels. By accounting for both personal attributes and contextual influences, the theory offers a robust foundation for analyzing the complex mechanisms linking AAHIs to psychological well-being. Specifically, Ecological Systems Theory posits that mental health is shaped by a series of nested and interconnected factors, including the microsystem, mesosystem, exosystem, macrosystem, and chronosystem (Bronfenbrenner, 1974, 1977). This multi-level perspective enables a nuanced understanding of how environmental and social factors jointly affect mental health in the context of racialized experiences.

Microsystem. The microsystem includes an individual's immediate surroundings and direct experiences that influence their psychological functioning (Bronfenbrenner, 1974, 1977). In the context of this study, three key components of the microsystem are considered: physical proximity to the AAHI location, similarity between victim and respondent characteristics (e.g., gender), and respondents' race/ethnicity.

Proximity to an AAHI can shape whether an individual directly witnesses the incident, hears about it firsthand, or interacts with someone affected. Such close exposure may increase the psychological salience of the event, potentially intensifying emotional responses, such as fear, anxiety, or stress. Although existing research has not directly examined the relationship between physical proximity to AAHIs and mental health during the COVID-19 pandemic, studies on general crime support the significance of spatial context. For example, Pak and Gannon (2023) classified crime exposure into immediate neighborhood versus surrounding areas, finding that closer exposure had a stronger association with mental well-being. Similarly, psychological proximity defined as believing one is at risk of victimization in a current location was significantly linked to heightened worry and distress (Mellberg, 2024).

Another critical microsystem factor is victim-respondent similarity, particularly shared characteristics such as gender. Perceived similarity can influence how individuals process incidents of violence or discrimination. When individuals identify with victims, they may experience heightened empathy or fear, which in turn can amplify mental health impacts. For example, anxiety and fear of crime often vary by crime type and victim gender (Schafer et al., 2006; Sutton & Farrall, 2005). Women are more likely to fear crimes involving personal relationships, such as stalking or sexual assault (Fox et al., 2009), while men exhibit greater fear when crimes primarily target males (Schafer et al., 2006). These dynamics may similarly apply in the context of AAHIs, as gender-matching between victims and observers could intensify psychological distress. However, empirical research remains limited on the mental health effects of gender matching in AAHI contexts, particularly during the COVID-19 pandemic.

Prior research further suggests that individuals often experience stronger psychological reactions to hate incidents and racialized violence when victims belong to their own racial or ethnic group, reflecting intergroup empathy bias and heightened perceptions of shared vulnerability. Population-based evidence shows that highly publicized racial violence can generate measurable mental health spillover effects among members of the targeted group even without direct victimization; for example, police killings of unarmed Black Americans were associated with significant increases in poor mental health days among Black adults, but not White adults (Bor et al., 2018), and exposure to widely publicized anti-Black violence similarly predicted worse mental health among Black Americans (Curtis et al., 2021). Beyond direct exposure, indirect pathways such as media coverage and social network transmission may amplify distress by fostering anticipatory stress, vigilance, and fear of future victimization, contributing to vicarious trauma and broader community-level psychological burden (Moody et al., 2023). Consistent with this mechanism, scholarship on intergroup empathy demonstrates that individuals typically display reduced empathic responses to outgroup suffering compared to ingroup suffering, which may partially explain heterogeneous psychological responses across racial/ethnic groups (Cikara et al., 2011, 2014). At the same time, research highlights that repeated exposure to discriminatory content and racialized incidents through online and mainstream media can adversely affect mental health, including among Asian Americans during the COVID-19 pandemic, indicating that indirect exposure can produce substantial distress even in the absence of personal contact with hate incidents (Layug et al., 2022).

Mesosystem. The mesosystem encompasses the interactions among multiple elements of an individual's microsystem (Bronfenbrenner, 1974, 1977). In this study, COVID-19 death rates are conceptualized as a key component of the mesosystem, reflecting a broader, health-related stressor that simultaneously affects various domains of daily life—including community, healthcare, and interpersonal relationships. During the pandemic, high mortality rates intensified societal fear, stress, and uncertainty. This period also saw a notable rise in anti-Asian racism and hate incidents (Tan et al., 2022), further complicating the psychological environment in which individuals experienced and responded to AAHIs. As a result, COVID-19 death rates are not only markers of physical health risk but also serve as contextual amplifiers of emotional and psychological vulnerability.

Increased COVID-19 death rates often triggered stricter public health measures (e.g., social distancing and lockdowns), which contributed to reduced social interaction and heightened loneliness, factors strongly associated with depression (Banerjee et al., 2020; Kovacs et al., 2021; Owczarek et al., 2022). Furthermore, elevated mortality heightened anxiety about personal safety, fear of contagion, and feelings of helplessness amid a prolonged crisis (Menzies & Menzies, 2020). When experienced alongside AAHIs, these overlapping stressors may compound the risk of adverse mental health outcomes by exacerbating feelings of isolation, fear, and perceived threat.

Exosystem. The exosystem refers to environments or systems that individuals may not directly engage with, but that indirectly influence their thoughts, behaviors, and emotional responses by shaping the broader context in which their immediate environments (microsystems) function (Bronfenbrenner, 1974, 1977). In this study, components of the exosystem include AAHIs involving racial/ethnic minor offenders, male victims, elderly victims, and injured victims. Although individuals may not interact personally with those involved in these incidents, such factors can still shape perceptions and responses to AAHIs through media portrayals, institutional reactions, and broader community narratives.

When considering the psychological consequences of AAHIs, the perceived identity of the perpetrator may be a critical element. Integrated Threat Theory (Stephan et al., 2000) posits that intergroup dynamics are shaped by two overarching forms of perceived threat: realistic threats, which include fears of bodily harm or economic disadvantage, and symbolic threats, which involve worries that an out-group undermines an in-group's values and sense of identity. Acts of aggression carried out by members of racially or ethnically dominant groups have the potential to elicit both types of threat simultaneously, thereby intensifying victims' feelings of exposure and marginalization. Although Pallone and Hennessy (2000) examined the racial distribution of offenders in the U.S. crime statistics, no research has yet addressed how the racial identity of offenders in AAHIs specifically affects mental health outcomes, pointing to a significant gap for future inquiry.

Victim characteristics, including gender and age, may moderate the link between AAHIs and psychological outcomes. Social Role Theory (Eagly & Wood, 2012) suggests that men are often expected to embody protective roles; thus, when men experience vulnerability, the incongruity with this social expectation can exacerbate psychological distress—particularly when AAHIs accumulate. Age norms operate in parallel ways. Older adults are frequently associated with authority, wisdom, and heightened physical frailty (North & Fiske, 2012), whereas younger adults are generally perceived as stronger, more self-reliant, and better equipped to safeguard themselves. Consequently, AAHIs involving younger victims may violate age-based social expectations, intensifying observers’ empathic responses and perceived injustice.

Moreover, the occurrence of physical injury can markedly intensify the psychological impact of AAHIs. When such incidents involve bodily harm, victims face not only physical consequences but also heightened emotional trauma, increasing the risk of post-traumatic stress symptoms, persistent feelings of vulnerability, and chronic fear (Kerig et al., 2016). Injuries may also create enduring challenges, including ongoing pain, limited mobility, and financial pressures stemming from medical treatment or loss of income, all of which can further compromise psychological well-being. Importantly, the effects extend beyond those directly harmed: individuals who witness or learn about violent AAHIs may also experience heightened anxiety and distress, contributing to a pervasive climate of fear and amplifying the collective psychological burden of racially motivated violence.

Macrosystem. The macrosystem encompasses the broader cultural, societal, and ideological context in which individuals and communities are embedded (Bronfenbrenner, 1974, 1977). This includes prevailing social norms, institutional ideologies, and political orientations. In the context of this study, political affiliation, particularly at the county level, serves as a key macrosystem factor. Research suggests that counties with higher Republican vote shares may foster more hostile or unsupportive environments for Asian American and Pacific Islander (AAPI) communities, often characterized by systemic racism and negative societal attitudes (Malcom et al., 2023; Yoo et al., 2023), which can exacerbate psychological distress.

Political affiliation may moderate the relationship between AAHIs and mental health by shaping both the frequency of discriminatory incidents and the extent of social or institutional support available to affected individuals. In conservative-leaning areas, reduced support for hate crime legislation (Malcom et al., 2023) and higher levels of implicit bias against AAPIs (Yoo et al., 2023) may contribute to a greater prevalence and visibility of AAHIs. Furthermore, when prejudiced rhetoric is normalized or endorsed by political leaders (Newman et al., 2021), it can embolden discriminatory behavior, thereby increasing stress, fear, and anxiety among AAPI individuals. In such environments, the absence of strong legal protections and clear societal condemnation may deepen feelings of vulnerability and social exclusion, compounding the mental health burden associated with hate incidents.

Chronosystem. Chronosystem captures changes in anxiety and depression over time as AAHI frequencies fluctuate or societal responses evolve (Ertorer, 2024; Harii et al., 2025). This temporal dimension highlights how ongoing exposure and contextual shifts influence mental health outcomes.

1.4. The current study

Previous research has investigated how exposure to crime affects mental health outcomes (Dustmann et al., 2016; Morrall et al., 2010). Yet, little is known about the moderating factors related to the relationship between hate incidents and individuals’ psychological well-being, particularly in the context of AAHIs during the COVID-19 pandemic. This study is innovative as it examines how AAHI counts during the pandemic are associated with residents' anxiety and depression at an individual level, using a nationally representative panel dataset. This study contributes to existing literature by leveraging high-frequency panel data collected at bi-weekly intervals during most of the study period. This approach enables researchers to track changes in anxiety and depression levels associated with variations in AAHI counts over time during the COVID-19 pandemic. Additionally, this study examines heterogeneous relationships between AAHI counts and mental health outcomes using moderators informed by Ecological Systems Theory. To the best of our knowledge, this is the first study to analyze the differential effects of offender and victim characteristics, the similarity between victims and respondents, and the physical proximity of incidents to respondents of AAHIs that occur on the relationship between the AAHIs and mental health. Understanding these factors is critical for informing mental health interventions, community support efforts, and policy responses aimed at addressing the psychological toll of AAHIs.

2. Methods

2.1. Data and sample

This study utilizes individual-level data from the Understanding America Study (UAS) Coronavirus in America (COVID) dataset, combined with data on AAHIs made publicly available by the Federal Bureau of Investigation (FBI). Details regarding these datasets and their use in the analysis are provided in the following section.

2.1.1. Law enforcement data

This study utilized data on AAHIs from the FBI for the years 2020 and 2021. The FBI dataset provides comprehensive hate crime information, including bias type and motivation, details of criminal acts, demographics of offenders and victims (such as race/ethnicity, age, and gender), incident types, and locations. The data collection process employed by the FBI involved a two-step procedure: first, a responding officer determined whether an incident was suspected to be bias-motivated; if so, the case was reviewed by a designated judgment officer or unit before being submitted to the FBI's Uniform Crime Reporting (UCR) program (GLESS & CLESU, 2022).

Although the Hate Crime Statistics Act of 1990 requires the Attorney General to collect hate crime data from law enforcement agencies at all levels, participation in the FBI UCR program remains voluntary (GLESS & CLESU, 2022). As a result, the dataset may not fully capture all bias-motivated offenses, and underreporting is likely to vary systematically across counties. Prior research suggests that zero counts in FBI hate crime data may reflect non-reporting rather than true absence of incidents, and that reporting patterns may be associated with county political context (Mills et al., 2024). Thus, measurement error in AAHI exposure may be nonrandom and could bias estimated associations. Additionally, incidents lacking address information were excluded, which may further reduce completeness.

2.1.2. Individual-level data

This study utilized waves 3 through 34 of the UAS longitudinal panel sample. The analysis began with the third wave, as it was the first to include questions on the outcome variables of interest: anxiety and depression levels. Administered by the University of Southern California (USC), the UAS COVID-19 survey began collecting data on March 10, 2020, using a probability-based internet panel comprising approximately 7,100 nationally representative participants aged 18 or older (Kapteyn et al., 2020a). Surveys were conducted bi-weekly until February 16, 2021, after which they transitioned to a monthly schedule from February 17 to July 20, 2021 (Kapteyn et al., 2020b). The study achieved a high participation rate of approximately 88% (Kapteyn et al., 2020b). To minimize sampling attrition and ensure inclusivity, UAS provided tablets with internet access to households without the necessary technology (Kapteyn et al., 2020b).

The UAS data measure respondents' physical and mental health, employment, education, future outlooks, and COVID-19-related factors. (USC Dornsife Center for Economic and Social Research, 2024). The UAS data are particularly relevant for individual-level analyses in the event of AAHIs because the high frequency of data collection allows for tracking changes in mental health over time. Additionally, the dataset provides rich information about COVID-19-related factors, offering a comprehensive understanding of the pandemic's impact on individuals.

2.1.3. Merging the two datasets

For most of the empirical analyses, this study merged individual-level UAS data with AAHI exposure measures derived from the FBI UCR hate crime database. The FBI UCR data are reported to the FBI on a monthly basis by participating agencies and include the dates of incidents (FBI, 2001), whereas the UAS COVID-19 survey was conducted bi-weekly until February 16, 2021 and monthly thereafter (Kapteyn et al., 2020b). To harmonize these frequencies, we aggregated FBI hate crime incidents into totals corresponding to each UAS survey wave using the wave-specific field dates. Specifically, for each wave, we summed the number of AAHIs occurring during the time interval covered by that wave. This wave-level exposure measure was then merged to all respondents who completed the survey in the subsequent wave to minimize timing mismatch.

In addition, to examine heterogeneity based on geographic proximity, we constructed a county-level exposure measure using respondents’ county of residence (available in the restricted UAS dataset) and FBI hate crime incidents geocoded at the county level. This county-level measure was merged to respondents by wave and county of residence. To access the restricted data for analysis, our research team completed the required application process, ensuring full compliance with data access protocols. Together, these approaches allow us to capture both broader population-level exposure to AAHIs during the pandemic (e.g., through media and social networks) and geographically proximate exposure.

2.1.4. Sample characteristics

After merging the FBI's AAHI data with the UAS's individual-level data, the baseline analyses included 6,552 respondents and 152,276 observations during the study period. The analytic sample includes respondents from a wide range of U.S. counties, with coverage across all major Census regions. Because identification relies on within-person changes over time, geographic distribution is not central to model estimation, but county identifiers are used to link contextual exposure and to cluster standard errors. The sample characteristics are summarized in Table 1. Proportions and means were calculated without weighting, taking into account the panel structure of the data. Note that these summary statistics are based on observations, not respondents. Consequently, time-variant variables, such as household income and COVID-19 diagnosis, reflect the characteristics of the observations, while time-invariant factors, such as gender and race/ethnicity, correspond to the respondents' characteristics.

Table 1.

Sample characteristics.

	%/Mean (S.D.)
Gender
Men	42.16
Women	57.84
Age	53.25 (15.63)
Race/ethnicity
Whit	74.11
Black	7.56
Hispanic	9.82
Asian	3.59
Native American	1.00
Mixed	3.92
Educational attainment
Less than a high school diploma	4.50
High school diploma	17.12
Some college education	36.46
Bachelor's degree	23.77
Graduate	18.14
Marital status
Married	58.81
Separated/Divorced/Widowed	22.59
Never married	18.60
Employment status
Employed	51.81
Unemployed	18.66
Retired	22.03
Mixed	7.50
Household income
< $30K	22.39
$30K-$59,999	25.25
$60K-$99,999	25.99
$100K or more	26.36
Renters	25.04
COVID-19 diagnosis	1.60
Health insurance ownership	91.71
Number of household members	1.66 (1.41)

Note. Unweighted estimates. N = 6,552; N-wave: 152,276; 3-34 waves of the UAS COVID dataset. Standard deviations and group membership percentages are calculated across all respondent–wave observations, reflecting the panel structure of the data.

Slightly more than half of the respondents were female (58%). Approximately two-thirds of the respondents self-identified as non-Hispanic White (74%). The proportion of Hispanic respondents (10%) appeared higher than in the general population, likely due to the oversampling of California residents. Slightly more than two-thirds of the observations reported having attained at least some college education (78%). Additionally, more than half of the observations were married (59%) and employed (52%) during the study period. Only a small proportion of observations were diagnosed with COVID-19 during the study period (2%), while most had health insurance (92%).

Table A1 of the Online Supplementary Materials compares characteristics of respondents included in the analytic sample and those excluded due to missing county-level geographic identifiers. Overall, excluded respondents differ modestly but systematically from included respondents along several dimensions. The excluded sample is, on average, younger, reports slightly higher PHQ-4 scores, and is more likely to be unemployed or never married. Excluded respondents are also less likely to hold a graduate degree and more likely to rent their homes. In contrast, differences in gender, race/ethnicity, household income, COVID-19 diagnosis, and health insurance coverage are generally small and statistically insignificant. Importantly, while some differences reach statistical significance due to the large number of observations, the absolute magnitudes of most differences are modest. These patterns suggest that exclusion due to missing geographic identifiers is associated primarily with life-course and socioeconomic characteristics rather than extreme differences in mental health or health status.

Although Table A1 indicates modest differences between included and excluded respondents, these differences do not threaten the internal validity of the analyses. The primary analyses rely on individual fixed-effects models, which identify associations from within-person changes over time and are therefore unlikely to be biased by cross-sectional differences between samples.

2.2. Variables

2.2.1. Dependent variable

The dependent variable in this study is anxiety and depression, measured using the four-item Patient Health Questionnaire (PHQ-4). The UAS questionnaire asks, “Over the past two weeks, how often have you been bothered by any of the following problems?” The four items are: “feeling nervous, anxious, or on edge,” “not being able to stop or control worrying,” “feeling down, depressed, or hopeless,” and “little interest or pleasure in doing things” (Kroenke et al., 2009). Responses range from 0 (“Not at all”) to 3 (“Nearly every day”). The PHQ-4 composite score was calculated by summing the responses to the four items ranging from 0 to 12. In general population samples, a PHQ-4 score around 3 corresponds to approximately the 75th percentile of the distribution, and higher scores (e.g., ≥6) occur only in the upper percentiles of community samples (Löwe et al., 2010). The average PHQ-4 score among the observations was 1.67 (SD = 2.74), indicating that the average respondent in the survey exhibited a normal level of anxiety and depression.

2.2.2. Explanatory variable

This study uses the count of AAHIs occurring in the prior survey wave (t-1) as the explanatory variable. Using lagged AAHI counts ensures that all incidents temporally precede the measurement of respondents’ mental health outcomes. The FBI provides incident-level data on anti-Asian hate crimes, categorized by bias motivation codes that indicate anti-Asian intent. These data are aggregated by survey wave to calculate the number of AAHIs occurring during each wave. On average, individuals were potentially exposed to 17 AAHIs per survey wave during the study period (SD = 16) (see Table 2).

Table 2.

Summary statistics of the explanatory variable and moderators.

	%/Mean (S.D.)
Explanatory variable
Number of AAHIs (FBI) in t	16.99 (15.95)
Moderators
Distance (in miles) between AAHI and respondents if (>0)	335.05 (361.43)
Monthly COVID-19 death rate (per 1000 population)	0.31 (1.00)
Similarity of victim-respondent characteristics
Same gender	46.07
Respondent race/ethnicity
Non-Hispanic White	74.11
Non-Hispanic Black	7.56
Hispanic	9.82
Non-Hispanic Asian (same race/ethnicity)	3.59
Non-Hispanic Other	4.92
Offender characteristics
Racial/ethnic minorities	31.29
Victim characteristics
Male	76.39
Older adults (≥60)	53.30
Any injury	80.40
Republican vote shares	39.69 (20.47)

Note. Unweighted estimates. N = 6,552; N-wave: 152,276; 3-34 waves of the UAS COVID dataset. Standard deviations and group membership percentages are calculated across all respondent–wave observations, reflecting the panel structure of the data.

2.2.3. Moderators

To examine the heterogeneous effects of AAHIs on individuals' anxiety and depression in relation to contextual factors, this study selected moderators informed by Ecological Systems Theory. For the microsystem, this study utilized the physical proximity of AAHIs to respondents' residences, the similarity between victim and respondent characteristics, and the respondents' race/ethnicity. When multiple AAHIs occurred within a given survey wave, the variable was coded based on the closest distance between an AAHI and the respondent. If no AAHIs were reported, the variable was coded as zero. To address the non-normal distribution of the distance variable, this study applied a natural logarithm transformation. If any incidents occurred, the average shortest distance between an AAHI and respondents during a given survey wave was 335 miles (SD = 361) (See Table 2). A gender match indicator was used to measure the similarity between victim and respondent characteristics. If any AAHI involved a victim of the same gender as the respondent, the variable was coded as one; otherwise, it was coded as zero. Approximately 46% of observations involved exposure to AAHIs with victims of the same gender (see Table 2). Respondents’ race/ethnicity was operationalized using four indicator variables—non-Hispanic Black, Hispanic, non-Hispanic Asian, and non-Hispanic other—with non-Hispanic White serving as the reference group. As shown in Table 1, Table 2, approximately 3.6% of the analytic sample identifies as non-Hispanic Asian, the group directly targeted by AAHIs.

The mesosystem was represented by county-level monthly COVID-19 death rates retrieved from the Johns Hopkins CSSE repository. These data were merged with the individual-level UAS dataset based on respondents' county of residence and survey months. The average monthly county-level death rate per 1000 population among the observations was 0.31 (SD = 1.00) (see Table 2).

For exosystem factors, offender and victim characteristics were examined. Offender characteristics included an indicator coded as one if the offender was from a racial or ethnic minority group and zero otherwise. Approximately 31% of observations involved exposure to AAHIs with offenders from racial or ethnic minority groups during the study period (see Table 2). Victim characteristics included gender, age, and the presence of physical injuries caused by the incident. If any AAHI during a survey wave involved male victims, the victim's gender variable was coded as one; otherwise, it was coded as zero. For incidents involving older adult victims aged 60 or above, the victim's age variable was coded as one, and zero otherwise. If victims suffered physical injuries during an incident, the injury variable was coded as one; otherwise, it was coded as zero. During the study period, 76%, 53%, and 80% of observations involved exposure to AAHIs where the victims were male, older adults, or injured as a result of the incidents, respectively (see Table 2).

Finally, the macrosystem was represented by the county's political affiliation. Data on Republican vote shares from the 2020 Presidential election were retrieved from the MIT Election Data + Science Lab. The vote share was calculated by dividing the number of Republican votes by the total votes cast. On average, the observations resided in counties with 40% Republican vote shares in the 2020 Presidential election (SD = 20) (see Table 2).

2.2.4. Control variables

The control variables in this study included respondents' age, marital status, employment status, household income, renter status, COVID-19 diagnosis, health insurance ownership, number of household members, along with individual- and time-fixed effects. Individual-fixed effects accounted for all time-invariant characteristics, such as gender, race/ethnicity, and educational attainment, which were excluded from the empirical models. Time-fixed effects controlled for unobserved seasonal and temporal factors, including attitudes toward and perceptions of COVID-19 and hate incidents. Definitions of each control variable are presented in Table A2 of the Online Supplementary Materials.

2.3. Empirical models

To examine whether changes in anxiety and depression levels are associated with AAHI counts in the U.S., this study employed individual-fixed effects while controlling for time trends, as specified below:

$${AD}_{it}={\alpha }_1{AAHI}_{t-1}+{\alpha }_2{X}_{it}+i_i+w_w+{\epsilon }_{it}$$ (1)

In this model, ${AD}_{it}$ represents the PHQ-4 score, which measures the anxiety and depression levels for individual i during period t. ${AAHI}_{t-1}$ denotes the number of AAHIs that occurred during period t-1. $X_{it}$ is a vector of time-variant individual characteristics as described in Section 2.2.4. $i_i$ and $w_w$ represent individual- and wave-fixed effects, respectively, while ${\epsilon }_{it}$ is the idiosyncratic error term. The primary coefficient of interest in this model is ${\alpha }_1$.

To examine the heterogeneous effects of AAHIs on anxiety and depression levels based on factors informed by Ecological Systems Theory, the following individual-fixed effects model is estimated:

$${AD}_{it}={\beta }_1{AAHI}_{t-1}+{\beta }_2{M}_{ict}+{\beta }_3{AAHI}_{t-1}{M}_{ict}+{\beta }_4{X}_{it}+i_i+w_w+{ϵ}_{it}$$ (2)

In this model, $M_{ict}$ represents a moderator variable interacting with AAHI counts, informed by Ecological Systems Theory, including the microsystem, mesosystem, exosystem, and macrosystem. The coefficients of interest are ${\beta }_1$, ${\beta }_2$, and ${\beta }_3$.

Because AAHI exposure is shared within survey waves, observations may exhibit within-wave dependence due to common exposure and period-specific shocks. Accordingly, in all model specifications we report multi-way clustered standard errors, clustering by respondent to account for within-person correlation across repeated observations, by county of residence to account for geographic dependence, and by survey wave to account for shared exposure and wave-specific common shocks.

3. Results

3.1. The relationship between AAHIs and anxiety and depression levels

The individual-fixed effect estimates from Equation (1) are presented in Table 3. The analysis revealed a positive correlation between the lagged number of AAHIs and anxiety and depression levels. Specifically, a one-unit increase in AAHI counts was associated with a 0.0635-point increase in the PHQ-4 score (range 0-12). Given a sample mean of 1.67 (SD = 2.74), this corresponds to an increase of approximately 2.3% of one standard deviation, indicating a small but measurable within-person effect on depression and anxiety symptoms.

Table 3.

The relationship between the number of AAHI and anxiety and depression levels.

	DV= PHQ4 scores
	Coef. (S.E.)
Number of AAHI in t-1	0.0635∗∗∗ (0.0029)
R-squared	0.0217
N	6552
Obs.	152,276

Note. Unweighted individual-fixed effects regression estimators. 3-34 waves of UAS. PHQ-4 scores assess respondents' levels of anxiety and depression, with higher scores indicating greater levels of anxiety and depression. The study's sample includes respondents for whom county of residence information is available. Covariates include the respondent's age, marital status, employment status, household income, renter status, health insurance coverage, number of household members, diagnosis of COVID-19, and wave-fixed effects. Standard errors are multi-way clustered by respondent, county of residence, and survey wave. ∗∗∗p < 0.001.

3.2. Heterogeneity in the relationship between AAHIs and anxiety and depression levels

Table 4 presents the individual-fixed effects model estimates, incorporating interactions between AAHI counts and selected moderators from Equation (2). Panel A includes microsystem factors as moderators, Panel B incorporates a mesosystem factor, Panel C examines exosystem factors (i.e., offender and victim characteristics), and Panel D applies a macrosystem factor.

Table 4.

Vulnerability towards anxiety and depression associated with the number of AAHIs.

	Coef. (S.E.)		Coef. (S.E.)		Coef. (S.E.)		Coef. (S.E.)
Panel A: Microsystem factors
(1) Number of AAHI in t-1	0.0591∗∗∗ (0.0030)	(1) Number of AAHI in t-1	0.0624∗∗∗ (0.0029)	(1) Number of AAHI in t-1	0.0632∗∗∗ (0.0029)
(2) ln(Physical proximity to the AAHI location in miles)	0.0172∗ (0.0070)	(2) Gender similarity between victims and respondents	−0.0621∗∗ (0.0185)	(1) × Black	0.0014 (0.0013)
(1) × (2)	−0.0003 (0.0003)	(1) × (2)	0.0012∗ (0.0001)	(1) × Hispanic	0.0002 (0.0008)
				(1) × Asian	0.0039∗∗ (0.0013)
				(1) × Other	0.0004 (0.0013)
R-squared	0.0217	R-squared	0.0219	R-squared	0.0218
N	6552	N	6552	N	6552
Obs.	152,276	Obs.	152,276	Obs.	152,276
Panel B: Mesosystem factors
(1) Number of AAHI in t-1	0.0630∗∗∗ (0.0029)
(2) COVID-19 death rate per 1000 population	0.0097∗ (0.0038)
(1) × (2)	0.0012 (0.0013)
R-squared	0.0217
N	6552
Obs.	152,276
Panel C: Exosystem factors
Offender characteristics		Victim characteristics
(1) Number of AAHI in t-1	0.0635∗∗∗ (0.0029)	(1) Number of AAHI in t-1	0.0020 (0.0013)	Number of AAHI in t-1	0.0635∗∗∗ (0.0029)	Number of AAHI in t-1	0.0020 (0.0012)
(2) Racial/ethnic minority offenders (ref = other)	0.0642 (0.0589)	(2) Male victims (ref = female)	1.7983∗∗∗ (0.0688)	Older adult victims (ref <60)	0.0642 (0.0589)	Victims with physical injury (ref = other)	1.7983∗∗∗ (0.0688)
(1) × (2)	−0.0635∗∗∗ (0.0028)	(1) × (2)	−0.0443∗∗∗ (0.0019)	(1) × (2)	−0.0635∗∗∗ (0.0028)	(1) × (2)	−0.0443∗∗∗ (0.0019)
R-squared	0.0217	R-squared	0.0217	R-squared	0.0217	R-squared	0.0202
N	6552	N	6552	N	6552	N	6552
Obs.	152,276	Obs.	152,276	Obs.	152,276	Obs.	152,276
Panel D: Macrosystem factors
(1) Number of AAHI in t-1	0.0631∗∗∗ (0.0029)
(2) Republican vote shares in %	−0.0035 (0.0027)
(1) × (2)	−0.0000 (0.0000)
R-squared	0.0216
N	6552
Obs.	152,276

Note. Unweighted individual-fixed effects regression estimators. 3-34 waves of UAS. PHQ-4 scores assess respondents' levels of anxiety and depression, with higher scores indicating greater levels of anxiety and depression. Covariates include the respondent's age, marital status, employment status, household income, renter status, health insurance coverage, number of household members, a diagnosis of COVID-19, and wave-fixed effects. Standard errors are multi-way clustered by respondent, county of residence, and survey wave. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

Among microsystem factors (Panel A), microsystem-level salience—particularly gender similarity and shared racial identity—modestly shapes how exposure to AAHIs translates into psychological distress. Victim–respondent gender similarity modestly moderates the association between lagged AAHIs and PHQ-4 scores, with distinct level and frequency effects. Incidents involving gender-concordant victims are associated with slightly higher baseline levels of psychological distress, reflecting a small level effect. In contrast, the interaction term indicates that the marginal effect of additional AAHIs differs only minimally by gender concordance. Specifically, under gender concordance, the estimated marginal effect of an additional AAHI is approximately 0.0015 PHQ-4 points (=0.0624 − 0.0621 + 0.0012), compared with the overall effect of lagged AAHI exposure. Although this moderation effect is statistically detectable, its magnitude is extremely small—on the order of a few thousandths of a PHQ-4 point per additional AAHI—indicating that gender concordance primarily influences psychological distress through baseline salience of incidents rather than through cumulative effects of repeated exposure. This pattern is consistent with the interpretation that shared gender identity modestly heightens the symbolic relevance of hate incidents without substantially amplifying the impact of incident frequency.

Additional analyses incorporating respondents’ race/ethnicity indicate that the association between lagged AAHIs and PHQ-4 scores is modestly stronger for Asian respondents. Specifically, the interaction effect implies a total marginal effect of 0.0671 PHQ-4 points per additional AAHI (=0.0632 + 0.0039), representing a slightly larger—though still small—effect relative to non-Hispanic White respondents. In contrast, interaction terms for non-Asian racial/ethnic groups are small and statistically insignificant, indicating that the effect of AAHI exposure on PHQ-4 scores does not differ meaningfully from that observed for non-Hispanic White respondents.

Physical proximity to the AAHI location exhibits a small positive main effect on PHQ-4 scores; however, its interaction with AAHI frequency is not statistically significant. This indicates that geographic distance does not meaningfully moderate the association between lagged AAHI exposure and mental health outcomes.

The selected mesosystem factor (Panel B), the monthly county-level COVID-19 death rate exhibits a small additive association with anxiety and depression, while its interaction with AAHI exposure is not statistically significant. Lagged AAHI exposure remains positively associated with PHQ-4 scores, with each additional AAHI in the prior wave associated with an increase of 0.063 PHQ-4 points in the absence of COVID-19-related deaths in the community. Independently, higher COVID-19 mortality is associated with increased psychological distress: a one-unit increase in the monthly COVID-19 death rate (per 1000 population) corresponds to a 0.0097-point increase in PHQ-4 scores. Although the interaction term between AAHI counts and COVID-19 death rates is positive, it is not statistically significant, indicating that pandemic severity does not meaningfully moderate the association between AAHIs and mental health.

Panel C examines whether exosystem-level incident characteristics moderate the association between AAHI exposure and anxiety and depression symptoms. Regarding offender characteristics, the interaction between AAHI counts and whether incidents involved racial/ethnic minority offenders is statistically significant. Specifically, when AAHIs involve non-minority offenders (the reference category), additional incidents are associated with a 0.0635-point increase in PHQ-4 scores. In contrast, when incidents involve racial/ethnic minority offenders, the interaction term indicates that additional increases in AAHI frequency are not associated with further increases in psychological distress, as the marginal effect of additional incidents is attenuated (=0.0635 − 0.0635 ≈ 0). However, the positive main effect for minority offenders (0.0642) suggests that incidents involving racial/ethnic minority offenders are associated with higher baseline PHQ-4 scores, independent of incident frequency.

A distinct pattern emerges for victim characteristics. While the frequency of AAHIs is positively associated with PHQ-4 scores overall, incidents involving male victims, older victims, or victims with physical injuries are associated with substantially higher levels of psychological distress, largely through strong level effects rather than incremental effects of incident frequency. Specifically, the total effect of AAHI exposure on PHQ-4 scores is approximately 1.76 points when incidents involve male victims (=0.0020 + 1.7980-0.0443) and 1.76 points when incidents involve victims with physical injuries (=0.0020 + 1.7983-0.0443), reflecting large increases relative to the 0–12 scale. Incidents involving older victims are associated with a smaller but positive total effect of approximately 0.06 PHQ-4 points (=0.0635 + 0.0642-0.0635). Although the interaction terms indicate that the marginal effect of additional AAHIs is attenuated in these cases, the dominant influence on mental health arises from the presence of these victim characteristics themselves, rather than from changes in incident frequency.

Finally, this study finds no interaction effect between AAHI counts and macrosystem factors (Panel D), such as political affiliation, indicating that the effect of AAHI counts on anxiety and depression levels is consistent across political affiliations.

4. Discussions

The results support that higher frequencies of AAHI are linked to elevated levels of anxiety and depression among respondents. This pattern is consistent with prior evidence showing that anti-Asian discrimination during the COVID-19 pandemic elicited widespread negative emotional responses, such as distress, anxiety, depression, and avoidance across both Asian and non-Asian American populations (Hahm et al., 2021). Additional research demonstrates that the harmful effects of anti-Asian discrimination extend beyond individual encounters to broader social contexts: for instance, state-level hate crime rates have been significantly associated with greater depressive symptoms among Asian, Hispanic, non-Hispanic White, and non-Hispanic Black individuals (Harii et al., 2025). These findings underscore that AAHIs—whether pandemic-related or occurring in other contexts—pose substantial risks to psychological well-being across communities. Because AAHI counts were positively correlated with anxiety and depression, the present study further examined factors that moderated this association across the microsystem, mesosystem, and exosystem levels, consistent with Bronfenbrenner's (1974) Ecological Systems Theory.

At the microsystem level, we initially expected that respondents who shared the same gender as AAHI victims would exhibit stronger associations between incident frequency and psychological distress, drawing on intergroup empathy theory, which posits greater empathic responses toward ingroup members (Vanman, 2016). The results provide only limited support for this expectation. While victim–respondent gender concordance is associated with slightly higher baseline levels of psychological distress, the moderating effect on the frequency of AAHIs is extremely small. In other words, shared gender identity appears to influence the salience of hate incidents at the level of psychological response, but it does not meaningfully amplify the impact exposure to AAHIs on that response. This pattern suggests that gender concordance operates primarily through symbolic or interpretive mechanisms, rather than through cumulative stress processes. Social Role Theory (Eagly & Wood, 2012) offers a useful lens for understanding this distinction. Incidents involving gender-concordant victims may activate role-based expectations—such as protection or caregiving—that heighten immediate emotional responses, thereby elevating baseline distress. However, these role-based responses may not substantially alter how additional incidents accumulate to affect mental health over time.

We also found that the association between lagged AAHIs and psychological distress is modestly stronger for Asian respondents relative to non-Hispanic White respondents, although the magnitude of this difference is small. This pattern aligns with prior research indicating that members of directly targeted racial/ethnic groups may experience heightened psychological vulnerability when exposed to hate incidents or discrimination directed toward their communities (Ertorer, 2024; Hahm et al., 2021; Harii et al., 2025). Shared racial identity may intensify perceived threat, collective stigma, and anticipatory stress, thereby amplifying emotional responses to hate exposure. At the same time, the modest size of the differential effect suggests that the mental health consequences of AAHIs are not limited to directly targeted groups. Although Asian respondents demonstrate slightly greater vulnerability, the overall association between AAHI exposure and psychological distress is observed across the broader sample. These results underscore the importance of combining targeted outreach for directly affected communities with broader mental health support during periods of elevated hate incidents. In particular, culturally competent and trauma-informed screening, assessment, and intervention approaches for Asian American individuals may be especially important during periods of increased hate exposure, as such strategies can help address stigma, culturally specific stress responses, and barriers to care.

At the exosystem level, incident characteristics shape how exposure to AAHIs translates into psychological distress, though not always in the direction initially hypothesized. Contrary to expectations, the association between AAHI frequency and respondents’ symptoms of anxiety and depression is stronger when incidents involve non-minority offenders, whereas increases in incident frequency are less strongly associated with distress when incidents involve racial or ethnic minority offenders. This finding is consistent with Integrated Threat Theory (Stephan & Stephan, 2000), which emphasizes that threat perceptions depend on both the source and interpretation of intergroup encounters. Hate incidents perpetrated by members of marginalized racial or ethnic groups may be interpreted through complex social narratives that diffuse perceived intent or responsibility, whereas repeated incidents involving non-minority offenders—who occupy structurally dominant social positions—may reinforce perceptions of systemic vulnerability and social exclusion over time. Together, these dynamics help explain why offender characteristics shape not only the presence of psychological distress but also its responsiveness to repeated exposure.

Turning to victim characteristics, the findings indicate that AAHIs involving male victims or victims with physical injuries are associated with substantially higher levels of anxiety and depression, whereas age-related differences are more modest. Importantly, these patterns reflect strong level effects rather than stronger marginal effects of incident frequency. That is, incidents involving male or injured victims elevate psychological distress regardless of how frequently AAHIs occur, suggesting that the salience of these incidents drives distress more than cumulative exposure. Social Role Theory (Eagly & Wood, 2012) provides a useful lens for interpreting these findings. Acts of violence against men may violate normative expectations that men should be able to protect themselves and others, thereby heightening perceptions of social disorder and vulnerability. Such norm violations may trigger elevated distress even in response to a single salient incident. Age-related norms may operate differently. Although violence against older adults is often perceived as particularly troubling due to associations with frailty and vulnerability (North & Fiske, 2012), the present results suggest that AAHIs involving older victims do not substantially alter the relationship between incident frequency and psychological distress, indicating a more limited role for age-related moderation. Finally, incidents involving physical injury appear to be especially salient, producing large increases in distress independent of incident frequency. Visible evidence of harm may amplify fear and perceived threat to personal and community safety, reinforcing psychological responses to hate exposure (Gover et al., 2020). These findings underscore that victim characteristics shape mental health responses primarily through the severity and symbolic meaning of incidents, rather than through cumulative exposure alone.

4.1. Limitations of this study

This study has several limitations that should be acknowledged. First, the FBI hate crime database used to measure AAHIs is based on voluntary reporting and almost certainly underrepresents the true incidence of hate incidents. Importantly, this underreporting may not be random: prior research suggests that zero counts in FBI hate crime data may reflect non-reporting behavior rather than a true absence of incidents, and that political context is a strong predictor of reporting completeness (Mills et al., 2024). If underreporting is systematically correlated with other county-level characteristics, the estimated associations may be biased rather than merely attenuated. Consistent with our prior work comparing FBI hate crime data with media-based measures (Yoon et al., 2025), future research should triangulate AAHI exposure using multiple data sources to improve measurement validity. Additionally, we excluded cases lacking complete geographic information, which may have further underestimated exposure. Second, several key measures are proxies and may only partially capture the constructs of interest. Offender and victim characteristics recorded by law enforcement are subject to misclassification, and county-level Republican vote share represents a broad indicator of sociopolitical climate that cannot fully reflect lived experiences of community hostility or support. Finally, as an observational study, causal inferences cannot be made. While the use of individual and time fixed effects helps mitigate confounding, unmeasured time-varying factors—such as other stressors, changes in therapy or medication, salient news events, or local policies—could still influence mental health outcomes. Future research should integrate multiple sources of hate incident data, utilize finer-grained geographic indicators, incorporate richer measures of sociopolitical context, and apply designs capable of supporting stronger causal claims.

4.2. Implications for public health practices and policy

The findings of this study provide significant implications for public health practice and policy. They reaffirm that race-related discrimination, including AAHIs, is closely linked to psychological distress across racial and ethnic groups. Addressing this risk requires proactive preventive efforts. Public education initiatives, such as social media campaigns and community-based programs, should highlight the mental health consequences of discrimination to increase awareness of its societal costs. Disseminating theoretical frameworks such as Social Role Theory (Eagly & Wood, 2012) and Integrated Threat Theory (Stephan & Stephan, 2000) can further help the public understand why characteristics such as victim gender, victim age, offender group membership, and broader contextual stressors (e.g., COVID-19 death rates) may intensify psychological impact of race-based hate incidents. These perspectives may also empower individuals to better recognize and process their emotional responses.

In addition, community-level interventions such as bystander intervention training can equip individuals with strategies to safely address discrimination, harassment, and violence when they occur. Following violent or highly visible incidents, rapid, trauma-informed outreach should be activated through hotlines, mobile or pop-up counseling, and clear safety messaging. During local spikes, brief mental health screenings and warm handoffs across health systems, campuses, and community hubs can facilitate timely care. Investment in culturally and linguistically tailored services, in partnership with trusted AAPI organizations, is essential. When these practices are implemented routinely with stable funding, real-time monitoring, and clear objectives, communities can transition from ad hoc responses to a reliable, evidence-based system that reduces psychological distress while strengthening safety and trust.

In particular, given the contextual factors identified in this study as influencing mental health, intervention strategies should be developed with close attention to the conditions that may intensify the psychological impact of hate incidents. For example, gender similarity between victims and respondents may shape threat perception, disclosure, and access to support, intensifying the mental health impact of racially motivated violence. Additionally, targeted interventions, such as trauma-informed counseling and culturally tailored outreach for male and younger adult victims, along with coordinated medical and mental health services for victims with physical injuries, are necessary.

At the policy level, strong anti-hate crime legislation must be enacted and enforced to deter race-based violence, while funding should be directed toward community services that provide victims with crisis counseling, mental health support, and legal resources. Together, these efforts can mitigate the mental health burden of race-related discrimination and foster safer, more inclusive communities. Policy makers can use these findings to justify dedicated funding streams for hate-crime mental health response teams, mandatory data sharing between law enforcement and public health departments, and the integration of anti-hate education into school and workplace policies. Together, these measures can reduce the psychological burden of race-based violence and strengthen community resilience in future public health crises.

4.3. Conclusion

Altogether, our results show that county-level AAHIs are linked to elevated anxiety and depression symptoms in ways that are shaped by victim, offender, and contextual characteristics. Our findings underscore the importance of moving beyond incident-level responses to address the broader ecological factors that shape vulnerability to psychological harm. Public health strategies must integrate hate-incident surveillance with real-time mental health support, trauma-informed outreach, and culturally responsive services, especially in high-risk periods or communities experiencing concurrent stressors. Interventions tailored to vulnerable subgroups, coupled with stronger policy infrastructures and community partnerships, are essential for mitigating the mental health burden of AAHIs. Ultimately, reducing the psychological toll of race-based hate will require coordinated action across public health, social services, and policy environments to foster safer and more resilient communities.

CRediT authorship contribution statement

Su Hyun Shin: Writing – review & editing, Writing – original draft, Validation, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization. Young Ji Yoon: Writing – review & editing, Writing – original draft, Validation, Project administration, Funding acquisition, Data curation, Conceptualization. Junfei Lu: Writing – review & editing, Writing – original draft. Hee Yun Lee: Writing – review & editing, Writing – original draft, Validation, Supervision, Funding acquisition, Conceptualization.

Ethical statement

This research used de-identified secondary data from the Understanding America Study (UAS) COVID-19 panel, including restricted geographic identifiers obtained through USC Dornsife's approved data access procedures. UAS secures informed consent from all participants and operates under oversight of the University of Southern California Institutional Review Board. The hate crime data from the Federal Bureau of Investigation's Uniform Crime Reporting Program contain no personally identifiable information. In accordance with U.S. federal regulations for research involving human subjects, the present study was determined to be exempt from IRB review under 45 CFR 46.104(d)(4), as it involved analysis of existing, de-identified data. No additional ethical approval was required.

Declaration of competing interest

The authors declare no competing financial or personal interests.

Appendix A. Supplementary data

The following is the Supplementary data to this article:

Data availability

The data that has been used is confidential.