Microaggressions in the United States

“Microaggressions” is the term scholars and cultural commentators use to describe the ways that racism and other systems of oppression are upheld in everyday interactions. Although prior research has documented the types of microaggressions that individuals experience, we have lacked representative data on the prevalence of microaggressions in the general population. We introduce and evaluate five new survey items from the 2018 General Social Survey intended to capture five types of microaggressions. We assess the prevalence of each microaggression as well as a constructed microaggression scale across a key set of sociodemographic characteristics. We find that black Americans experience more microaggressions than other racialized groups, twice the rate of the general public for some types. Younger people report more microaggressions than older people. Women are more likely to report some types of microaggressions, and men others. Experiencing microaggressions is associated with an array of negative physical and mental health outcomes.

O VER the past several decades, social scientists have documented a shift in the expression of racism in everyday life: as overt racism has become less socially acceptable among whites, 1 scholars argue that more covert, though no less powerful, expressions of racism now predominate (Bobo et al. 2012;Bonilla-Silva 2013). Microaggressions, or everyday, subtle experiences of racism, have become a useful concept for studying these forms of racism, and scholars and public commentators alike have increasingly incorporated them into discussions about contemporary racism and racial inequality in America (Embrick, Domínguez, and Karsak 2017;Limbong 2020;Yoon 2020). Microaggressions help capture the ways that white people express and reinforce racial domination in everyday life.
A large body of research indicates that people of color regularly experience microaggressions, and some scholars have expanded the concept to include other types of oppression, including those based on gender, sexual orientation, nativity status, and disability status (Nadal 2011). Though these scholars have vividly documented the experience and harmful consequences of microaggressions, most data come from unrepresentative samples, often drawn from undergraduate university students. Data from unrepresentative samples leave open questions about the frequency and types of microaggressions experienced in the broader population.
In this article, we introduce and evaluate five new survey items from the 2018 General Social Survey (GSS) intended to capture five types of microaggressions identified by prior research: being given less respect than one deserves, receiving poorer service than others at restaurants or stores, being threatened or harassed, others acting as if you are not smart, and others acting as if they are afraid of you. We assess the prevalence of each microaggression as well as a constructed microaggression scale across a key set of sociodemographic characteristics. We found that black people experienced more microaggressions than other racialized groups, for some types at twice the rate of the general public. Younger people of all racialized groups experienced more microaggressions than older people did, suggesting a generational difference in Americans' expectations in public settings. We also found a complex story regarding gender: women and men differ in the types of microaggressions that they experience. Furthermore, microaggression is not just talk; it is aggression. We found that microaggressions were associated with an array of negative physical and mental health outcomes. Our analysis provides a first representative portrait of microaggression experience in the United States and raises important questions for future research on the topic.

Background
The concept of microaggressions was developed by black psychiatrist, doctor, and scholar Chester Pierce. Writing about the impact of racism in the everyday lives of black Americans, Pierce wrote, "Most offensive actions are not gross and crippling. They are subtle and stunning. The enormity of the complications they cause can be appreciated only when one considers that these subtle blows are delivered incessantly" (Pierce 1970:265-66). Pierce drew attention to the cumulative effect that "subtle" experiences of racism can have when experienced consistently as a part of everyday life. Psychologist Derald Wing Sue and colleagues' recent work has reinvigorated the study of microaggressions and sought to further classify different types of microaggressions. Sue (2010:3) defines microaggressions as "the everyday verbal, nonverbal, and environmental slights, snubs, or insults, whether intentional or unintentional, that communicate hostile, derogatory, or negative messages to target persons based solely upon their marginalized group membership." In contrast to aggressive acts like overt discriminatory treatment or racist statements, microaggressions take subtler (and more common) forms, such as a white student telling a black student that they were admitted to college only because of affirmative action. This focus on the subtlety of contemporary racism aligns with sociological notions of "everyday racism" (Essed 1991) and the linguistic form of color-blind racism, summarized succinctly by the subtitle of Bonilla-Silva's 2002 article "How to Talk Nasty about Blacks without Sounding 'Racist"' (Bonilla-Silva 2002).
There is by now a large body of scholarly work on microaggressions, coming largely from psychiatry, psychology, education, and, more recently, sociology. Researchers have assessed the forms of microaggressions experienced by many marginalized groups, including black Americans (Torres et al. 2010;Lamont et al. 2016;Keith et al. 2017), Latinx Americans (Huber and Cueva 2012;Ballinas 2017;Zambrana et al. 2017), Asian Americans (Sue et al. 2007a), Native Americans (Senter and Ling 2017), multiracial people (Nadal et al. 2013), women (Capodilupo et al. 2010), lesbian, gay, bisexual, transgender, and queer (LGBTQ) people Woodford et al. 2013), and persons with disabilities (Keller and Galgay 2010). This body of work has identified variations in both the context and content of the microaggressions different groups typically experience. For instance, whereas Sue et al. (2007b) identified invisibility as a key theme of microaggressions experienced by Asian Americans, Torres et al. (2010) identified hypervisibility compounded by an assumption of criminality experienced by black Americans. Microaggressions are also, as Peréz Huber and Solorzano (2015) argue, "layered assaults," or based on the intersection of race with other identities, including gender, class, sexuality, and immigration status. Scholars taking this intersectional approach have found that experiences of microaggressions are often shaped by multiple identities at once (Sue et al. 2007a;Nadal et al. 2014;Zambrana et al. 2017). For instance, McCabe (2009) found that Latina women were viewed as sexually available and exotic, but the same was not true for Latino men.
Microaggressions accumulate. Previous scholars have identified damaging psychological and physiological effects of microaggressions. People who experienced more microaggressions were more likely to be depressed and to have lower overall psychological well-being (Finch et al. 2000;Sellers et al. 2006;Torres et al. 2010). Smith, Allen, and Danley (2007), focusing on the cumulative effect of experiencing microaggressions as a part of everyday life, coined the condition of "racial battle fatigue." Microaggressions also harm physical health and have been linked to chronic stress, high blood pressure, and sleeping problems (Harrell, Hall, and Taliaferro 2003;Steffen and Bowden 2006;Pascoe and Smart Richman 2009). Microaggressions do harm because they are everyday manifestations of the broader system of racial domination that structures American society (Bonilla-Silva 2013; Peréz Huber and Solorzano 2015).
Scholarship on microaggressions typically draws on qualitative data from focus groups or in-depth interviews. These data are useful for understanding how individuals experience microaggressions and for discovering the forms that they take, but they are not representative and therefore cannot be used to estimate the prevalence of microaggressions (Wong et al. 2014). Nadal (2011) developed a set of 45 survey items to assess microaggressions called the Racial and Ethnic Microaggressions Scale (REMS), but, to our knowledge, the REMS has not been fielded on a representative sample. Representative data regarding experiences of microaggressions are needed to provide insight into everyday experiences of marginalized groups and for assessment across groups and individual characteristics. Representative data would also enable analyses that identify both individual and contextual factors that predict the occurrence of microaggressions.

Data and Methods
Our data come from a special supplement to the 2018 GSS. The supplement was proposed by Austin W. Ashe, Eduardo Bonilla-Silva, and Kenneth C. Land of Duke University and selected over other user-proposed modules by the GSS Board of Overseers. The Duke team proposed eight items; due to space and time limitations, the GSS Board, in consultation with the Duke team, trimmed the module to these five items: In your day-to-day life, how often have any of the following things happened to you? 2 1. You are treated with less respect than other people? sociological science | www.sociologicalscience.com 2. You are treated unfairly at restaurants or stores?
3. People act as if they think you are not smart?
4. People act as if they are afraid of you? 5. You are threatened or harassed?
The response categories were "almost every day," "at least once a week," "a few times a month," "a few times a year," "less than once a year," and "never." People had little trouble answering these questions; 1.2 percent of respondents could not provide an answer to "not smart," and less than one percent of respondents could not provide an answer to the other items.
To accommodate more content, the GSS puts most items on two of three ballots and randomly assigns cases to one of the three ballots. The microaggression questions appeared on ballots A and C of the 2018 GSS; 785 people completed ballot A, and 789 people completed ballot C. Of those 1,574 people, 1,539 gave a valid answer to all five microaggression questions, and between 1,555 and 1,564 answered each of them.
The questions were administered by professional interviewers; 89.4 percent of respondents were interviewed face-to-face, 10.4 percent were interviewed by telephone, and mode was not recorded for three cases. Interviews were conducted in either English or Spanish. 3 Of 4,376 contacts with households for the 2018 GSS, 53 (1.2 percent) were dropped due to language difficulties; that is, no one in the household spoke English or Spanish well enough to do the interview.
Each GSS draws a representative sample of U.S. households and randomly selects an English-or Spanish-speaking adult from the people living in that household. As initial response fell in the early 2000s, costs rose. To economize, since 2006, half the initial nonrespondents are dropped, and the other half pursued aggressively. In 2018, this approach yielded an American Association for Public Opinion Research response rate (RR5) of 59.5 percent, low by the standards of previous GSSs but high compared with other surveys. Dropping half the nonrespondents introduces some bias that GSS recommends users correct by using "nr" weights provided with the data set. We used weights for descriptive statistics below, but we did not weight cases for the regression analyses.
Race and gender are crucial to our analysis below. Based on theory, we would also have liked to assess sexual orientation but did not have enough cases for reliable generalizations. 4 Reflecting its roots in the 1970s, the GSS usually classifies people as male or female, according to the interviewer's observation of the respondent during the interview, and we use that classification for our analysis. 5 For classification into racialized groups, the GSS follows the practices of the U.S. Census Bureau. Respondents were asked two questions: their Hispanic heritage, if any, and their racial identity from categories specified by the federal Office of Management and Budget. GSS reports up to three racial responses for each case. We classified cases who reported any Hispanic origin in response to the first question as Latinx, and we classified the remaining cases by the GSS race variable, which codes cases as non-Latinx black if any of the racial responses included black, Non-Latinx white if the first racial response was white and neither of the other two was black, and "other" for any other combination of responses.
We recoded the GSS's educational attainment and degree questions into five categories: no credentials, high school diploma, some college, college degree, and advanced degree. The GSS records each person's exact age but truncates the ages it reports at 89 to avoid disclosing personal identities; we treat age as a continuous variable. To measure political orientation, we recode the GSS question: "We hear a lot of talk these days about liberals and conservatives. I'm going to show you a seven-point scale on which the political views that people might hold are arranged from extremely liberal-point 1-to extremely conservative-point 7" to three categories: liberal (1-3), moderate (4), and conservative (5-7).
To assess the consequences of microaggressions for physical and mental health, we used four variables. Three items-quality of life, mental health rating, and physical health rating-were measured on a five-point scale with 1 being "excellent" and 5 being "poor." We measured quality of life using the following question: "In general, would you say your quality of life is excellent, very good, good, fair, or poor?" We assessed physical health using "In general, how would you rate your physical health?" We assessed mental health using a similar measure: "In general, how would you rate your mental health, including your mood and your ability to think?" Lastly, we assessed occurrence of emotional problems: "In the past seven days, how often have you been bothered by emotional problems such as feeling anxious, depressed or irritable? Never, rarely, sometimes, often, or always?"

Prevalence of Microaggressions
We begin our analysis by examining the prevalence of each type of microaggression in the adult population. Table 1 reports the weighted distribution for each of the five microaggressions.
Most Americans never or rarely experience microaggressions, as we see in the top panel of Table 1, where the modal reply to four of the five items was "never." The exception was disrespect. Eight percent of Americans experienced being disrespected daily, and another 12 percent were disrespected weekly. Six percent of Americans were treated as if they were not smart daily, and another six percent experienced that weekly. One in 10 Americans receives poor service monthly, and five percent weekly. The majority of Americans never experience being threatened or harassed (62 percent) or never perceive that others are afraid of them (55 percent).
Microaggression theory addresses how racism is experienced. Black people are expected to experience more microaggressions than whites, though, of course, white women, whites who are LGBTQ, and whites with disabilities may experience nonracial microaggressions. Because much prior research has focused on black Americans, we assess their microaggression reporting here in detail before assessing other groups below. Black people experience significantly more frequent microaggressions than other Americans experience. Twenty-eight percent of black people experienced disrespect weekly or more, compared with 20 percent of all Americans (top panel). One in 10 black Americans reported others being afraid Note: Data weighted for nonresponse.
of them weekly or more, compared with seven percent of all Americans. For all other microaggressions, black Americans reported occurrences of weekly or more at approximately twice the rate of the full population. For instance, whereas 11 percent of black Americans reported poor service on at least a weekly basis, only six percent of the general public did so. Whereas 13 percent of all Americans reported being treated as not smart at least weekly, almost one in four black Americans did so (23 percent). For all microaggressions, black respondents report "never" experiencing the microaggressions at lower rates than the general public.

Microaggression Scale Construction
In short, microaggressions vary from relatively frequent disrespect to mercifully rare threats. Theory expects each of them to enforce a common status group hierarchy, implying that they will have high interitem reliability. Cronbach's alpha is the classic measure of interitem reliability; a five-item scale should have an alpha of at least 0.7 (on a scale from zero to one). The alpha for the five microaggression items is 0.76. It would decrease if any of the five items were to be removed from the scale. This implies that we can think of exposure to the risk of microaggression as a single concept, realized in individuals' lives in different ways. A principle components factor analysis further confirms that inference; the first eigenvalue was 1.88, and the second eignenvalue was a trivial -0.05. After assessing the overall distribution of each item, we next create a summary scale measure by summing responses to all five microaggressions questions. Each microaggression was coded such that 0 = "Never" and 5 = "Almost every day," giving the scale a possible range of 0 to 25 with higher scores indicating more frequent experience of microaggressions. The actual range of the scale in the data is  Figure 1 shows the distribution of the scale. The modal scale score is 0 (10 percent of sample), the mean is 6.50, and the distribution is right-skewed. Figure 2 shows the results of regressing the microaggression scale on each microaggression item by racialized group. For all racialized groups except the "all others," the "afraid of" item has a shallower slope than the other items, indicating it is less related to the other four items than they are to each other.

Microaggression Prevalence across Demographic Groups
We next assess the reported frequencies for the five microaggression items as well as the scale across five sociodemographic characteristics: racial identity, gender, educational attainment, age, and political orientation. See Table 2. We pay particular attention to racial identity and gender because previous research indicates that microaggressions are regular features of life for people of color and women. In our discussion, we focus on the scale score differences across groups but note where patterns for individual items differ.
Among racialized groups, black respondents report experiencing microaggressions most frequently (scale mean = 7.9). Latinx respondents have the second highest scale mean, 6.5, though this is only slightly higher than white respondents'  mean of 6.4. Respondents in the racial other category, which includes Asian Americans, Native Americans, and multiracial individuals who were not Latinx, reported microaggressions least frequently (mean = 5.1). Four of the five separate items follow the same racial pattern as the scale; "threaten" was the exception. Latinx respondents were threatened or harassed most often, followed by white respondents, black respondents, and finally racial others. Being threatened was rare among all groups.
In contrast to findings from some qualitative research on microaggressions (Capodilupo et al. 2010), men and women scored very similarly on the microaggression scale, with men (mean = 6.6) scoring slightly higher on average than women (mean = 6.5). Unlike for race, the individual items do not all follow the scale pattern. Women more often reported receiving poor service (16 percent vs. 14 percent) and being treated as not smart (24 percent vs. 21 percent), though these differences are small. Though a rare event, women were twice as likely as men to report being threatened or harassed (six percent vs. three percent). Men, on the other hand, more often reported others being afraid of them (19 percent vs. 11 percent). Men and women were equally likely to report being disrespected (both 35 percent). Small gender differences on experiential items like this are a common outcome in survey research. For example, in a 2016 GSS module, men and women were equally likely to report bullying (authors' calculations). Sexual harassment is an exception; women were roughly three times more likely to experience it. In the 2018 GSS, women who were sexually harassed reported being threatened with the same frequency as women who were not sexually harassed (again, our calculations).
Education was not much of a factor in microaggressions. The range of means across educational categories was not monotonic, and it was less than half the range across racialized groups. Given how education correlates with both race and age, we will not explore the educational differences further until we can statistically control for other factors. Note: Data weighted for nonresponse.
Microaggression scale scores decreased consistently across age groups, from a mean of 7.8 among 18-to 29-year-olds to a mean of 2.6 among people aged at least 80 years. The separate items also declined with age, with slight variations.
Finally, assessing scale scores across political orientation, we find small acrossgroup differences with moderates (mean = 6.7) scoring slightly higher than liberals (mean = 6.6) and conservatives scoring lowest (mean = 6.3). For each individual item, moderates reported more frequent experiences of microaggressions, though conservatives reported receiving poor service and others being afraid of them more often than liberals, in contrast to the pattern seen for the scale. All of the political differences were quite small.

Multivariable Model of Microaggressions
Next, we include all sociodemographic characteristics in one multivariable linear regression model to assess their conditional associations with the microaggression Notes: Standard errors in parentheses. Higher political orientation values correspond to more conservative identification. * p < 0.05.
scale. As shown in Table 3, when all characteristics are included in the model, only racial identity and age are significant. Black respondents report significantly more frequent microaggressions than white respondents. However, the magnitude of the slope is small: controlling for other characteristics, black respondents on average score 1.1 points higher on the microaggression scale. Latinx and racial other respondents do not differ significantly from whites. An increase in one year of age is associated with a 0.07 decline in the microaggression scale score, controlling for other characteristics. Much to our surprise and inconsistent with previous literature, women do not significantly differ from men in this model.

Relationship between Microaggressions and Health Outcomes
Prior research identifies negative effects of microaggressions on both emotional wellbeing and, through stress mechanisms, physical health. To complete our analysis, we examine associations between the microaggression scale and an array of mental and physical health outcomes included in the GSS. This analysis is not meant to be a definitive assessment of the effects of microaggressions but instead a starting point to encourage further research. We examine four outcomes: quality of life rating, mental health rating, frequency of emotional problems, and physical health rating. Answer choices for all four items are on a five-point scale (described in more detail above). The rating items are reverse coded such that 5 indicates an "excellent" rating. We use ordinary least squares regression to predict each outcome using the microaggression scale, race, gender, age, educational attainment, and political orientation. Figure 3 presents the predicted outcomes by frequency of microaggressions (see Table 1 in the online supplement for full model results). More frequent microaggression experience is associated with lower quality of life, worse mental health, more emotional problems, and worse physical health. The strongest association appears for emotional problems: An increase of one standard deviation in the microaggression scale score is associated with a 0.2 standard deviation increase in frequency of emotional problems.

Discussion
Microaggressions were identified in clinical settings decades ago. Scholars explored their forms and theorized about their antecedents and consequences. Commen-tators expressed doubts of various sorts. Until now, the nationally representative data needed to estimate prevalence correctly was lacking. The 2018 GSS included five questions about microaggressions, ranging from the relatively prevalent experience of being disrespected to slightly less common experiences of being ignored or passed over in public settings or being treated as "not smart" to the thankfully rare experiences of being feared or threatened. One in five Americans was disrespected at least weekly; six in 10 Americans were never threatened. However, because microaggressions serve to support power hierarchies, those who experienced one type of microaggression were more likely to experience others. Thus, we constructed a microaggression scale to capture Americans' overall experience of microaggressions.
Microaggressions are everyday manifestations of systems of oppression, and we assessed their prevalence across various demographic groups. Black Americans experienced microaggressions most frequently, followed by Latinx Americans, then whites, then those of other racial identities. Native Americans, Asian Americans, and multiracial Americans-the groups combined in the "other" racial category due to small sample sizes-reporting fewer microaggressions than whites is unexpected given previous qualitative findings (Sue et al. 2007a;Nadal et al. 2013;Senter and Ling 2017). This discrepancy could be read as suggesting that previous qualitative samples were biased toward higher reporting of microaggressions, or-and this is where we fall-it suggests that the GSS items do not capture the microaggressions that these groups experience well. For instance, Sue et al. (2007a) document that Asian Americans experience invisibility in social settings, and perhaps items such as "disrespect" do not capture their experience. This speculation raises a larger point: the content and forms of microaggressions experienced by different groups vary widely, and these items may not capture the forms experienced by some groups. Given our empirical findings and the strong interitem associations of reliability, these items may capture microaggressions experienced by black Americans better than those experienced by other groups of color.
To our surprise, but consistent with some previous work on stigmatization experiences (Lamont et al. 2016), there are not large gender differences in microaggression scale scores; however, assessment of the individual items suggests that this is because the scale consists of microaggressions more likely to be targeted at women (poor service, treated as not smart, being threatened), at men (others being scared of you), and at both genders equally (disrespect). Gender differences across most items, though small, comport with claims that microaggressions are often gendered in their forms and expression (McCabe 2009;Capodilupo et al. 2010). More research is needed to investigate the gendered nature of microaggressions.
Microaggression reporting decreases with age. This strong age gradient deserves more study. Our hunch is that younger people interpret the microaggressions questions differently than their elders do. Younger people are probably more aware of concepts like microaggressions, disposing them to identify with the experiences the questions describe. Microaggressions were experienced pretty equally across educational categories and across the political spectrum.
Microaggression theories focus on the cumulative effects of microaggressions. The consistency of small events was hypothesized to build up and harm those who experience them. Our results are consistent with these claims: higher frequencies of microaggressions were associated with worse quality of life, physical health, and mental health and more emotional problems. The association was largest for emotional problems. Of course, these data can only establish associations, not cause-effect relationships. But the associations do reinforce the construct validity of the microaggressions measures.
The 2018 GSS provided a first foray into the field of microaggressions in the general population. The five items scaled well and were reliable. Our empirical findings suggest that the scale measures the concept well for some groups-particularly black Americans-but that more research is necessary to capture how oppression is experienced in everyday encounters for other groups, particularly Asian Americans, Latinx Americans, and Native Americans. We were surprised to see that younger people experienced more microaggressions than older people. More research into both the particularities of the five items in the GSS and how expectations differ by age is needed to understand this pattern. Furthermore, the presence of gender differences on individual items but not in the overall scale score requires more investigation to understand how gender shapes content and experiences of microaggressions.