‘People of Color’ as a category and identity in the United States

ABSTRACT

Although the category ‘people of color’ has not been adopted by the U. S. Census, it has assumed critical importance in public and social-scientific understanding of population change. But do Americans understand it as a neutral demographic category or a political identity? We provide the first U.S. national estimates for self-identification as a person of color (PoC ID). We also measure dissonance between individuals’ self-identification and the reflected appraisals of others, as well as where Americans draw the boundaries around PoC. PoC ID ranges from 95% among Blacks to 61% among Asian Americans and Pacific Islanders and 45% among Hispanics. Liberalism significantly increases PoC ID among AAPI, while conservatism reduces PoC ID among Hispanics, but political ideology is unassociated with PoC ID among Blacks. Stronger ethnic group consciousness raises PoC ID among all three groups. Demographic factors (particularly gender, age, and nativity) also influence PoC ID and PoC ID dissonance, but effects vary sharply in direction and significance from one group to another. Our estimate that only 27% of American adults self-identify as PoC suggests caution in assuming that growing populations of Hispanics and Asians will necessarily lead to a national majority who see themselves as people of color.

KEYWORDS:

• People of color
• race
• identity
• panethnicity
• multiracial

Reporting results of the 2020 census, the Washington Post declared that ‘the under-18 population is now majority people of color’ (Bahrampour and Mellnik 2021). The Census Bureau had not used the term ‘people of color’, much less defined it, but the Post’s usage reflected the widespread journalistic practice of using the term as an umbrella category for nonwhite minorities. The Post was also not unusual in counting all Hispanics as PoC even though many self-identify on the Census as white. It also assumed that all children with just one non-Anglo parent identify as PoC.

In these and several other respects, the boundaries of the category ‘people of color’ are ambiguous. Does it include, for example, all of the growing number of Americans who mark both ‘white’ and another race, the multiracial category that the Census calls ‘whites in combination’? Does it extend to those with Middle Eastern and North African (MENA) origins whom the Census continued in 2020 to classify as white? Depending on definitions, the PoC population may be larger or smaller than conventional assumptions suggest. Members of groups officially classified as white may think of themselves as PoC, while others may not self-identify as PoC despite being routinely categorised that way. If the self-defined PoC population is larger than conventional classifications indicate, the United States may be closer to a PoC majority, at least by self-identification, than most estimates project. If, on the other hand, the self-defined PoC population is smaller than conventional classifications indicate, the United States may be further from that point – indeed, further away than Richard Alba (2021) suggests in The Great Demographic Illusion.

The uncertainties about who is and who is not a person of color arise basically for two reasons: blurred boundaries to the category and mismatches between attributions of identity and self-identifications. The boundaries are especially blurry for individuals in panethnic categories such as Hispanic and MENA that span color lines and for multiracial people with white ancestry. Self-identifications may also depart from conventional notions of PoC for social and political reasons. The term ‘people of color’ has long been associated chiefly with Black Americans, but some members of other ethnoracial minorities may not see themselves as sharing an identity with them. ‘People of color’ has also been part of the vocabulary of progressive movements in the United States and taken to have political implications, which those with more conservative beliefs may reject. As an identity, ‘people of color’ therefore does not necessarily map on to ‘nonwhite’. It may carry connotations about the equivalence of minority groups that not all members of those groups share. Black leaders and journalists have been among the critics: During the 2020 Black Lives Matter protests, several prominent figures repudiated the term ‘people of color’ on the grounds that Blacks face distinctive threats such as police violence (Code Switch 2020; Edwards and McKinney 2020).

Lack of clarity and agreement about the meaning and use of PoC is a concern not just because of its popular and political currency. The term presents a dilemma for the language of social science. ‘People of color’ has become a routine category in the social sciences despite its uncertain boundaries. For example, in the JSTOR database, the number of articles in social science journals using the terms ‘people of color’, ‘person of color’, or ‘persons of color’ rose from the single digits in the 1960s to 45 in the 1970s, 188 in the 1980s, 1,095 in the 1990s and 1,504 in the 2000s.¹

We designed a U.S. national survey to clarify some of the perplexities about ‘people of color’ that previous research has not addressed. Besides gathering basic descriptive data, the survey includes items designed to test the extent to which political and demographic variables influence self-identification and the categorisation of others as PoC.² Our findings show large and systematic divergences between the conventional use of the term ‘people of color’ and individual self-identifications. Political variables have stronger effects in some groups than in others. Dissonance between self-identification as PoC and beliefs about others’ perceptions is especially high among Hispanics and whites in combination. These results help illuminate how ethnoracial identities and boundaries are evolving along the U.S. people-of-color line.

People of color as a social category and identity

The distinction between social categories, on the one hand, and groups and identities, on the other, is fundamental to the study of race and ethnicity. Racial and ethnic categories may be widely used in a society (for example, in official statistics) but not correspond to ‘experienced group boundaries or social identities’ (Loveman 1997 ). Emotional attachments to a category are likely to develop when ‘members of an ethnic category self-identify and are identified by others as “belonging” to a “group” with little ambiguity, when they share easy-to-identify cultural repertoires of thinking and acting, and when they are tied together by strong alliances in day-to-day politics’ (Brubaker 2004 , 46–47). Uncertainties about the match between categories and identities are especially likely to arise when populations change, political alliances shift, and new categories and identities emerge.

Since the eighteenth century, the social category ‘people of color’ has been a prime example of the historical variability of racial and ethnic categories and identities. The term has arisen twice from the mixing of peoples in the United States (Starr 2022). From the late 1700s to the mid-1800s, ‘people of color’ – a translation of the French gens de couleur – referred primarily to an intermediate group in slave societies: free, mostly mixed-race people of African, white, and (in some contexts) indigenous descent. The category was especially important in Louisiana and South Carolina, whose laws, like those in the French Caribbean, recognised ‘people of color’ as a distinct, third social stratum between enslaved Blacks and free whites. This was not the same as ‘mulatto’, as most mulattoes were enslaved, and some free people of color were solely African-descended. Since many free people of color were educated and owned property, the term had positive associations. In the 1820s, northern Black leaders and journalists adopted ‘people of color’ as a collective name for Blacks, whether of exclusively African or mixed descent. During the antebellum period, ‘people of color’ also sometimes served, particularly in the North, as an umbrella category for all nonwhites. But as a collective name for Blacks, ‘people of color’ began to give way to ‘colored’ even before the Civil War, and with the end of slavery ‘people of color’ virtually disappeared as a social category of any kind (except in Louisiana), until the term was revived in an expanded, umbrella usage beginning in the 1970s. The revival came largely through the Black, progressive, and feminist movements as they sought to build interracial coalitions that included growing numbers of Hispanics and Asian Americans as well as the indigenous. Within those movements, ‘people of color’ signified people who suffer racial oppression, but as the term has spread into mainstream journalism and social science, it has become a more neutral synonym for ‘nonwhites’, though it is still typically shunned or mocked on the right (Starr 2022).

The formation of higher-level, panethnic categories and identities has been a repeated pattern in the United States, from the early development of the categories ‘Black’ and ‘Indian’ to the emergence of ‘Hispanic’ and ‘Asian American’ in the 1970s. Rather than simply being an external ideological imposition, such categories also arise from structural conditions, such as segregation in work, residence, and marriage, and from panethnic movements and collective action. Panethnicity, as Okamoto (2014) defines it, is a ‘process through which multiple ethnic groups relax and widen their boundaries to forge a new, broader grouping and identity’. The adoption of panethnic terms typically stirs political conflict. Community activists and politicians often use panethnic terms in building alliances, but panethnicity tends to arouse tension with subgroups (Okamoto and Mora 2014). At the time that ‘Hispanic’ was introduced in the 1970s, Mexican Americans, Puerto Ricans, and Cubans had little connection with one another; Cubans particularly resisted identifying themselves as ‘Hispanic’ (Mora 2014). The diffusion of ‘Asian American’ as a category and identity had to overcome longstanding tensions among the Chinese, Japanese, and Korean communities, stemming in part from hostilities among their countries of origin. ‘Asian American’ – introduced in the late 1960s by university students as an alternative to earlier terms such as ‘Oriental’ – illustrates another aspect of panethnicity: Social movements often seek to use a new name free of derogatory associations to give a more positive meaning to a panethnic identity.

Unlike ‘Hispanic’ and ‘Asian American’, ‘people of color’ has not been adopted as an official governmental classification. But as a kind of super-panethnicity, it too involves a widening of ethnic group boundaries to forge, in Okamora’s terms, ‘a new, broader grouping and identity’. As an alternative to the terms ‘nonwhites’ or ‘minorities’, which imply whites as a reference point, ‘people of color’ is also a positive renaming, an example of what Wimmer (2008) calls ‘normative inversion’. Whether or not it is officially adopted, it has already become central to both public and social-science conceptions of the changing U.S, population.

Higher-level categories and identities do not necessarily replace lower-level ones. Since the 1970s, surveys have found that growing percentages of Hispanics and Asian Americans accept panethnic identities, but when offered a choice, most prefer to identify with their national-origin or ethnic group – for example, choosing ‘Mexican’ over ‘Hispanic’ or ‘Latino’ (Mora 2014, 34; see also Lu 2020). Similarly, instead of replacing panethnic identities, ‘people of color’ is a superordinate identity at a still higher level, which individuals may integrate into their self-understanding and relation to others. Some, however, may not see themselves as PoC, if only because it is one further step removed from their primary ethnic or racial identities and because of the extreme heterogeneity among all the groups that PoC embraces.

Rather than viewing race and ethnicity as externally fixed, contemporary work in the social sciences recognises a role for choice and agency at both the individual and group levels (Roth 2018). Even at the individual level, however, racial identity is not solely a matter of psychological introspection. Individuals may integrate into their self-image how others view them – the ‘reflected appraisal’ of others (Khanna 2004; see also López 2014). But since individuals may also reject how others (including the state) view them, they may confront ‘perceptual dissonance’, and that dissonance may determine whether a racial identification is stable or fluid.³ Similarly, the assignment of whole groups to a category may be subject to conflicting beliefs about ethnoracial boundaries. For example, in an analysis of survey data, Lee and Ramakrishnan (2020) find that contrary to official classifications and the views of South Asians themselves, other Asians as well as non-Asians tend not to think of South Asians as ‘Asian Americans’. Concerned that they are marginalised among Asian Americans, ‘told that they are not “Asian enough”’ and ‘stereotyped as inferior or uncivilized’, some Filipino Americans, Southeast Asian Americans, and South Asians have organised as ‘Brown Asians’ to demand recognition and representation (Nadal 2020).

Who counts as PoC is even more ambiguous than who counts as Asian American because neither the government nor non-governmental associations authoritatively define the category ‘people of color’. As a result, the questions of who self-identifies as PoC and who counts as PoC in the eyes of others are genuinely wide open.

Literature review

Two bodies of research are directly relevant to questions about PoC as a category and identity: research explicitly concerned with individual self-identification as a person of color and studies of felt political commonalities across racial and ethnic groups.

Research on people-of-color identity

The most substantial research on self-identification as a person of color comes from Pérez (2021), who argues that a distinct identity (‘PoC ID’) has emerged that is separate from identification with a particular racial or ethnic group: ‘As a cross-cutting form of identity’, Pérez (2021, 31) writes, ‘PoC ID re-categorizes minority outgroups under a common banner of identity’ so the members of each group see the interests of the others as their own. Although Pérez (2021, 27) claims that each racial identity is ‘comfortably nested’ under PoC ID, his own data are in tension with that interpretation in two respects.

First, Pérez’s findings indicate that identifying as a person of color is a politically inflected choice, associated with liberal views. The self-identified people of color he interviewed ‘let me know that if you are a racial minority and you practice conservative politics, you are considered a persona non grata by other PoC’ (Pérez 2021, 50). Yet while acknowledging this and other bases of exclusion such as ‘acting white’, Pérez (2021, 180) holds that PoC ID is an identity that ‘all U.S. racial and ethnic minorities display to a degree’ (emphasis in original). In a 2018 online nonprobability survey of Blacks, Hispanics, and Asians, Pérez used four agree/disagree statements to construct a scale for stronger or weaker PoC ID: ‘The fact that I am a person of color is an important part of my identity’; ‘Being a person of color is a major part of how I see myself’; ‘I often think about the fact that I am a person of color’; ‘I am glad to be a person of color’ (Pérez 2021, 139). These items, however, seem to prime respondents to assume ‘the fact’ that they are people of color.

Second, despite the framing of these questions, the data from Pérez’s survey nonetheless suggest a marked difference in PoC ID between Blacks, on the one hand, and Hispanics and Asians on the other. For example, 90% of Blacks ‘somewhat agreed’, ‘agreed’, or ‘strongly agreed’ that they are ‘glad to be a person of color’, but the comparable percentages were only 60% among Asians and 58% among Latinos (Pérez 2021, 42). According to other data Pérez (2021, 157–158) presents, members of all three groups agree that Blacks are the ‘most prototypical’ people of color – that is, best exemplifying the category – whereas Asians are the least prototypical. His interviews indicated that a darker skin tone is associated with higher PoC ID, though a light tone is not an absolute bar to it.

But does PoC ID vary only by degree among U.S. racial and ethnic minorities? A survey of Hispanics conducted by two liberal political pollsters, Joshua Ullibari and Celinda Lake, found that when offered two other alternatives, only 25% chose to describe Hispanics as ‘people of color’, who, ‘like African Americans, remain distinctive over generations’. The two alternative self-descriptions were as ‘white ethnics’, who ‘like European Americans, over generations become part of the American mainstream’ (32%) and as ‘bootstrappers’, who ‘over generations can get ahead through hard work’ (28%) (Haney López and Gavito 2020; Project Juntos 2020). It is difficult to know what to make of these results since the survey’s framing of choices suggests that identifying as ‘people of color’ is less consistent than the two alternatives with aspirations of social mobility.

Two studies of self-identification as a ‘woman of color’ indicate substantial variations by ethnicity and race. In an initial online survey, Matos, Greene, and Sanbonmatsu (2021) found 92% of Black women and 64% of Hispanic women self-identify as WOC. The same authors then used data from the larger 2020 Collaborative Multiracial Post-Election Survey and put self-identification as WOC at 91% among Black women and 58% among Asian American but only at 36% among Hispanic women despite excluding any respondent who listed white as ‘one of their identities’ (Greene, Matos, and Sanbonmatsu 2022).

Although this body of research is suggestive, the results are inconclusive. No national study, to our knowledge, analyses how PoC ID varies among both men and women in the population at large.

Research on felt political commonalties

Another relevant body of research concerns ‘felt political commonalities’ among ethnoracial groups. Survey researchers have measured political commonalities in various ways, asking, for example, about feelings of ‘closeness’ to other groups or about how much one group has ‘in common’ politically with others. A felt political commonality is not the same as a shared ethnic or racial identity; whites, after all, may feel political commonalities with Blacks. But the reverse of felt commonalities – feelings of distance or hostility – may well undermine the tendency of minorities to identify with one another as people of color. As shown in some social psychological research (González and Brown 2003 ), the forging of a superordinate identity that bridges groups can help foster feelings of commonality. ‘People of color’ is such a bridging identity for Blacks, Hispanics, Asian Americans, and others. The same factors that predict stronger or weaker felt political commonalities among them may also hypothetically predict variations in PoC ID.

Much of the work on political commonalities has been sparked by the troubled history of Black-Hispanic and Black-Asian relations and an interest in the potential for cross-racial coalitions. Arguing that ‘perceived commonality between blacks and Latinos is integral to constructing political associations’, Kaufmann (2003, 199) sought to explain why there was ‘little evidence of formal or even informal coalitions between the nation’s two largest minority groups’. Using data from a 1999 survey, she pointed to an asymmetry: While 75% of Blacks felt they had a ‘fair amount’ or ‘a lot’ in common with Hispanics, only 33% of Hispanics reciprocated those feelings. But she also found several factors that predicted Hispanic-Black commonality. These included ‘pan-Latino affinity’ (Hispanics who saw themselves as sharing more in common with other Hispanics were also more likely to see themselves as sharing commonalities with Blacks) and acculturation (those born in the United States and with greater English proficiency were more likely to see commonalities with Blacks). Several later studies have supported these findings on Hispanic group consciousness (Sanchez 2008; Jones-Correa, Wallace, and Zepeda-Millán 2016 ) and on the differences in feelings of commonality with Blacks between foreign- and U.S.-born Hispanics (Sanchez 2008; Wilkinson 2014, 910–912).

Research on Black-Asian commonalities has explored the same issues (Nicholson, Carter, and Restar 2020). In 2008, 2012, and 2016, the National Asian American Survey – conducted in English and as many as ten Asian languages – asked, ‘Thinking about government services, political power and representation, would you say Asian Americans have a lot in common, some, little in common, or nothing at all in common’ with each of the other major ethnoracial groups. Ramakrishnan (2014) reports that in the 2008 and 2012 NAAS surveys, the proportion of Asian American respondents who said they had ‘a lot’ or ‘some’ in common with another racial group was just 33% with Blacks and only 37% with Hispanics, compared to 43% with whites. But in the same surveys, Asian Americans registered higher levels of support than did whites for social and economic policies benefiting other communities of color. In other words, Asian Americans had more in common on policy with Blacks and Hispanics than the measure of felt political commonality suggested.

Moreover, from 1992 to 2012, according to presidential election exit polls, Asian Americans moved 40 points toward the Democrats, another sign of growing political convergence with Blacks and Hispanics (Ramakrishnan 2016). Strikingly, the 2016 NAAS then found that, compared to four years earlier, the share of Asian Americans perceiving a lot or some in common with Hispanics rose 13 percentage points, and the share seeing a lot or some in common with Blacks rose 17 points (Arora, Sadhwani, and Shah 2020). This sequence suggests that earlier scholars may have gotten the causality backward. Kaufmann (2003, 201) had written that ‘until blacks and Latinos see their respective groups as having much in common, they are unlikely to engage in sustainable political coalitions’. But the data on Asian Americans suggest that felt political commonalities may at least in part result from coalitions (voting for the same party) rather than serving as a precondition for political alliance.

Research on both Hispanics and Asian Americans has also found significant subgroup variations in felt political commonalities with Blacks. Among Hispanics, feelings of commonality with Blacks have been higher among Puerto Ricans and Dominicans than among Mexicans and Cubans (Kaufmann 2003). Among Asians, feelings of commonality with Blacks have been higher among Bangladeshi, Pakistani, and Hmong respondents than among Chinese and Japanese (Arora, Sadhwani, and Shah 2020) – a finding that corresponds to the distinction between ‘Brown’ Asians and East Asians. Furthermore, just as U.S.-born Hispanics tend to see greater commonality with Blacks than foreign-born Hispanics do, so U.S.-born Asian Americans tend to see greater commonality with both Hispanics and Blacks than their foreign-born coethnic counterparts do (Arora, Sadhwani, and Shah 2020).

The research on cross-racial commonalities is directly relevant to PoC ID because much of the impetus for use of the term ‘people of color’ beginning in the 1970s came from coalition-building efforts seeking to unite Blacks, Hispanics, and Asians in progressive movements (Yuen 1997; Starr 2022). For example, the term ‘women of color’ first gained national attention in 1977, when a delegation to the National Women’s Conference in Houston that had originally called for a ‘Black Women’s Agenda’ secured support from other minority women for an expanded resolution on the ‘double discrimination’ faced by ‘women of color’, a new phrase to most Americans at the time (Perlstein 2020, 183; Ross 2011). Identifying under the same superordinate category as PoC was closely related to finding cross-racial commonalities.

Theoretical expectations

Previous theoretical and empirical work suggests how we should expect ‘people of color’ to vary as an identity and category. The dimensions of individual identity include self-identification as PoC (PoC ID), beliefs about others’ perceptions of oneself (reflected PoC ID), and whether these two aspects of identity coincide or conflict (PoC ID dissonance). Individuals may also vary in the extent to which they see their fate as being tied to that of other PoC (PoC linked fate). At the collective level, the key issue concerns boundary definitions and whether members of different groups categorise themselves and other groups as PoC (PoC categorisation). These dimensions of PoC ID and PoC boundaries may be expected to vary across major ethnoracial groups as well as subgroups. We focus on the impact of demographic and political variables.

Demographic variation

Given the historic association of the term ‘people of color’ with Blacks and continued perceptions of Blacks as being the most prototypical PoC, we expect that PoC ID, reflected PoC ID, and PoC linked fate will be highest among Blacks and that Blacks will be nearly universally categorised by others as PoC. The evidence on other groups is conflicting. On the one hand, Pérez’s (2021) data about prototypicality suggest that Hispanics should be next highest in PoC ID, reflected PoC ID, and PoC linked fate and that, of the three major nonwhite groups, Asian Americans should be lowest and least likely to be categorised as PoC. On the other hand, the WoC research suggests that Asian Americans may be higher than Hispanics in PoC ID (Greene, Matos, and Sanbonmatsu 2022). PoC ID dissonance should vary inversely with PoC ID.

Within groups, we expect variation according to national origins and ethnicity; in particular, among Hispanics, we expect lower PoC ID among Cubans than among Puerto Ricans and Mexicans. In addition, the association of higher education with liberalism suggests that education may predict stronger PoC ID. Since the term ‘people of color’ has a distinct meaning in the United States that may be unfamiliar to first-generation immigrants or attractive to those who see little or no political commonality with Blacks, foreign-born immigrants are likely to be associated with lower PoC ID among both Hispanics and Asians.

Political variation

Given the political inflection of the term ‘people of color’, we expect that within each racial group, liberal political views and voting Democratic in the 2020 US presidential election will correlate positively with PoC ID. In line with the findings on felt political commonalities, we also expect stronger panethnic group consciousness to be associated with higher PoC ID.

Data and methods

Data were collected by the Princeton University Survey Research Center through an online, nonprobability survey (the Princeton PoC ID survey) conducted between 10 August and 15 September 2021.⁴ The Center engaged Bovitz, Inc. to gain access to its U.S. national panel of survey respondents, originally aiming for 1,000 white, 500 Black, 500 Hispanic, and 500 Asian usable responses. We exceeded our targets for three of the four strata, but fell short of our goal for Asian Americans, even after extending the field period by several weeks. From the initial set of responses, we identified and dropped several test cases, responses with too much missing data, and two duplicate responses. Sample members who completed the survey questionnaire in less than two minutes (the median completion time was five minutes seven seconds) were tagged as speeders. After eliminating the speeders, the total number of complete, usable interviews came to 2,630. Defined by their racial and ethnic self-identification, the sample comprises 1,062 whites (non-Hispanic whites alone), 593 Blacks (non-Hispanic Blacks alone and in combination), 569 Hispanics (all those giving a positive response to the Census-style question on Hispanic ethnicity), 372 Asians and Pacific Islanders (non-Hispanic, non-Black AAPI alone and in combination), 31 American Indians and Alaska Natives (non-Hispanic, non-Black AINA alone and in combination), and 3 others.

Although the Princeton PoC ID survey is based on a sample from an opt-in national panel, it is fit for our purpose, which is chiefly to analyse factors related to PoC ID and its covariates across and within ethnic (Hispanic/non-Hispanic) and racial groups (Vehovar, Toepoel, and Steinmetz 2016; Dutwin and Buskirk 2017). The distribution of respondents by age, gender, region, and education in the unweighted data matches up closely with national distributions for each race/ethnic stratum in the sample.

The sample for the survey was stratified in a way that would allow us to make parameter estimates for the U.S. population as a whole as well as estimates for separate ethnoracial subgroups. As a result, we use two weight variables: Weight₁ (rakedwght1 in the dataset) accounts for the stratified design based on race and Hispanic ethnicity. The target population around which this weight is constructed is the U.S. adult (age 18+) non-institutional population. Weight₁ is used only once below (in Table 1 where we include an estimate of PoC ID for the U.S. population as a whole). Weight₂ treats each ethnoracial subgroup as if it were a representative sample of each group’s population. This weight is used when comparing survey responses across ethnoracial groups. Where we estimate the proportion within each major racial-ethnic group, we use weight₂.

Table 1. Self-identification as a Person of Color (PoC ID)^a.

	Do you consider yourself a person of color? (Yes)	Weighted n
Full weighted U.S. sample	27%	2,630
Race and ethnicity (by self-ID)^b
Hispanic
Hispanic (white alone)	28%	324
Hispanic (black alone)	93%	11
Hispanic (other nonwhite or multiracial)	65%	235
Hispanic (any race)	45%	569
Black Non-Hispanic
Black NH Alone	96%	521
Black NH in combination	89%	72
Black NH alone and in combination	95%	593
AAPI Non-Hispanic
AAPI NH Alone	64%	319
AAPI NH in combination	48%	68
AAPI NH alone and in combination	61%	387
White
White NH alone	3%	1,065
White NH in combination	51%	105
White NH alone and in combination	7%	1,170
White inclusive (Hispanic and non-Hispanic alone and in combination)	13%	1,535

^a Estimated proportion for full weighted sample is based on total population weight (weight₁) for all cases in the sample. All other estimated proportions in the table are based on sub-group weights (weight₂).

^b Sample members who self-identify with more than one race appear in more than one category in the table.

Supplemental data from the panel provider and imputation are used for a small number of cases where there are missing data for any variables to be used for weighting. To generate the two weights for the data analysis, we use the 2020 5-year ACS file (Ruggles et al. 2022) to create target proportions for gender, age, Hispanic ethnicity, race, education, and geographic division. In spite of these weight adjustments, there remain three sources of potential bias in the Princeton PoC ID survey that also affect many recruited surveys: (1) underrepresentation of nonvoters; (2) underrepresentation of 2020 Trump voters; and (3) underrepresentation of foreign-born immigrants with low English proficiency. To address the first two potential sources of bias, we incorporate a weight adjustment based on whether and how the sample member voted in the 2020 election for U.S. president; these weights are based on U.S. Census estimates of turnout rates (Fabina 2021; Fabina and Scherer 2022; Ramakrishnan and Sadhwani 2021) and Pew Research estimates of presidential voting (Igielnik, Keeter, and Hartig 2021). We discuss the potential problem of underrepresented foreign-born immigrants in qualifying our results.

Dependent variables

The primary dependent variable of interest is self-identification as a person of color (PoC ID), measured by a simple yes/no response to the question, ‘Do you consider yourself a person of color?’ We measure reflected PoC ID by the question, ‘Do you think other people see you as a person of color?’ which has four possible responses: ‘1. Yes, I think others see me as a person of color; 2. No, I don’t think others see me as a person of color; 3. It depends on circumstances and how I present myself; 4. I’m not sure’.

The combination of PoC ID and reflected PoC ID yields the second dependent variable of interest, PoC ID dissonance, the extent to which individuals believe others’ perceptions match or conflict with their own self-conception. PoC dissonance may come in two types. Type 1 dissonance exists when respondents say they consider themselves a person of color, but they think that others do not or might not see them that way. Type 2 dissonance exists when respondents say they do not consider themselves a person of color, but they think that others do or might.

These two questions about identification as PoC come before other questions about racial or ethnic identity to avoid priming the respondents to think about those categories in answering the questions about PoC ID. After standard Census-style questions about ethnic and racial identity, an additional item serves as a measure of PoC group identification: ‘Do you think that what happens generally to people of color in this country will affect what happens in your life?’ (PoC linked fate).

A fourth dependent variable of interest concerns the categorisation of members of different minorities as PoC. In a later part of the survey, respondents receive the following prompt: ‘Let’s say you work for an organization where a meeting has been called for employees who are people of color. On a 1–7 scale (1 = not welcome; 7 = very welcome), please indicate whether you think the following people would be welcome at the meeting’. The putative employees listed have the following characteristics: Black, Puerto Rican identifying as Black, Cuban American identifying as white, Chinese, Pakistani, Lebanese Christian, Egyptian Muslim, American Indian (e.g. Navajo), and an employee claiming American Indian ancestry but not a registered member of a tribe. The final item in this question asks: ‘And what about you? Would you expect to be welcome at a meeting that has been called for people of color?’

Independent variables

We measure race two ways. First, we use the Census-style, two-question format. The first question asks, ‘Are you of Hispanic, Latino/a, or Spanish origin?’ and then gives the same options as the Census; the second question asks, ‘What is your race or origin?’ and then lists the standard Census categories (condensing, however, the smaller Asian American and Pacific Islander sub-groups). Obtaining responses to these questions allows us to weight our sample according to ACS estimates and to compare our results to the many studies that have used these questions.

Near the end of the survey, we also ask respondents about ‘the race or origin’ of their grandparents. Here the survey employs a single-question format that the Census used in its National Content Test in 2015. Asking about ancestry this way has several advantages. It enables us to identify the many respondents with mixed-race ancestry who may not check off more than one race on the standard Census self-identification question. It provides an alternative measure of Hispanic descent and allows us to identify respondents with MENA ancestry. We also use the single-question format in asking about the ‘race or origin’ of ‘the person you're most involved with – your spouse, domestic partner, or a friend with whom you have a close relationship’. We treat racial and panethnic group consciousness as a political variable, measured by asking about ‘how important’ the respondents’ self-identified race or ethnicity is to their ‘personal identity’. Questions on 2020 presidential voting and political ideology (liberal/conservative) serve as measures of political views.

A final pair of questions asks for textual responses from respondents who identify in seemingly anomalous ways. The survey invites respondents who identify as Hispanic, Black, AAPI, American Indian, or multiracial but not as PoC to explain their thinking. Conversely, if respondents identify as white but do consider themselves PoC, the survey also invites them to explain why they do.

Analytical plan

We begin by presenting descriptive statistics by self-identified ethnicity and race on the primary outcomes of interest: PoC ID, PoC ID dissonance and PoC linked fate, and PoC categorizations. We then focus our bivariate and multivariate analyses on Blacks, Hispanics, and AAPI to identify the factors that drive PoC ID within each of those groups.

Results

Descriptive statistics by self-identified ethnicity and race

Table 1 shows variation in PoC ID by self-identified ethnicity and race. We estimate that 27% of the U.S. adult population self-identifies as PoC. As expected, nearly all non-Hispanic Blacks (95% of all Blacks alone and in combination) do so, while nearly all non-Hispanic whites alone (97%) do not. Self-identification as PoC proves to be lower among Hispanics (45%) than among non-Hispanic AAPI (61%). PoC ID varies among Hispanics from 28% among those identifying as white to 65% of those who identify as neither white nor Black (often writing in a Latin American origin term) to 93% among those identifying as Black (only 2% of the Hispanic population). Due to the small number of respondents self-identifying as American Indian or Native Alaskan, we do not show data on them in this section, though we discuss them later.

Table 2 presents the data on PoC ID dissonance, collapsed into four mutually exclusive ethnoracial categories: all Hispanics (regardless of race); non-Hispanic Blacks whether alone or in combination; non-Hispanic, non-Black AAPI alone and in combination; and non-Hispanic whites alone. As expected, PoC ID dissonance is low among both Whites alone and Blacks. But two out of five Hispanics and one out of three AAPI show some degree of dissonance. As more detailed analysis shows,⁵ the most common response among those who report type 1 dissonance (that is, respondents who consider themselves PoC but say others do not or might not), is that ‘it depends on circumstances and how I present myself’, which suggests they see themselves as having a degree of agency. (This response is not as common among the Hispanics or Blacks with type 2 dissonance.)

Table 2. Consonant and dissonant identity as a Person of Color (PoC)^a.

	Consonant PoC (self + other, yes)	Consonant NOT PoC (neither by self nor by others)	PoC Dissonance: Type 1	PoC Dissonance: Type 2	Unweighted n
Do you consider yourself a person of color?	Yes	No	Yes	No
Do you think other people see you as a person of color?	Yes	No	No/It depends/Not sure	Yes/It depends/Not sure
Self-identified race/ethnicity
Hispanic (any race)	27%	33%	17%	23%	567
Black Non-Hispanic	83%	2%	12%	3%	593
Asian PI Non-Hispanic	46%	22%	14%	18%	371
White Non-Hispanic	1%	87%	1%	10%	1,061
Other Non-Hispanic	1%	46%	35%	18%	34

^a n = 4 cases with missing data. Some tabulations do not add up to corresponding numbers in Table 1 because of rounding error, missing cases, or exclusive assignment of cases to a single category in this table.

The measure of linked fate follows the same relative order as PoC ID: 77% of non-Hispanic Blacks, 66% of non-Hispanic AAPI, and 50% of Hispanics agree that ‘what happens generally to people of color’ will affect their lives.⁶ Notably, though, Black PoC-linked fate is 18 points lower than Black PoC ID, while Hispanic PoC-linked fate is 5 points higher than Hispanic PoC ID – a compression of differences that is consistent with other research on linked fate (Gay, Hochschild, and White 2016).

Table 3 presents the data on how welcome respondents believe individuals of different racial, ethnic, and religious backgrounds would be to a workplace meeting called for persons of color. All groups agree that Blacks would be the most welcome at such a meeting; they also tend to agree that the Black Puerto Rican and Navajo employees would be the next most welcome. Regarding other employees (the white Cuban, Asian, MENA, and American Indian without tribal citizenship), the pattern is more mixed, though the ratings tend to fall in the same range.⁷ In general, Black respondents are the least likely to see others as welcome at a meeting for PoC, whereas AAPI respondents are the most likely to view members of other non-Black race-ethnic groups as welcome. The contrast in how the Black and AAPI respondents view the Chinese and Pakistani employees is particularly striking. AAPI respondents have a less restrictive understanding of PoC (notably including themselves), while Blacks tend to have a more restrictive understanding of the other groups’ PoC ‘credentials’.

Table 3. Categorisation as ‘People of Color’^a.

Putative employee	Mean Rating by Race/ethnicity
	Non-Hispanic Black	Hispanic (any race)	Non-Hispanic AAPI	Non-Hispanic White
Black (on −3 to +3 scale)	1.45	1.59	1.76	2.22
Puerto Rican (Black)	.96	1.25	1.54	1.86
Cuban (white)	.50	.71	1.15	.74
Chinese	.60	.75	1.49	.98
Pakistani	.30	.62	1.25	1.03
Lebanese Christian	.42	.57	1.31	.87
Egyptian Muslim	.45	.56	1.21	.94
Amer Indian (Navajo)	.76	.82	1.61	1.27
Amer Indian w/o tribal citizenship	.44	.47	1.12	.41
Yourself	1.63	1.08	1.47	-.88

^a Prompt: ‘Let’s say you work for an organization where a meeting has been called for employees who are people of color. On a 1–7 scale (1 = not welcome; 7 = very welcome], please indicate whether you think the following people would be welcome at the meeting’. Scale of 1–7 in questionnaire recoded to −3 to +3 for results displayed in the table.

Bivariate and multivariate analysis

We turn now to more detailed analysis of PoC identification and categorisation within and across the major racial groups. The bivariate proportions in Table 4 show some unanticipated patterns, particularly for gender and age. Women are less likely to identify as PoC among Hispanics, but more likely to do so among Blacks. Younger Hispanics and AAPI are more likely to identify as PoC, but younger Blacks are less likely. Higher education raises PoC ID among AAPI but appears to have little effect among Hispanics or Blacks. Political ideology and 2020 voting are strongly correlated with PoC ID among AAPI, but ideology makes hardly any difference among Blacks and has a seemingly anomalous effect among Hispanics, with both liberals and conservatives lower than moderates on PoC ID.

Table 4. Self-identification as Person of Color (PoC ID): weighted proportions for race/ethnic groups.

	Hispanic, any race (n = 569)	Non-Hispanic Black (n = 593)	Non-Hispanic AAPI (n = 372)
Total	44.6%	95.1%	60.2%
Male	52.3%	92.9%	59.8%
Female	36.9%	97.0%	60.6%
Age
18–34	53.0%	90.5%	64.3%
35–64	44.6%	97.0%	58.1%
65+	13.8%	100.0%	58.3%
Education
Less than BA	45.2%	94.6%	54.3%
BA or more	41.5%	96.9%	66.3%
2020 Vote
Nonvoter	44.3%	91.1%	47.8%
Biden	50.5%	98.6%	76.6%
Trump	35.6%	83.2%	52.1%
Political Orientation
Liberal (1–2)	35.7%	95.7%	81.1%
Moderate (3–5)	49.1%	94.9%	58.1%
Conservative (6–7)	34.0%	94.3%	32.5%
Nativity
Native born	47.8%	95.8%	62.2%
Foreign born	30.0%	83.6%	58.4%

Separate logistic regressions for PoC ID among Hispanics, AAPI, and Blacks (Table 5) test whether these bivariate correlations hold up. Following Mansournia et al. (2021), we use robust standard errors for estimating the odds ratios. In each case, we first show a model with basic demographic variables and then introduce political ideology and panethnic group consciousness. Although we show the results for Blacks, it must be remembered that 95% of Blacks self-identify as PoC, so there is not much non-identification to explain.

Table 5. Logistic regressions for predictors of self-identification as a Person of Color (Poc ID).

	Odds ratios
	Hispanic/Latino/a		Asian American		Black
Variable	(1)	(2)	(1)	(2)	(1)	(2)
Gender (Female)	.478*	.421**	.956	.756	3.18^†	3.53^†
Age	.886	.851**	.952	.936	1.37***	1.42***
Foreign born (1st generation)	.269**	.313**	.894	1.24	.283	.185^†
Cuban	.244^†	.517
Puerto Rican	1.58	1.81
White	.161***	.199***
Some other race	.587	.816
Partner diff race (white)	.532	.510†	1.21	1.53
Partner diff race (non-white)	.840	.632	1.79	2.20†	.792	1.48
Education	1.32*	1.40**	1.21^†	1.28*	1.26	1.34
Importance of being Latino		1.49***
Importance of being Asian				1.32†
Importance of being Black						2.53***
Importance of being biracial						2.11*
Liberal		.552		3.49**		.746
Conservative		.275**		.443		1.64
Constant	4.63**	1.37	.826	.203^†	2.24	.057*
N	569	564	372	362	586	586
Wald χ^2	47.2***	67.2***	13.4	18.4***	23.9***	30.7***
Pseudo R^2	.20	.23	.06^†	.11*	.16	.31

^† p < .10; *p ≤ .05; **p ≤ .01; ***p ≤ .001.

Demographic variation

Not only does PoC ID vary across the minority groups (Table 1); the factors that drive these differences also vary as well (Table 5). Several demographic variables are significant in the models for Hispanics. In both Models 1 and 2, three characteristics (being female, being born outside the US, and identifying as white) significantly reduce the odds for Hispanics of identifying as PoC, while one factor (education) increases it. In Model 2, the negative effect of age also becomes significant, with every increase in age by five years reducing the odds of identifying as PoC by 15%. Among all these variables, self-identification as white stands out as both having an extremely strong negative effect and being highly significant. But for AAPI, the pattern is different. The basic demographic variables are not significant, except for education, which increases PoC ID. Among Blacks, gender and age work the reverse way they do for Hispanics, though foreign birth also reduces the likelihood of PoC ID.

Based on previous work by Vasquez-Tokos (2017) showing an impact of interracial marriage on racial identity, we had anticipated race of partner might be associated with PoC ID. Indeed, among Hispanics, having a partner who is white and non-Hispanic reduces the odds of identifying as PoC, while among the AAPI, having a partner who is neither white nor AAPI increases the likelihood of identifying as PoC.

Contrary to expectations, national sub-group variations among Hispanics did not prove statistically significant (being Cuban was significant in Model 1 but not in Model 2). National sub-group variations also did not prove statistically significant among AAPI (not shown).

Political variation

Model 2 brings out varying levels of significance for political ideology (base: moderate) and ethnic group consciousness. Political ideology is significant for both Hispanics and AAPI, though the effect shows up in different ways. Among Hispanics, conservatism reduces PoC ID, while the effect of liberalism is not significant, whereas among AAPI, liberalism is positively associated with PoC ID but conservatism is not significant. In contrast, among Blacks, political ideology does not have a significant effect on PoC ID. Group consciousness (how important being Hispanic or Latino, Asian or Pacific Islander, or Black is to the respondent’s personal identity) raises PoC ID among all three groups. The effect is especially strong among Blacks and highly significant among both Blacks and Hispanics, though less so within the AAPI sample.⁸

A note on multiracial ancestry

Besides classifying respondents on the basis of their racial self-identification, we were also able to classify them on the basis of the race or origin they gave for their grandparents, according to the single-question, alternative Census format. This allows us to identify a larger number of whites in combination, including those with mixed Hispanic-Anglo as well as mixed Black–white and AAPI-white grandparents. We are also able to identify respondents with Native American and MENA ancestries, most of which are mixed.

We did not find, however, that introducing a measure of multiracial ancestry adds to the explanatory power of the logistic regressions, even though descriptive statistics indicate an association between the number of Anglo grandparents and PoC ID. For example, among those with at least one white (Anglo) grandparent, the proportion saying they identify as POC rises from 2.1% among those with no non-Anglo grandparents to 10.5% among those with one non-Anglo grandparent, 32.4% among those with two, and 50.1% among those with three (for the full table, see https://www.personofcolorsurvey.org/). But other variables in the logistic regressions already appear to capture these effects.

Grouping together all those with mixed American Indian/Native Alaskan and white ancestry (n = 57), 12.1% self-identify as PoC. If instead we limit this mixed group to those who self-identify as both American Indian/Native Alaskan and white (n = 40), the proportion self-identifying as PoC is 28 percent, compared to 64 percent among those who self-identify as Native American alone. The number of respondents with any MENA ancestry is only 22, seven of whom self-identify as PoC. More targeted surveys will be necessary to estimate PoC ID reliably among both the Native American and MENA populations.

Discussion

Several limitations of the survey suggest caution about some of the results. As mentioned earlier, the foreign born are underrepresented in the Princeton PoC ID survey. They account for 18% of Hispanics in our sample, compared to 47% in the US adult Hispanic population.⁹ Our data show a significant reduction in PoC ID among foreign-born Hispanics; if we had weighted our sample to compensate for their underrepresentation, PoC ID among Hispanics would have fallen from 45% to 35%.

Among AAPI in our survey, 42% were foreign born, compared to 77% in the U.S. adult AAPI population.¹⁰ Our regressions did not indicate a statistically significant effect on PoC ID from foreign birth among AAPI, but a more complete representation of foreign-born AAPI, particularly of low-English-proficiency immigrants, might have produced different results. In addition, the size of the AAPI sample may have limited our ability to detect differences among national-origins groups that larger surveys have found in felt political commonalities with Blacks and Hispanics. The Brown Asian American Movement (Nadal 2020) might increase PoC ID among those of Filipino, Southeast Asian, and South Asian descent.

The survey’s results, particularly for AAPI, may also have been affected by timing. Anti-Asian violence during the COVID-19 pandemic may have heightened Asian concerns about racial bias and raised PoC ID. Other studies of voting patterns and felt political commonalities, however, had already found a trend among Asian Americans in recent decades toward closer affiliation with Blacks and Hispanics. The strong positive effects on PoC ID of education, political ideology, and panethnic group consciousness among AAPI are consistent with earlier research on political commonalities. It is striking that the results from the American Experiences with Discrimination Survey (Lee and Sheng 2023) coincide so closely with ours on the proportion of Asian Americans who consider themselves people of color (63 percent compared to our estimate of 61 percent). Lee and Sheng (2023) also find a strong effect of political ideology and political affiliation among Asian Americans.

Although we did not anticipate that PoC ID would be lower among Hispanics than among AAPI, the results are not entirely surprising. The earlier survey by Ullibari and Lake (Project Juntos 2020) suggested Hispanic self-identification as PoC might be only 25%, whereas the survey reported here puts it at 45%. Most of the factors driving down PoC ID – racial identification as white, political conservatism, foreign birth, and age – are consistent with expectations based on earlier research. But the lower rate of PoC ID among Hispanic women relative to men was unexpected and needs further analysis.

We get some insight into the reasons why a majority of Hispanics did not self-identify as PoC from the textual answers by the Hispanics who did not identify as PoC and accepted our invitation to explain their thinking (259 out of 273 did so). Color consciousness appears to be critical. As Table 6 shows, in the most common response, 42.5% said they see themselves as white, light-skinned, or Caucasian (often adding that others see them as white), while 7.7% referred to their European ancestry (a few who cited both their physical appearance and ancestry are counted under both categories). Nearly all of the non-PoC-identifying AAPI respondents (69 out of 78) also accepted the invitation to explain their reasoning, but relatively few of them cited physical appearance or ancestry. Their most common explanations were an insistence on color-blindness (26.1%) or a belief that the term PoC applied only to Blacks (20.3%).

Table 6. Explanations for not self-identifying as PoC.

Reason	Hispanic		AAPI
	Proportion^a	Example	Proportion^b	Example
See self as or is seen as white/light skinned/ Caucasian	42.5%	‘I am not a person of color. I am light skinned’.	8.7%	‘Because my color is fair’
Have European or other white ancestry	7.7%	‘my parents are white caucasians europeans born in south America’	2.9%	‘I am Filipino and Spaniard but raised in the United States’
PoC a term for Blacks only	10.4%	‘Because I'm latino and the term person of color is used to describe black people’	20.3%	‘when you hear people of color you think of a black person not an asian’
Identify as Hispanic/mixed	6.9%	‘I’m not white but not black neither, I’m mix’
Identify as Asian			10.1%	‘I’m lao Thai Chinese and Vietnamese’
Reject PoC term; color-blind, etc.	18.5%	‘I don't look at the color of someone skin’	26.1%	‘i believe in all come from same God’
Identify only as American	4.2%	‘Because I am an American citizen’	11.6%	‘I'm just an ordinary guy that was born in the U.S.’
Unsure, unclear, other	13.5%		20.3%

Note: Totals sum to more than 100% because some respondents offered more than one reason.

^a Denominator for Hispanics is the 259 respondents who gave textual explanations (out of 273 non-PoC identifiers).

^b Denominator for Asians is the 69 respondents who gave textual explanations (out of 78 non-PoC identifiers).

A few angry responses from Trump voters had a more political character. A male Hispanic Trump voter in his late thirties wrote ‘fuck all this ‘identifing’ [sic] pc bullshit’. A male Japanese over age 70 wrote that ‘people of color only want a free ride and do nothing to better themselves’. A third Trump voter, a Hispanic male is his late twenties who identified as an American, declared that the very term PoC ‘in of itself is racist’. Conversely, a white-alone-identifying woman in her late twenties who also identified as PoC said she did so ‘because I support the black movement’. But overt political language in the textual answers was relatively rare, perhaps held back by social desirability bias.

Some of the respondents who identified as both white and PoC responded to the open-ended textual question with ambivalent, personal explanations. A Hispanic woman in her fifties, with a white father and Puerto Rican mother, wrote:

… nobody can figure out what the heck I am. I look Middle Eastern. Because of my racial ambiguity I have first-hand experienced racial profiling but at the same time I am in white passing enough that I enjoy many benefits of white privilege as well. Depends how dark I look at the time and how I am dressed. Definitely treated different when I am out in public with my white husband. Treated better than when I am alone. So I don’t feel like a white person … 

This kind of dissonant and contextual identity, with its political overtones (‘racial profiling’), is inherently difficult for surveys to capture. But rather than erasing these complexities in search of black-and-white distinctions, sociologists and demographers ought to incorporate them into their work.

Conclusion

Although ‘people of color’ has become part of Americans’ everyday vocabulary, its boundaries are ill-defined, and its use runs the risk of imputing an identity to people who do not accept it and inflating perceptions of shared identity across ethnoracial lines. Social scientists should recognise that, in using the term, they are using a category that more than half of Hispanics and nearly two out of five Asian Americans do not see as part of their identity, at least partly for political reasons. The results of this survey suggest caution in assuming that growing populations of Hispanics and Asians will necessarily lead to a national majority who self-identify as people of color. If we add together Census figures for Hispanics and all nonwhites and designate them all ‘people of color’, the U.S. population 18 years of age and over is now 39.2 percent PoC. In contrast, our estimate of those who self-identify as PoC is only 27 percent. U.S. demographic practice is to count people according to their self-identifications. If that practice were applied to the category ‘people of color’, the United States would be understood to be further away from a PoC majority than many observers currently believe.

This survey, however, at most captures only one moment in a larger transformation. Panethnic identities that were introduced with political purposes have evolved into generally accepted demographic categories; the same may happen with ‘people of color’ and, to some extent, the process may have already begun. Media and social-scientific use of PoC and related terms (e.g. students of color, voters of color) has become routine. The greater tendency of younger, more highly educated, and U.S.-born Hispanics to identify as people of color suggests that PoC ID may well increase in the Hispanic population. Perceptions may also change if the Census substitutes a single question on race or origin for the two-question format for Hispanic ethnicity and race. Officially counting those who check Hispanic alone as nonwhite will strengthen the tendency toward viewing all Hispanics as PoC, even though, if this survey is correct, only a minority currently see themselves that way.

Tracking how many Americans self-identify as PoC could be an important indicator about how the racial order in the United States is changing. We see the Princeton PoC ID survey as a preliminary effort to establish a baseline estimate. Especially on surveys that use a single question on race, an additional question could minimise misunderstandings about whether individuals who self-identify as Hispanic, Asian, MENA, native American, or multiracial also self-identify as PoC. Simply asking, ‘Do you consider yourself a person of color?’ could clarify whether they do.

Acknowledgments

We want to thank Princeton University and Richard Alba, Kristopher Velasco, Naila Rahman, G. Cristina Mora, John McWhorter, Gregory Leslie, and Janet Xu.

Data availability statement

Access to the Stata data and code for the Princeton PoC ID survey is available at https://www.personofcolorsurvey.org/

Disclosure statement

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

Additional information

Funding

This work was supported by Princeton University.

Notes

1 The JSTOR social-science search, conducted via Constellate on 6 October 2021, shows a slight tailing off of articles in the 2010s relative to the 2000s, but this may be due to journals from recent years not yet being included in the database. The search is at https://constellate.org/builder/?start=1900&end=2021&keyword=%22people%20of%20color%22%20OR%20%22person%20of%20color%22%20OR%20%22persons%20of%20color%22&provider=jstor&doc_type=article&category=Social%20sciences

2 We abandoned an early plan to survey usage of ‘BIPOC’ because of concerns about how well understood the term was. The survey did ask, however, whether respondents knew what BIPOC means: 56.8% said they were unfamiliar with the term, and only 18.6% were able to identify what is generally regarded as the correct meaning (‘Black, Indigenous, and People of Color’).

3 We borrow the term ‘perceptual dissonance’ from Rodríguez (2000, 136), who uses it for differences between individuals’ racial perceptions and the perceptions of others such as interviewers. We do not have any interviewer ratings and instead use ‘dissonance’ for the difference between self-identifications and reflected appraisals.

4 The questionnaire used for the survey is available at https://www.personofcolorsurvey.org/.

5 See Table 2a. Consonant and Dissonant Identity as a Person of Color (PoC) (Detailed) at https://www.personofcolorsurvey.org/

6 See supplemental Table 2a at https://www.personofcolorsurvey.org/.

7 See supplemental Figure 1 with confidence intervals for the means at https://www.personofcolorsurvey.org/

8 More detailed results for the logistic regressions in Table 5 are available at https://www.personofcolorsurvey.org/

9 The weighted estimate of 47.6% for the proportion of foreign born among U.S. Hispanic adults comes from the 2016–2020 ACS 5-year estimates.

10 The weighted estimate of 77.2% for the proportion of foreign born among U.S. Asian American and Pacific Islander adults comes from the 2016–2020 ACS 5-year estimates.