Abstract
We identify two dimensions of race for the Latino/Hispanic population in the United States—Latinos as one category among the various categories of the U.S. “ethno-racial pentagon” and racial or color differences among Latinos. In a major change from the previous (two-question) format, the Census Bureau recommends a one-question format for capturing ethno-racial distinctions in the 2020 census, which efficiently captures the Latino population on the first dimension and is consistent with racial classification and identification in the real world. At the same time, it nearly eliminates the problem that the two-question format fostered of classifying many Hispanics as “some other race” while maintaining a similar number of Americans classified as Hispanic or Latino. Whether the Census Bureau adopts the one- or two-question format is yet to be decided as of this writing. However, neither format is sufficient for capturing racial distinctions among the fast-growing Latino population, thus precluding effective monitoring of racial disparities in the United States.
In 1977, the U.S. Congress passed the Office of Management and Budget’s (OMB) Directive 15, which regulates the classification of Latinos or Hispanics 1 in the census and other official surveys (Bean and Tienda 1988). OMB Directive 15 deems that Hispanics/Latinos are an ethnic but not a racial group in the U.S. Census. In census publications and in social practice, however, Hispanics or Latinos are often treated as a separate category, whether it is called racial or ethnic, apart from black, Asian, white, and native Americans. More to the point, the Census Bureau attempts to treat race and ethnicity as separate dimensions for Hispanics since the 1980 census go against Latinos’ own cognitive understandings of race and ethnicity—concepts that are often considered indistinguishable.
There are two dimensions of race and ethnicity among Hispanics, both of which are important and the census needs to adequately capture (Telles and Murguia 1990; López 2013): whether the respondent is in the category Hispanic/Latino, and how the person may be perceived in traditional “racial” terms. The first dimension is adequately captured by the current census two-question format and would be captured by a single question. The second dimension is also important, but it is inadequately captured by the census. Whether it was the Census Bureau’s intent, analysts have used census data to model those two dimensions, but researchers have shown the capture of the second dimension to be woefully inadequate.
The Census Two-Question Format and the Hispanic Response
A two-question format in the census since 1980 has sought to implement the notion that Hispanics are an ethnicity and not a race. The Census Bureau assesses Hispanic ethnicity through a separate question and forces Hispanics to identify in the census’s race categories, where Hispanic is not a category (Bean and Tienda 1988). The census introduced a question in the 1980 census that simply asks whether respondents are Hispanic or Latino in addition to the standard race question. Several versions of that two-question format were tried between 1980 and 2010, with changing wording of the questions and categories and order of the two questions (the 2010 version is reproduced in Alba, Beck, and Sahin, this volume). The “race” question never included the category Hispanic/Latino, and the large majority of Hispanics, faced with the available choices, chose either white or “other” (i.e., “some other race”). Less than 5 percent chose black, and far fewer chose any of the other terms.
Presumably the two-question format sought to eliminate confusion of race with ethnicity. The notion that race and ethnicity can be clearly separated is a common one and has been widely accepted. The Institute of Medicine defines race as “a sociocultural concept wherein groups of people sharing certain physical characteristics are treated differently based on stereotypical thinking, discriminatory institutions and social structures, a shared worldview, and social myths,” whereas ethnicity is defined as “shared culture and way of life … reflected in language, folkways, religious and other institutional forms” (Smedley, Stith, and Nelson 2003; Allen et al. 2011). However, these definitions do not seem to hold up on close inspection. So-called ethnic groups, such as Jewish Americans, have been characterized as having distinct physical features and discriminated against accordingly or discriminated against by their religion or culture (Roediger 1999), and so-called racial groups, such as African Americans, have been described as having a shared culture and way of life.
For such reasons, many analysts understand race and ethnicity as overlapping concepts (Cornell and Hartmann 2006; Hughes et al. 2006). Whereas ethnicity was traditionally defined as based on culture and specifically as social groups with common cultural attributes, Norwegian anthropologist Frederik Barth (1959/1998) redefined ethnicity as based on the salient social boundaries that humans create in their social interactions. Nearly 50 years later, Wimmer (2008) would expand Barth’s classic definition and argue that race should be considered as one component of ethnicity.
Analysts have used the two questions together in their attempts to understand “racial” differences among Hispanics (Denton and Massey 1989; Logan 2010). Denton and Massey (1989), for example, examined segregation among Hispanics in Northeast metropolitan areas and assumed that those who self-identified as black are phenotypically black, those who identified as white are actually white, and those who identified as other or some other race are mixed. Logan made the same division and called these groups black Hispanics, white Hispanics, and Hispanic Hispanics. Notably the racial designations were based on self-classification, but analysis of surveys has shown that self-identified race among Hispanics does not reflect phenotypic or color differences (Landale and Oropesa 2002; Roth 2012; E. Vargas et al. 2016). Although dark and phenotypically African Hispanic respondents experience discrimination on the basis of phenotype, their racial self-identification may be a poor proxy for their racial ascription.
Note the example of Dominicans, who largely appear black. In the 2000 U.S. Census, only 13 percent of Dominicans also identified as black, and in 2010 that number increased to only 18 percent. Most Dominicans are probably considered black in the American understanding of the term (Roth 2012; Candelario 2007), but they have been routinely classified as Indio in Dominican identity cards (Howard 2001). 2 Since the early twentieth century, the black racial category, African culture, and blackness became associated with Haitians, which were seen as antithetical to Dominican-ness (Moya Pons 1986; Howard 2001; Candelario 2007). Only recently, have some (non-Haitian) Dominicans begun to embrace blackness and identify with black and especially mulatto categories (Howard 2001; Simmons 2009; Telles and Paschel 2014).
The Hispanic Response: Cognitive Understandings of Race and Ethnicity
The two-question format used by the census does not capture the way that Hispanics tend to think about race; the census’s separate notions of race and ethnicity simply do not coincide with the lived experience/worldview of many Hispanics (Hitlin, Brown, and Elder 2007). Social science research has found that cognitively Hispanics consider Hispanic/Latino as both race and ethnicity and, racially, that Hispanic is a group separate and in addition to black, white, Asian, and American Indian (Hitlin, Brown, and Elder 2007). This is consistent with a popular U.S. system of racial classification, which Hollinger (2006) described as America’s ethno-racial pentagon. Even the Census Bureau treats Hispanics as a (de facto) fifth race, by typically publishing its reports to show Hispanic demographics as a fifth column alongside those of the four traditional (non-Hispanic) racial groups. In practice, though, Hispanics are often understood in both ethnic and racial terms, which some analysts prefer to call ethnic-racial or ethno-racial because of their overlapping and conceptually indistinguishable uses. 3 This lived experience in the United States, which contributes to understanding Hispanic as a separate race/ethnic category, is supported by the finding that, for immigrants, longer residence in the United States, or, for U.S. natives, more generations since immigration, is positively correlated with identification as Hispanic/Latino and rejection of the white racial category (Golash-Boza and Darity 2008).
Census design should obviously consider the social psychology of how respondents interpret questions, question formats, and response categories. Hispanics cognitively understand Hispanic or Latino as a race, while the census seeks to parse out race and ethnicity (Hitlin, Brown, and Elder 2007). Campbell and Rogalin (2006) find that when self-identified Hispanics were faced with a U.S. Census–style two-question format, nearly 80 percent identified as white; but when the category Latino was added to race, 78 percent of them switched to Latino. Rather than a conceptual confusion of race and ethnicity by Hispanics (see for example Mays et al. 2003), this switch is at least as much a reflection of how some analysts and the Census Bureau are out of step with popularly held notions of race and ethnicity.
The Census’s Proposed One-Question Format for 2020
Based on extensive testing of questions in very large census survey trials, including the Alternative Questionnaire Experiment in 2010 and the National Contents Test in 2015, the Census Bureau has proposed the biggest change in the race and ethnicity question(s) since at least 1980: a one-question format that eliminates the Hispanic ethnic question and adds a Hispanic/Latino category in the race question (Perlman and Nevada 2015). In other words, Hispanic would be treated as a race (and possibly MENA [Middle Eastern and North African] would also become a new racial category) in addition to the traditional categories. Census Bureau analysts have found that the overall count of Hispanics is the same with either format and that the new format leads to a dramatic reduction in the selection of “other” or “some other race.” Moreover, identification as black under either format remains the same, which means that there are similar undercounts of the Afro-Latino population under either format.
As of this writing, though, a sudden change is afoot. Despite years of analysis by the Census Bureau and its recommendation to create a single ethno-racial question, the OMB has recently recommended that the previous two-question format be continued into the 2020 Census (Wang 2018). However, the question itself, like all other items on the census, is subject to final approval by Congress.
Two Dimensions of Race and Ethnicity for Hispanics
Notwithstanding the Census Bureau’s appropriate decision to move to a one-question format, there is still a need to capture the dimension of race among Hispanics that refers to their phenotype or racial ascription. But that cannot be done in the way the Census Bureau has envisioned. The design of such a question needs to consider how census respondents understand the item.
Dimension 1: Racialization as Hispanic/Latino
For reasons cited above, Hispanics/Latinos (and other Americans) often see themselves as a race/ethnic group regardless of the OMB directive. This is a first-order cognitive understanding of race or ethnicity. Also, much social science research reveals that Hispanics/Latinos are often treated as a group (Cobas, Duany, and Feagin 2015) and commonly considered and treated as nonwhite (Menchaca 2001; Roth 2012; Vargas 2015). And in some places, the category Hispanic has real-world understandings and consequences. Note that we also could substitute national categories for Hispanic/Latino but the census favors a single panethnic category.
Since 1980, the U.S. Census has used the panethnic category, Hispanic or Latino, whereas previously they had used national categories, particularly Mexican, Puerto Rican, and Cuban, which were the primary groups in the United States from Latin America. These three groups were regionally segregated, had differentiated interests, and their own spokespeople; but in the 1970s, driven by census, business and media interests, a coalition of these groups created a new ethnic category (Hispanic) to further their increasingly common political and economic interests (Mora 2014). While recent immigrants tend to identify with their national group, they increasingly choose the panethnic category especially with increasing time in the United States (Itzigsohn and Dore-Cabral 2000) and over subsequent generations since immigration (Golash-Boza and Darity 2008). 4
The oft-repeated phrase by the census and some analysts that “Hispanic/Latino (like Mexican, Puerto Rican, etc.) is not a race” is anachronistic and harks back to racial science designations of “the races” and the biology books that until the 1960s essentialized this notion (Morning 2011). At the same time, these textbook notions go against other biologically based decisions by the U.S. Supreme Court and by local courts in the nineteenth century to classify Mexicans racially, as either white or a separate race (Menchaca 2001). More recently, police and other institutions continue to classify suspects and others in a scheme where Hispanic is a major racial category. In institutions such as California prisons, inmates are “segregated by race” to diminish violence, reflecting the sharp racial boundaries of street gangs (Goodman 2008). In that highly racialized environment, Hispanics, African Americans, and whites are kept apart from each other, regardless of the insistence that Hispanics are not a race or the Census Bureau’s notions and their neat preordained categories.
Race is a social construction and if members of a society deem certain individuals as belonging to a distinct race, then that view reflects the racialization of that society, whereby its members are classified in particular categories and then treated accordingly. In other words, society often distinguishes Latinos as a race apart but there is another dimension. Society, including Latinos themselves, also often racializes Latinos who appear African and those with dark skin.
Dimension 2: Racialization by color or as black, white, mestizo, and so on
Social science data show that U.S. Latinos and Latin Americans are socially ranked along a racial hierarchy of skin colors and phenotypes (Telles and Murguia 1990; Rodriguez 2000; Telles and the Project on Ethnicity and Race in Latin America [PERLA] 2014) in much the same way as there is a racial hierarchy in the United States. But that hierarchy within the Hispanic population is not adequately captured in the census. When asked about one’s race and ethnicity, Hispanics/Latinos are likely to “racially” self-identify as Hispanic/Latino or under a nationality group like Mexican or Dominican. At the same time, their lived experience also reflects their treatment in American society, where phenotype, especially a black phenotype, affects their life chances. Yet they often do not identify as black even though they may be perceived as such (Candelario 2007; Roth 2012). The second dimension, which we refer to as a second-order category of race, is fluid, while the first-order identification as Hispanic is not. For example, Hannon and DeFina (2015) find in a longitudinal study that only 2 percent of respondents switched classification as Hispanic from one year to another, but fully 44 percent of Hispanics switched their second-order racial classification.
The U.S. Census is inadequately equipped to capture the phenotypic and color gradients that characterize racialization within the Hispanic population. As we previously noted, the white and black categories that have been used to make such distinctions among Latinos (Logan 2010; Denton and Massey 1989) are poorly captured by the U.S. Census. For many persons considered black Latino, Afro-Latino or Latino of African descent, black self-identification in the U.S. Census is a second-order identity. For example, Landale and Oropesa (2002) find that when asked to identify themselves by race, 47.5 percent of Puerto Ricans identified as Puerto Rican, 40.6 percent identified as Hispanic or Latino, and less than 1 percent identified as black. 5 Thus, many such persons fail to acknowledge second-order racial identity in the census. It is possible to capture this additional dimension along with the Hispanic dimension of race/ethnicity with the “mark more than one” option in the new one-question race format but often only one, “Hispanic,” is checked. In the two-question option, many such persons probably choose “some other race.” Moreover, the census does not capture phenotype or skin color among Latinos (or anyone else) beyond the black category. Latinos rarely identify as indigenous except in the relatively small number of cases in which they identify with an indigenous culture or speak an indigenous language.
Latinos in the United States are immigrants or descend from Latin Americans coming from twenty countries with their own, often complex, racial histories. (Latinos also include U.S.- and island-born Puerto Ricans, who are U.S. citizens, and a few who are actually descendants of Mexicans living in areas conquered by the United States [not immigrants].) Latinos/Latin Americans span a wide range of phenotypes comprising varying degrees of European, indigenous, and African ancestries, and many, perhaps most, are racially mixed. Indeed, many of the nations from which they originated (including Puerto Rico) stress racial mixture as an essential part of the national character. Despite the rhetoric of race mixture leading to homogenization and racial egalitarianism among the national populations, the origin societies are racially stratified, characterized as pigmentocracies where skin color gradients tend to align along a social and economic hierarchy (Telles, Flores, and Urrea-Giraldo 2015). While some Latin American countries tend toward the European phenotype or origins (e.g., Argentina), others toward black (Brazil and Panama), and others toward indigenous (Bolivia, Guatemala, Mexico), most have incorporated ideas of mixture based on the idea that racial distinctions are generally difficult to make because of a history of mixing. In Mexico, for example, most are considered mestizo (mixed race) yet they vary widely by skin color, which is a primary axis of stratification in Mexican society (Martínez Casas et al. 2014; Villarreal 2010). Thus, the so-called racial categories fail to capture such racialized differences.
Note that there are also racialized differences based on color or phenotype and referred to as colorism among African Americans, but African Americans are captured in a single category in the census. Skin color and other phenotypical differences result in differential treatment in schools, the labor market, the dating and marriage market, the criminal justice system, and other dimensions of society (Dixon and Telles 2017). The census thus does only an approximate job of monitoring discrimination by capturing the first order of identification, but it fails to capture the racialized dimension of phenotype variation within the black population, as it does among Hispanics.
Despite the U.S. tendency to lump all of Latin America together, the twenty nations in the region have quite diverse racial histories, understandings of race, racial terminology, and patterns of racial classification/identification (Telles and PERLA 2014; Telles and Paschel 2014). Racial conceptions, racial identification, racial variation, and racial stratification in countries where many Latino Americans originate may vary substantially. In Mexico, where 64 percent of U.S. Latinos originate, the race question may sound odd. For example, fully 20 percent of the population when asked about their race or ethnicity does not understand the question, and another 20 percent considers it to be “Mexican” (Flores and Telles, n.d.). Cuba, Puerto Rico, and the Dominican Republic compose the next largest groups. The Caribbean countries are largely of African origin yet they and the U.S.-bound emigrants they spawned have quite distinct racial histories and politics (Sawyer 2006; Duany 2005; Candelario 2007).
Racial Identification in the U.S. Census among Hispanics
The identification of about half of Hispanics as white, and an increase of “white Hispanics” from 1990 to 2010, has been used as evidence of how Hispanics are or have become white (Cohn 2014; Yancey 2003), perhaps a result of mobility or an acceptance by whites. Indeed, the Department of Justice recently reported that most persons caught illegally crossing the U.S.-Mexico border to the United States were white, based on similar information. Hispanics who are of lighter skin color are more likely to be of higher socioeconomic status (Telles and Murguia 1990; Montalvo and Codina 2001). However, Hannon and DeFina (2015), utilizing the General Social Survey, find that for Latinos of similar skin color, identifying as white is unrelated to socioeconomic status or closeness to whites, suggesting that their reasons for identifying as white are not because they have socioeconomically assimilated or acculturated but rather because of inadequate questionnaire design. On the contrary, analysts have found that among persons of similar skin color, Latinos with higher income were more likely to refuse to identify as white (Frank, Akresh, and Lu 2010; Golash-Boza and Darity 2008). Dowling (2014) finds that Mexican Americans often identify as white not because of their phenotype—many understand that they are considered nonwhite—but because of ideologies and discursive strategies about citizenship and aspirations where white is understood as American. Moreover, identification as white occurs because the current format does not offer a Hispanic category for race even though more than 40 percent choose “other,” apparently rejecting the census’s racial categories (Pastor 2014; Vargas 2015). Pastor (2014) notes that slight changes in wording, particularly the admonition in 2010 that “For this census, Hispanic origins are not races,” led to increases in identification as white among Hispanics.
As mentioned earlier, at the other end of the white-black continuum, only 5 percent of Hispanics and about 15 percent of Dominicans select the black category, even though some argue that most Dominicans in the United States would probably be considered black (Candelario 2007; Roth 2012). As in the Dominican Republic, Brazil, Puerto Rico, and other parts of Latin America (Candelario 2007; Telles 2004; Duany 2005), many Latinos avoid the black category because black is the most stigmatized category, instead self-classifying in U.S. Census categories such as other and white. On the whole, those who do identify as black may be the more assimilated than those who are treated as black. Golash-Boza and Darity (2008) find that greater experience with color discrimination is associated with greater identification as black. However, they and others (Landale and Oropesa 2002; Hannon and DeFina 2015) find that skin color is only loosely associated with identification as black. The U.S. Census thus does not reliably capture race differences within the Hispanic population.
Conclusion
The U.S. Census has come a long way in efficiently enumerating the Hispanic/Latino population, but it has a way to go in capturing racial distinctions among it. Racialization among Latinos or Latin Americans is often based on phenotypical differences, particularly skin color, yet these differences are often not considered consequential in the United States or the many countries of Latin America. Extensive research has shown that dark skin color and African origin negatively affect social mobility and the opportunities available to Latinos in the United States, just as they do throughout Latin America. The U.S. Census should pay more attention to ways of effectively capturing such characteristics. Considering that the phenotypically diverse Latino population is now the nation’s largest minority, counting that diversity is now critical for understanding racial disparities.
The U.S. Census laudably attempts to designate Afro Latinos. However, it has not been able to fully identify that population, in either the old or the new format. In the old format, Latinos—identified in the Hispanic question—could also identify as black in the race question, and in the new format, respondents can check both the Latino/Hispanic and black categories. Indeed, the Census Bureau has found that the change from the old to the new format does not really change the number of persons that are classified as Afro-Latino. The problem has not so much to do with the census question format as with self-identifications. Latinos, even when they are seen as black, often do not identify that way. In the U.S. context, where single identities are most common, they may be seen as primarily Latino and see themselves that way; perhaps more worrisome, many see blackness as an identity to be avoided.
So how do we overcome this problem? Perhaps a long-term strategy, which has been tried by the Afro-Latino Forum in New York City, is to raise awareness and consciousness through TV spots about race in the Latino community for Hispanic media markets. By exposing Hispanics to the importance of checking off both black and Latino and by reducing the stigma of blackness, self-identification as Afro-Latino might be enhanced. The use of supplemental surveys that measure skin color and hair type also provide direct measures of observed racial status, which can be used to assess an Afro-Latino population as well as distinguish the very heterogeneous Hispanic population by phenotype. Surveys might also consider a question on reflective race like “What race do others think you are?” or “Do others ever consider you black or of African origin?” Finally, the Census Bureau should consider publishing data specifically on Afro-Latinos, while seeking to specifically improve its ability to capture that population.
The problem of identifying Afro-Latinos notwithstanding, the Census Bureau has recommended the one-question format for the ethno-racial question in the 2020 Census. This new format, which considers Hispanic or Latino one category among the various racial terms of the U.S. “ethno-racial pentagon,” is more consistent with racial classification in the real world. At the same time, it nearly eliminates the problem that the old (two-question) format that fostered classifying many Hispanics in the residual “some other race” category while it maintains the number of Americans classified as Hispanic or Latino the same as in the old format. Unfortunately, as of this writing, the Trump administration, through the OMB, threatens to roll back these advances.
Footnotes
NOTE:
The author would like to thank Emilce Santana and Richard Alba for their comments. At the time of this writing, the Office of Management and Budget (OMB) decided to retain the two-question format for the 2020 Census despite the Census Bureau’s recommendations to move to the one-question format. By the time this article is published, it will likely have congressional approval to keep the old format. This decision, nevertheless, does not change my fundamental argument here.
Notes
Edward Telles is distinguished professor of sociology at the University of California, Santa Barbara. He has authored Pigmentocracies: Ethnicity, Race and Color in Latin America (The University of North Carolina Press 2014); Race in Another America: The Significance of Skin Color in Brazil (Princeton University Press 2004); Generations of Exclusion: Mexican Americans, Assimilation and Race (Russell Sage Foundation 2008); and many articles in leading academic journals.