|Home | About | Journals | Submit | Contact Us | Français|
The racial and ethnic makeup of the American people is in flux. New immigrants from Asia and Latin America have added a large measure of cultural and phenotypic diversity to the American population in recent decades, just as waves of immigrants from Southern and Eastern Europe did a century ago (Bean and Stevens 2003; Higham 1988; Lieberson and Waters 1988: Ch. 2; Thompson and Whelpton 1933: Ch. 2). Moreover, the boundaries between racial and ethnic groups are becoming blurred by high rates of intermarriage and the growing number of persons with mixed ancestry (Lee and Bean 2004).
Descriptions and projections of the racial and ethnic composition of the American people appear kaleidoscopic, with varied accounts and interpretations. Some commentators anticipate a new melting pot, often labeled as the “browning of America,” characterized by continued blurring of once-distinct racial and ethnic divisions (Rodriguez 2003). This interpretation is consistent with the thesis of the declining significance of race and ethnicity in American society. Others see new racial divisions arising as some immigrant groups are allowed to integrate with an expanded and privileged white population, while other groups are “racialized” as disadvantaged brown and black minorities (Bonilla-Silva and Glover 2004; Golash-Boza 2006). These conflicting accounts arise, in part, because of differing ideological presuppositions, but also because racial and ethnic identities are not mutually exclusive or immutable (Barth 1969; Alba 1999).
The US Census Bureau recently released population projections showing that non-Hispanic whites will no longer be the majority of the population in 2042 (US Census Bureau 2008b; Roberts 2008). Most media accounts of these forecasts neglect to report that whites (as opposed to non-Hispanic whites) are actually projected to remain the large majority (upwards of 70 percent in 2050) of the US population (for earlier accounts, see Pellegrini 2000; US Census Bureau 2004). Census Bureau projections by race are flawed, however, because they ignore the relatively high levels of intermarriage and the variations in racial and ethnic identities of mixed-ancestry descendants (Hirschman 2002; Perlmann 2002). More nuanced population projections, produced by the 1997 National Research Council Panel on the Demographic and Economic Impacts of Immigration, incorporate alternate assumptions about current and future trends in immigration, intermarriage, and identity choices (Smith and Edmonston 1997: Ch. 3). These projections, subsequently updated by Edmonston, Lee, and Passel (2002), also show a decline in the proportion of non-Hispanic whites, although not as rapid as in Census Bureau projections. Population projections by race are heavily dependent on the identity choices of persons of multiple racial and ethnic origins (ibid.: 249). Assuming current trends continue to 2050, about a quarter of Asian Americans and African Americans will have recent mixed ancestry, as will nearly half of all Hispanic Americans (ibid.: 246−247). These levels of racial and ethnic mixing and an increased awareness of multiracial ancestry are likely to reshape racial and ethnic boundaries in the coming decades.
Many Americans have multiple identities that reflect complex ancestral origins, tribal and communal associations, and varied ideological outlooks on race and culture. In general, people do not change their ethnicities as a matter of fashion, but they may emphasize different aspects depending on the circumstances. For instance, a person who identifies as Mexican among relatives might identify as Hispanic at work and as American when overseas. A person of mixed heritage might be Native American in one context, but white in another. These possibilities exist in census data, just as they do in informal conversations and settings, because of the opportunities for varied responses to different census questions about race and ethnicity.
In this article, we compare different accounts of the racial and ethnic composition of the American population and measure the degree of overlap of identities for the largest racial and ethnic groups. Our analysis relies on responses to questions about race and ethnicity in the 2000 census, although we argue that these data should not be viewed uncritically. Our interpretation draws upon a historical perspective and emphasizes the inherent subjectivity of census measures of race and ethnicity. Most Americans, except for recent immigrants, are probably descended from multiple geographic, ethnic, and racial origins. Even with strong sanctions against intermarriage, there is considerable historical, literary, and genetic evidence of ethnic and racial mixing among all of the peoples who have settled in the United States (Davis 1991; Hollinger 2003; Gordon-Reed 1998). Yet, many Americans tend to downplay—or are unaware of—this complexity.
America was a multiethnic and multicultural society from the outset. The original American colonies were formed during the seventeenth and eighteenth centuries as frontier societies composed of multiple founding populations (Klein 2004: Ch. 2). First among these were the indigenous peoples of North America, who were gradually displaced or absorbed by the more numerous European settlers and indentured servants from various parts of the world. Africans were imported primarily as slave labor from the Caribbean and West Africa, although some arrived as indentured servants on terms similar to whites. In the middle decades of the seventeenth century, some blacks became free settlers, but by the close of the seventeenth century, slavery and African heritage became nearly synonymous (Fredrickson 1981). With unbalanced sex ratios in frontier settings, large populations of mixed ancestry soon emerged, particularly in Southern colonies (Davis 1991). While some unions were the result of intermarriage or consensual liaisons, there was also widespread sexual exploitation of black women by white slave owners (Fredrickson 1981: Ch. 3).
The ethnic and racial landscape became even more complex during the nineteenth century. Continental expansion added lands that had been home to Native Americans and peoples of mixed indigenous and Spanish origin, and successive waves of immigration from Europe and Asia fueled the rapid growth of an increasingly diverse population. Tracking the mixed and un-mixed descendants from these many threads is a theoretical possibility, but not one that can be easily accomplished with historical or contemporary data. The problem is that the differential rates of settlement, natural increase, and intermarriage (or sexual unions) that produced progeny of mixed ancestry are largely unknown. Small differences in assumptions about the relative magnitudes of these processes can lead to greatly different estimates of the ancestral origins of the contemporary American population.
An even greater obstacle to describing the ethnic makeup of the American people is the assumption that most people are able and willing to accurately report the origins of their parents, grandparents, and more distant ancestors. In many cases, knowledge of ancestral origins is passed along in families or communities, but in some cases these narratives are suppressed or simply lost to history. As a result, the racial and ethnic composition recorded in censuses, surveys, and administrative records reflects a large degree of subjectivity and even speculation, in addition to actual patterns of genealogical descent. Methodological studies of census questions about race and ethnicity, for instance, show that responses are affected, often remarkably so, by the format of questions, the listed choices, and the examples included in questionnaire instructions (Farley 1991; Hirschman, Alba, and Farley 2000).
In this article, we argue that responses to census questions about race and ethnicity measure identity, which is theoretically distinct from ancestry, the geographic origins of one's ancestors. While ancestral origins are potentially objective facts, identities are subjective articulations of group membership and affinity. Ancestry influences identities, but its impact is mediated by a number of factors, including ethnic admixture (blending), the awareness and preservation of knowledge about ancestral origins, prevailing ideologies about race and racial divisions, and the number of generations removed from the arrival of immigrant ancestors.
With an awareness of these limitations, we offer an in-depth portrait of the racial and ethnic composition of the American population, circa 2000, framed within a historical perspective of how racial and ethnic identities have evolved in the United States. For recent arrivals, especially from Asia and Latin America, we note the impact of immigration and report on the emergence of multiracial and panethnic identities. For populations long resident in the United States, we examine reports of ancestry to assess the residue of historical patterns of ethnic blending and ethno-racial hierarchies.
If race and ethnicity were purely cultural phenomena, with little attachment to stratification and political processes, we would expect the long-term outcome to be increasing racial and ethnic entropy—the gradual weakening and eventual disappearance of race and ethnicity as distinct groups with clear boundaries. Our analyses of the 2000 census reveal two patterns through which this increase in entropy can take place: a tendency toward multiple identities and another toward the “Americanization” of identity. By Americanization we refer to the replacement of detailed ethnic origins with simplified panethnic or racial categories, which are shaped and often reinforced by political and socioeconomic divisions. This process is most advanced for blacks and whites, who acknowledge relatively little ethnic complexity or detail and virtually no overlap with one another.
Recent immigrants from Asia and Latin America, by contrast, are more likely to claim national-origin identities, although there is evidence of emerging Americanized identities among the native-born, who are more likely to identify themselves simply as “Asian American” or “Hispanic.” Unlike whites and blacks, Asians and Latinos are also more likely to report multiracial ancestry, which reflects both mixed-ancestry diasporas and the rising levels of intermarriage within these communities.
A different pattern is evident for descendants of the indigenous peoples of North America and the Pacific Islands, who disproportionately report mixed-race ancestries. Although tribal and regional identities are still prevalent, a substantial minority reports panethnic or Americanized identities by simply identifying themselves as “American Indian” or “Pacific Islander.” These results are discussed only briefly here, but a longer report on the AIAN (American Indian and Alaska Native) and NHOPI (Native Hawaiian and Other Pacific Islander) populations is available from the authors.
Relative to other topics in the decennial census, race and ethnicity comprise a large portion of the questionnaire. As shown in Figure 1, the Census 2000 long form, which was sent to 1 in 6 households, includes three subjective measures—race, Hispanic origin, and ancestry—plus related questions on nativity (birthplace) and language (home language and English competence).
In recent years, the question on race has grown to include over a dozen categories as well as options for write-in responses (US Census Bureau 2002). For Census 2000, the question was further changed to allow respondents to choose multiple racial identities (Tofoya, Johnson, and Hill 2005). Hispanic/Latino origin was added as a sample item in the 1970 census and then moved to the 100 percent form starting in 1980, where it remains to this day. During that same year, the census introduced an open-ended ancestry question, which asked respondents to write in their “ancestry or ethnic origin” (Farley 1991; Lieberson and Waters 1988). At present, the Census Bureau releases data on more than 200 ancestry groups coded from the open-ended responses.1
To some observers, the preoccupation with increasingly detailed data on race and ethnicity has gotten out of hand (Prewitt 2002; Hochschild and Powell 2008: 89). The addition of questionnaire items, coupled with an ever-increasing number of categories, suggests that identities have grown so complex that the point of diminishing returns to measurement has long passed. Viewed through the lens of American political history, however, the current proliferation of racial and ethnic categories (and combinations) in the census is simply the latest chapter in the saga of a society long stratified by race and preoccupied with racial measurement (Anderson 1988, 2002; Prewitt 2005; Snipp 2003).
The idea of classifying and counting individuals by race was developed during the Revolutionary and Antebellum periods (Anderson 2002: 269−271). “Color” was deeply intertwined with legal status and citizenship rights at the time. The classification directed by Article I, Section 2 of the US Constitution distinguished between three groups for purposes of taxation and Congressional apportionment: “free persons” (including indentured servants), “other persons” (a euphemism for black slaves), and “Indians not taxed” (those living beyond areas of white settlement and control). Slaves were counted as only three-fifths of free persons, while “Indians not taxed” were not counted at all (Anderson 1988: 9; Klinker and Smith 1999: 25). While the Constitutional language obscures the color-coded nature of the classification (“Indian” is the only race referred to by name), the census classification was more explicit with respect to race. Between 1790 and 1810, census takers were instructed to report the number of free whites, free nonwhites, and slaves (Anderson 1988: 14; Snipp 2003: 564−565).
In the late nineteenth and early twentieth centuries, the classification by race expanded to include the measurement of “mulattos”—persons of mixed black and white ancestry—and the “blood quantum” (percent of white ancestry) of American Indians (Snipp 2003: 565−568). By 1890, the census racial classification scheme reflected a growing preoccupation with identifying persons with even the slightest hint of African ancestry, adding categories for “quadroon” (persons with one-fourth black ancestry) and “octoroon” (persons with one-eighth or less black ancestry). In 1930, Mexicans were added to the growing list of “nonwhites.” Fearing the move as an effort to stigmatize (and possibly deny naturalization to) Mexican Americans by labeling them a nonwhite racial group, the Mexican American population (and the Mexican government) strongly protested the change, and the racial category was soon disavowed by the director of the Census Bureau (Cortes 1980; Schor 2005: 92−93; Hochschild and Powell 2008: 80−81).
Despite the ambiguity suggested by the recurrent changes in categorization, the practice of measuring race by observer classification was routine procedure well into the twentieth century, since skin color and other aspects of physical appearance were thought to be obvious to any observer. Although early-twentieth-century Census Bureau reports acknowledge the limits of using enumerators’ perceptions as a basis for racial classification (US Bureau of the Census 1918: Ch. 11; Schor 2005: 91), racial ideology and government policies were constructed on the assumption that outward appearances signified racial origin. The changing racial classifications used by the Census Bureau reflect the deep ideological and political divides in American society, where color and culture were used at various times to justify slavery, exploitation, and official forms of discrimination (Davis 1991; Fredrickson 2005; Hochschild and Powell 2008). The social science of the late nineteenth century was almost completely dominated by Social Darwinist ideology that reinforced the assumptions of white superiority (Gould 1996). Administrative practices, including census classifications, reflected these biases (Perlmann 2001).
By the middle of the twentieth century, the conceptualization of race had shifted from a biological to a social trait, but interviewer observation remained the primary means of racial classification in censuses and face-to-face surveys. The most important change in measurement occurred in the 1960 census when a mail-out questionnaire replaced door-to-door enumeration for much of the United States (this procedure was adopted almost everywhere for the 1970 census). Although this change shifted the measurement of race from the perceptions of census enumerators to the subjective self-reports of household respondents, there is little evidence that these changes had a measurable effect on the racial composition of the United States (Campbell 2007: 922; US Bureau of the Census 1963: xi). The results of counts based on subjective perceptions of race, it seemed, were statistically indistinguishable from the perceptions of observers. The only major exception was American Indians, whose numbers increased dramatically following the change in measurement that allowed for self-identification (Eschbach 1993, 1995; Passel 1996).
Following the passage of civil rights laws in the 1960s, a new imperative emerged to measure race in census and administrative data (Farley 2004: 126). Federal laws declared discrimination illegal, and racial disparities could, in certain circumstances, be interpreted as evidence of discriminatory behavior. The 1965 Voting Rights Act gave the federal government the right to review electoral boundaries in areas where the potential voting power of racial groups and language minorities might be diluted by local governmental bodies. The Voting Rights Act left the definition of protected groups implicit, but 1975 legislation specified that in addition to blacks, the law was intended to protect the rights of “persons who are American Indian, Asian American, Alaska Native, or of Spanish heritage” (Edmonston and Schultze 1995: 147−148; also see Rumbaut 2006 on the 1976 legislation on gathering economic and social statistics for Americans of Spanish origin or descent). These new federal responsibilities could only be undertaken with detailed census data on racial and ethnic groups, tabulated by geographic area (Edmonston and Schultze 1995: Ch. 7; Edmonston, Goldstein, and Tamayo Lott 1996: 4−15).
In response to growing demands from the Latino community, a Hispanic-origin question was first included in the 1970 census (Choldin 1986; Rumbaut 2006). Whereas in 1930 the inclusion of a Mexican category in the census was seen as an effort to stigmatize Mexican Americans, in the post–Civil Rights era the inclusion of a Hispanic-origin question (separate from the race question) was welcomed as a source of data that could be used to protect the rights of Mexican Americans.
The addition of the ancestry question in the 1980 census was a result of efforts by the descendants of European immigrants to ensure that their national origins were included as part of the American racial and ethnic tapestry.2 A similar political effort was made in the late 1980s to include additional Asian and Pacific Islander groups in the 1990 census form. Initially, the Census Bureau proposed a generic Asian and Pacific Islander category that could be checked and a blank space so that individuals could write in their exact national origin. Representatives of Asian American communities argued that this format might lead to lower counts, and with the help of their Congressional representatives, they successfully had the list expanded to include eight specific origin groups as well as a residual “other Asian” category in the 1990 census form (US Bureau of the Census 1990).
In the midst of these political currents, and with a growing awareness that there was no clear conceptual framework to collect data on race and ethnicity, in 1977 the US Office of Management and Budget (OMB) stepped in by issuing Statistical Directive No. 15, “Race and ethnic Standards for Federal Statistics and Administrative Reporting” (Edmonston, Goldstein, and Tamayo Lott 1996: Appendix B). Statistical Directive 15 specifies five major racial and ethnic categories: American Indian or Alaska Native, Asian or Pacific Islander, Black, Hispanic, and White. The directive posits the basic principle of defining race and ethnicity as descent from peoples originating in distinct geographical parts of the world. For example, an Asian or Pacific Islander is defined as “a person having origins in any of the original peoples of the Far East, Southeast Asia, the Indian subcontinent, or the Pacific Islands” (ibid.: 65−66). For American Indians or Alaska Natives, however, there are two requirements: descent and affiliation: “a person having origins in any of the original peoples of North America, and who maintains cultural identification through tribal affiliation or community recognition” (ibid.: 65, emphasis ours). The most ambiguous definition is of Hispanics, which include any persons of “Spanish culture or origin, regardless of race.”
Although the need to standardize racial and ethnic measures was apparent, Statistical Directive 15 was riddled with inconsistencies. The most obvious flaw for purposes of measurement is that the categories specified are not mutually exclusive. Hispanics, for instance, include varied peoples of European, Amerindian, and African descent. OMB attempted to sidestep this problem by treating race and Hispanic origin as separate measures. Under this two-question scheme, Hispanics could be identified by race, and members of each race could be identified by Hispanicity (Hispanic vs. non-Hispanic). If a survey or administrative form had only one combined race and Hispanicity question (i.e., listed “Hispanic” among the standard racial categories), Statistical Directive 15 required that persons with any Hispanic origin be classified as a distinct group, which meant that persons reported as white or black were limited to non-Hispanic whites or non-Hispanic blacks, respectively.
The use of geographic boundaries to delineate the ancestral origins of the major OMB categories is also tenuous. Persons with origins in North Africa and the Middle East are classified as white, but those from the Indian subcontinent are classified as Asian. The line dividing these regions is far from clear. Even current census coding schemes are inconsistent. The racial codes used in Census 2000 (as well as current American Community Surveys) include “Afghanistani” under white, while the ancestry codes in the same document list “Afghan” under South Asia.3 Similar examples abound. There is no place in the initial OMB scheme for persons descended from the original peoples of Central or South America. Persons from Brazil, the largest Latin American state, cannot be counted as Latino because the definition of Hispanic/Latino does not include persons of Portuguese language. Spaniards from Europe, however, are classified as Hispanic/Latino, even though they have no ties to Latin America. Increasingly, criticisms were expressed by many groups who were dissatisfied with one aspect or another of the OMB classification. Pacific Islanders, especially Hawaiian natives, felt that their inclusion with Asians obscured the unique challenges faced by their communities (Edmonston, Goldstein, and Tamayo Lott 1996: 31). Other ethnic groups and national-origin populations, such as Arab Americans, believed they should be included among the nonwhite populations designated on the census form. Most vocal was a loose coalition of multiracial advocacy groups, who argued that the mutually exclusive categories on the race question forced mixed-race persons to choose only one racial identity (Farley 2002).
The framers of Statistical Directive 15 were undoubtedly aware of these potential problems. Indeed, OMB includes the disclaimer that “these classifications should not be interpreted as being scientific or anthropological in nature.” The guidelines also acknowledge that the classification will not provide an unambiguous identity for persons of mixed ancestry. When in doubt, the OMB suggests that persons should be classified under “the category which most closely reflects the individual's recognition in his community” (Edmonston, Goldstein, and Tamayo Lott 1996: 65−66).
In response to the criticisms of the initial Statistical Directive 15, OMB undertook a review of the measurement of race and ethnicity in federal statistics scarcely 15 years after the 1977 directive was issued. During the early 1990s, the review included a National Research Council workshop that brought researchers, administrators, and other interested parties together, a series of Census Bureau studies, interagency committees, and an invitation for public comments (Snipp 2003: 574−581). In 1997, after extensive study, OMB issued revised standards for the measurement and classification of racial and ethnic data (Office of Management and Budget 1997a, 1997b). The overall definition of race and ethnicity in the 1997 revision of Statistical Directive No. 15 did not change, but the list of racial categories was revised and a few group labels were changed as well. The major revision in the classification was the division of the Asian and Pacific Islander category into two groups—Asians and Native Hawaiians and Other Pacific Islanders—in part due to the influence of Senator Daniel Akaka of Hawaii (Farley 2004: 131). The 1997 classification includes: American Indian or Alaska Native; Asian; Black or African American; Native Hawaiian or Other Pacific Islander; and White. As before, Hispanic/Latino Origin remained a separate ethnic category.
The most important change, however, was not to the classification itself, but to the way in which individuals were instructed to locate themselves within that classification. In a major departure from the original OMB scheme (and from two centuries of census taking), respondents were now allowed to “mark one or more” races with which they identified. This change was prompted by a greater awareness of persons of mixed racial ancestry as well as political advocacy by intermarried families (Farley 2002, 2004).
As with their first effort, the 1997 OMB revision of Statistical Directive 15 generated as many criticisms as it silenced. In addition to the enduring criticism of ad hoc categories, the 2000 census revealed a new, if not entirely unforeseen problem—the profusion of data that resulted from multiple-race reports. Although nearly 98 percent of the population chose only one race, the combinations of the minimum six racial categories (the five OMB parent groups plus a residual “some other race”) created an almost nightmarish problem for census users who were accustomed to mutually exclusive groups. To the six single-race categories were added 15 two-way combinations, 20 three-way combinations, 15 four-way combinations, 6 five-way combinations, and 1 six-way combination, for a total of 63 racial groups, or 126 if cross-classified by Hispanicity (Grieco and Cassiday 2001). Worse yet, these combinations are themselves a simplification of the actual census questions on race and ethnicity, which include several specific origin groups (Mexican or Chinese) under each OMB parent category (Hispanic or Asian, respectively) and allow detailed write-ins (e.g. tribe) in various sections. With this added complexity comes a growing concern about the meaning and utility of contemporary data on race, not to mention practical questions about how to tabulate combinations or compare the revised data with much simpler racial data from earlier censuses (Perlmann and Waters 2002).
The quality of census racial and ethnic data is also affected by the problems of under-enumeration, item nonresponse, and unreliability of measurement. The census undercount, and differential undercount by race and ethnicity, has received considerable attention (Anderson and Fienberg 2001), but the implications for social science research remain inadequately studied. Less visible, and largely absent from most discussion of census measurement, is item nonresponse. Missing data are imputed by the Census Bureau, and our preliminary analysis of the IPUMS .01 file4 showed that 3 percent of the Census 2000 population did not fill in the question on race. Our earlier work showed that nonresponse to the census question on race could be reduced considerably by using a combined question on race and Hispanic origin (Hirschman, Alba, and Farley 2000).
Drawing up matched records from Census 2000 and the Census Quality Survey (a follow-up survey that replicated census questions), Jorge del Pinal (2004: Ch. 4) evaluated the reliability of racial measurement. In the follow-up survey 97−98 percent of whites, blacks, and Asians reported the same race (or had the same race reported by the household respondent) as in the census (ibid.: 27−28). There were, however, much lower rates of consistent reporting for American Indians and Alaska Natives (AIAN) and for Native Hawaiians and Other Pacific Islanders (NHOPI). The lowest rate of reliability is for the multiracial population: only 40 percent of persons reporting more than one race in the 2000 census were also reported to be multiracial in the follow-up survey (ibid.: 27). Similar studies by Harris and Sim (2002) and Perez (2008) reveal poor levels of reliability for self-identified multiracial and Hispanic youth, respectively.
That race and ethnicity are socially constructed is one of the axioms of contemporary social science (Omi 2001; Omi and Winant 1994). But how, exactly, are they constructed and by whom? At the individual level, intergenerational socialization is the primary mechanism for communicating group identities. Children and adolescents develop ethnic consciousness through interaction with parents, siblings, and other family members (Perry 2002). Nonverbal forms of socialization also take place through observation of family behavior, as well as informal interactions with friends and neighbors and in formal settings like schools, businesses, and institutions. These experiences foster a sense of the “ethnic self” through which children learn who they are and, just as important, who they are not.
While childhood socialization is the crucible of racial and ethnic identity formation, the boundaries of racial and ethnic categories and the history underlying their creation are much less straightforward. One view is that racial and ethnic identities are imposed from above. Terms like “Hispanic” and “Asian American” are unique to the United States and were created for data gathering and statistical tabulations by governmental agencies. But these categories also reflect the aims of panethnic coalitions and political advocacy groups, who played a direct role in the construction of racial and ethnic classifications adopted by government statistical authorities (Choldin 1986; Espiritu 1992: 99−103; Farley 2002). More importantly, individuals are free to report their racial and ethnic identity in the census, social surveys, and the vast majority of administrative forms that include a space for racial and ethnic identification. Respondents are instructed to mark the race or races they “consider themselves to be” (see Figure 1), and those who refuse to identify with the listed categories can write in one of their own. The assumption is of widely shared understandings (folk meanings) of racial and ethnic categories and their boundaries.
As noted above, the administrative expectation, given the wording of the category definitions, is that folk understandings of race and ethnicity will be consistent with ancestry—the geographic origins of one's ancestors. For a number of reasons, however, responses to questions about race and ethnicity only partially reflect ancestral origins. Ancestry is a potentially objective characteristic—the countries or regions of birth of a respondent's parents, grandparents, great grandparents, and so on. Identities, by contrast, are subjective claims of affiliation with groups that are recognized in society. Identities overlap with ancestries, but they are also shaped by knowledge, socialization, physical appearance, and culture, among other factors. Birthplace does not vary by social context, whereas identities are contextual by definition. The birthplaces of recent ancestors are often passed down in family conversations (unless there are conscious reasons to suppress them), but for individuals whose Old World roots are distant or complex, there may be only a dim awareness of, and minimal interest in, ancestral origins. Some people with the same ancestry will respond differently to census questions about racial and ethnic identity.
The idea that humankind shares common ancestry through evolution and prehistoric migrations “out of Africa” is widely recognized (Diamond 1993; Oppenheimer 2003). Less well known is that all human beings alive today are likely to share at least one common ancestor born a few thousand years ago, and that everyone alive today is likely to be descended from the same mother and father who lived a few thousand years earlier (Rhode et al. 2004). These conclusions are derived from simulations that consider a range of probabilistic assumptions about the likelihood of mating between adjacent and isolated populations throughout history. In fact, if human mating were fully random, everyone alive today would share a common ancestor just 20 generations back, around 1500 ce assuming 25 years per generation.
Common ancestry does not mean that populations share the same genotype, however. Most genes have only a 50 percent chance of being passed on to the next generation,5 so sporadic contact between isolated populations would not result in a large amount of genetic admixture. More importantly, even tiny differences in the genome, if differentially selected between populations, can account for variation in inherited physical features such as eye, skin, and hair color. These superficial differences, which arose in prehistory, probably resulted from natural selection in different climatic zones and can persist for many generations (Cavalli-Sforza and Cavalli-Sforza 1995).
Despite long stretches of relative isolation, there is evidence that geographically distant populations maintained some reproductive contact throughout history (Rhode et al. 2004), and the pace of contact and exchange has increased dramatically in the last one thousand years, and especially since 1500 (Davis 1974; Diamond 1997; Hoerder 2002; McNeill 1984). In addition to trade and warfare, long-distance contacts invariably led to intermarriage and other sexual relationships that produced offspring of mixed ancestry. Intermarriage and ethnic blending, in turn, diminished physical and cultural differentiation, a process nowhere more evident than in the New World where migrants from Europe, Africa, and Asia intermixed with indigenous peoples, creating entire continents of ethnically and racially mixed populations over the last few centuries (Harris 1964).
Figure 2 underscores the high probability of shared descent for most Americans: it plots the expected number of ancestors over the last 225 years for a person born in 2000 (time is shown along the horizontal axis). Because the number of ancestors doubles in each prior generation (2 parents, 4 grandparents, 8 great grandparents, etc.), the number of ancestors from any previous generation is equal to 2x, where x is the number of prior generations. If the length of each generation is about 25 years, then a person born in 2000 would have had 512 ancestors in 1800 and 1,024 around the time of the American Revolution. Continued exponential extrapolation will of course predict an impossible number of ancestors—more persons than were alive—by the close of the first millennium, a fallacy that results from double-counting persons who occupy multiple slots on a family tree. Most of our distant ancestors were related to one another, just as we are distantly related to most people alive today (Ohno 1996).
Assuming that the number of duplicate entries (related ancestors) is negligible in the short term, many, if not most, Americans probably have hundreds, even thousands, of New World ancestors. Even assuming low levels of intermarriage, most persons with deep roots in the United States are likely to have genealogical descent from at least one ancestor from a different part of the world. Yet in response to census inquiries on ancestry and race, most Americans tend to simplify their origins and report a single identity, even if they are aware of others (Waters 1990; Lieberson and Waters 1993).
In addition to a tendency to simplify, a reporting bias stems from the belief that not all identities are equally desirable. Historical discrimination and prejudice created strong incentives to repress or selectively ignore certain ancestries. On the other hand, identities that are associated with physical appearance, skin color in particular, are often more difficult to “forget” than those associated with language or culture (Fredrickson 2002). One distinctly American form of racial essentialism is the one-drop rule (Davis 1991), which holds that persons with any African ancestry, visible or otherwise, must be classified as black. In the period when discrimination against Southern and Eastern European groups was common, the children and grandchildren of immigrants could “escape” their disadvantaged heritage by changing surnames and cultural practices such as speech, dress, religion, and cuisine (Baltzell 1964). These options were rarely available to persons of partial African descent, regardless of socioeconomic status and other circumstances. In spite of this dominant pattern, large numbers of light-skinned persons of partial African and Native American ancestry are believed to have “passed” into the white community throughout the years (Burma 1946; Broyard 2007), despite considerable obstacles and the loss of family and community ties (Piper 1992). The contemporary descendants of persons who passed from one community to the other may not even be aware of their ancestry.
Prior to Census 2000, race was measured by mutually exclusive and exhaustive racial categories.6 This framework ignored the long history of racial and ethnic admixture in the New World. Populations of mixed European, African, and American Indian ancestry date back to the early years of settlement in North America (Davis 1991). The incidence of black/white intermarriage in seventeenth-century Virginia was substantial enough to prompt a ban on interracial unions (Harris 1964). While these bans probably did little to change behavior, they shifted the balance of interracial unions to the illicit type (Davis 1991; Spencer 2006)—obscuring them from public view and ensuring their absence in public records.
For example, the significant genealogical overlap of black and white Americans, once a topic of frequent study (Hart 1921; Herskovits 1928; Myrdal 1964 ; Stuckert 1958), received only fleeting attention in the decades that preceded the emergence of “ancestry informative” genetic markers, which can now be used to measure levels of admixture (shared ancestry) among ethnically blended populations (Shriver et al. 2003; Parra et al. 1998). Today, lingering denials of the shared ancestral heritage of black and white Americans have crumbled under the weight of biographical and DNA evidence. The long-hidden lineage of Thomas Jefferson and Sally Hemings, Jefferson's mistress who was a family slave, was not a unique case of “forgotten” racial mixing in American history (Gordon-Reed 1998; Dao 2003). Meanwhile, studies of the high and increasing rates of intermarriage among Asians, Latinos, and other groups suggest that ethnic blending will continue (Sandefur and McKinnel 1986; Qian 1997; Qian and Lichter 2007).
Even among nonstigmatized identities, there are many reasons why respondents may simplify complex ancestral origins in censuses and surveys (Lieberson and Waters 1993). Since not all ancestral ties carry equal meaning, many people simply report their primary attachment. With multiple weak ethnic affiliations, different contexts may elicit different primary attachments. There are also strong instrumentation effects on the reporting of identities. Respondents may interpret categories differently when filling out a survey for themselves than when an interviewer administers the questionnaire (Perez 2008). Listed racial and ethnic groups and even examples provided in questionnaire instructions are often interpreted as suggestions, which have been shown to affect the distribution of identities reported in the census and the Current Population Survey (Farley 1991; del Pinal 2004).
Simplification is also built into census data collection and coding procedures. As noted earlier, the original OMB Statistical Directive 15 guidelines relied on the traditional assumption that multiracial respondents should choose a single identity. While the 1997 revision (and the 2000 census) allowed multiple-race reporting, the Hispanic-origin question remains mutually exclusive. One can be Hispanic or non-Hispanic, but not both. Moreover, respondents are limited to a single Hispanic origin (Cuban, Mexican, etc.). The census question on ancestry allows respondents to write in as many identities as they want, but only the first two responses are coded.
Racist ideology undoubtedly affected popular consciousness about ethnic identities as well. Color barriers—whether formal, as in the South, or informal, as in much of the North—sharply limited opportunities for persons of African ancestry in employment, housing, education, and political participation. A linchpin of these practices was the ideological construct of the aforementioned one-drop rule. In many parts of the world, long-standing designations exist for peoples of mixed heritage (Fredrickson 1987, 2005; Nobles 2000; Telles 2004). In the nineteenth century, Americans of mixed African and European ancestry were even recognized as populations of distinction in Charleston, New Orleans, and other Southern cities. These populations largely “disappeared” in the late nineteenth and early twentieth century with the rise of one-drop ideology. This ideology received national sanction after the US Supreme Court affirmed in 1892 that Homer Plessy, a French Creole of less than one-eighth African ancestry, was legally black and thus confined by law to use segregated public services and facilities. By 1925, nearly every Southern state had a version of the one-drop rule on the books. These laws remained in force well into the latter half of the twentieth century and, until recently, were still being used to mandate black identity for persons of partial African descent. In 1982, Susie Phipps sued the state of Louisiana to change the race recorded in her vital records from black to white. In ruling against her, the court cited a state law that declared anyone with one-thirty-second African ancestry to be black (Omi and Winant 1994).
Census enumerations shadowed this construction. As late as 1960, census instructions held that “Negro... [includes] persons of Negro or mixed Negro and white descent...and persons of mixed American Indian and Negro descent unless the American Indian ancestry predominates...” (US Bureau of the Census 1963: x). Evidence suggests the one-drop rule similarly shaped popular consciousness among black and white Americans at the time, for even after the census changed from an enumerator-administered format to a self-reported mail-in questionnaire, there was little change in the proportions who reported themselves as black or white.
In summary, the major axioms of our theoretical argument are as follows. First, although ancestry can be defined as an objective attribute based on geographic descent or genetic markers, reported racial and ethnic identities are subjective articulations of group attachment and affinity. Ancestry influences identity, but its impact is mediated by ethnic admixture across generations, knowledge of ancestral origins, and the number of generations removed from the arrival of immigrant ancestors. Second, while family and community socialization are the primary mechanisms through which identities are reinforced, racial and ethnic categories are shaped by institutional and political forces through 1) laws and sanctions that regulate group rights and opportunities; 2) customs that affirm group claims for recognition and entitlements; and 3) systems of measurement and classification used in administrative records of race and ethnicity at the individual level or in aggregate counts of populations by race and ethnicity (censuses).
In the remainder of this article, we examine the contours and complexities of racial and ethnic boundaries using data from Census 2000. We show that two of the most prominent anomalies in contemporary census data on race—the large “some other race” category and a moderate fraction of multiracial responses—are largely an artifact of the administrative decision to treat race and Hispanicity as separate concepts and measures. We then show that recent trends in immigration and generational replacement are dramatically changing the racial and ethnic composition of the American population. Finally, we examine the prevalence of multiracial and multiethnic identities among major populations classified by race and ethnicity. Our findings suggest that the Americanization of identities and the intermixture of ethnic groups through assimilation and intermarriage obscure the boundaries between ethnic categories, particularly as racial or panethnic categories arise to take their place.
Table 1 shows the reported racial composition for the total population and for Hispanics in the 2000 census. In addition to the five standard racial categories specified by OMB (Office of Management and Budget 1997a, 1997b), we include the residual “some other race” (SOR) category used by the Census Bureau, as well as the largest multiple-race combinations. When asked to report their race in the 2000 census, three-fourths of Americans identified themselves as white only. Among the balance of the single-race identifiers, 12.2 percent are African American, 3.6 percent Asian, 0.9 percent AIAN, and 0.1 percent NHOPI.
As these figures show, the overwhelming majority of the population identified with a single OMB race. Still, over 8 percent of Americans—one in twelve—fell outside the orbit of the traditional list of mutually exclusive racial categories. About 2.6 percent chose more than one race, and more than double that number said “none of the above” by writing in a different racial identity that was coded as “some other race” (SOR). The second set of columns in Table 1 shows that much of this ambiguity is centered in the Hispanic population. Nearly half of all Hispanics are unwilling to identify with a single standard race—6.4 percent indicate multiple origins and 42.6 percent identify themselves as SOR. Presumably, most of the latter represent persons who used the “other race” write-in section to reiterate their Hispanic origins (Mexican, Puerto Rican, Cuban, etc.).
All told, 15.4 million Americans identify themselves solely as some other race, but 15 million (or 97 percent) of these are Hispanic. The SOR category is larger than the Asian, American Indian, and NHOPI populations combined. This quasi-race category is, however, an artifact of a data collection system that does not allow persons of Hispanic origin to be listed as a race (Grieco and Cassidy 2001). Hispanics also constitute 2.3 million of the 7.3 million Americans who reported that they were multiracial—mostly by checking white and writing in their Hispanic identity under “some other race.” As a result, Hispanics augment the multiracial population by 45 percent. Because Hispanics are a large and fast-growing segment of the US population—over 35 million (12.5 percent of the national population) in 2000—their disproportionate refusal to claim standard racial identities clouds any prospect of summarizing the racial composition of the US population without taking Hispanic identity into account (Campbell 2006).
The reported non-Hispanic mixed-race population numbers about 5 million—less than 2 percent of the total population. Although nontrivial in absolute terms—0.8 million persons with a bi-racial white/black identity, 1.3 million reporting a joint white/American Indian identity, and around 0.9 million reporting a combined white/Asian ancestry—these numbers are only a tiny fraction of the true level of Americans with mixed-race ancestry. Historical and contemporary estimates of mixed ancestry among whites, blacks, and American Indians (Myrdal 1964 ; Shriver et al. 2003; Stuckert 1958, 1976) suggest much higher levels of racial admixture within the US population. More than two decades ago, Yinger (1985: 156) summarized the prevailing evidence: “80 percent of black Americans have European ancestry; 50 percent or more of Mexican Americans have both [Latin American] Indian and European ancestors; perhaps 20 percent of ‘white’ Americans have African or Native American ancestors.” With rising levels of intermarriage in recent years (Farley 1999; Stevens and Tyler 2002), the fraction of the population with mixed racial ancestry has certainly increased.
Because most Americans, including many Hispanics, consider Hispanic ethnicity to be on a par with standard “racial” categories, Table 2 presents a revised classification of the American racial landscape that combines the four largest Hispanic groups with non-Hispanics classified by the major racial categories. This combined racial and Hispanic-origin classification reduces the share of multiracial persons and avoids the awkward (and artificial) inflation of the SOR category that occurs only because the Census Bureau does not accept “Hispanic” as a valid response to the question on race. One debatable element of the combined classification is that black Hispanics are counted as Hispanic and not as African American. This assumption, however, is of only minor demographic consequence since the number of black Hispanics (0.64 million) is less than 2 percent of either population.
Table 2, in our judgment, offers a clearer picture of American racial and ethnic diversity than the Census Bureau tabulations and reports that do not combine racial and Hispanic origin (e.g., Grieco and Cassidy 2001; Jones and Smith 2001). Even if Hispanic origin is ignored, the shift from mutually exclusive to “check one or more” racial categories created major analytical and interpretative problems for census users, especially among those who relied on census data for evaluation of civil rights compliance. The racial classification used in various Census 2000 reports and files ranges from 31 to 63 categories and combinations of categories. These classifications are incommensurate with earlier censuses, vital statistics, most survey data, and administrative records. By incorporating Hispanicity as part of a combined race/ethnicity classification, our revised format minimizes the problems of comparability by virtually eliminating the SOR category and substantially reducing the multiracial population.
With an awareness that comparability and interpretability were likely to be lost, the initial administrative response was to avoid the complexity of multiple-race reporting in the 2000 census by creating a statistical “bridge” to single-race categories (Office of Management and Budget 2000). The first generation of racial bridging methods relied on questionable assumptions about allocation procedures and the fractional assignment of persons (Goldstein and Morning 2002). The second generation of bridging methods was developed by the National Center for Health Statistics, which relied on National Health Interview Surveys that asked multiple-race respondents to choose a “primary” race (Ingram et al. 2003). Drawing upon individual characteristics as well as geographic and household features, binary and multinomial discrete-choice models were developed to estimate the likelihood of various single-race responses. The NCHS bridging parameters have recently been extended to public use microdata. The University of Minnesota IPUMS project now includes a “Racesing” (single-race identification) variable that recodes multiple-race respondents in the 2000 census (and American Community Surveys) to the single race with the highest predicted probability (Ruggles et al. 2008).
Racial bridging methods are a key contribution, though not without limitations. Many census users may not fully understand the assumptions and methods that underlie the allocation of multiracial persons to various single-race categories. Ingram and colleagues (2003: 12) accurately note, for instance, that “the goal of bridging is to correctly determine the size of single-race groups, not to correctly determine how each individual would have reported his or her race under a single-race system.” In addition, many data sources lack the necessary covariates and rich geographic measures needed to properly specify the bridging model (Liebler and Halpern-Manners 2008). More important than these methodological concerns, the bridging approach is premised on a need to “correct” (i.e., bypass) multiple-race identities. Our revised classification in Table 2, as noted above, all but eliminates the problematic “some other race” category and substantially reduces the number of multiracial persons without making assumptions about how census respondents might answer a different question.
The first column in Table 2 shows that under the revised classification, only 69 percent of the US population is non-Hispanic white. Blacks and Hispanics each comprise about 12 percent of the total, and, at nearly 35 million each, either of these minority groups exceeds the total population of Canada by several million persons. The rapidly growing Asian American population numbers 10 million and represents 3.6 percent of the total population. The original peoples of North America—American Indians and Alaska Natives (excluding those who reported multiple races)—today comprise less than 1 percent of the US population (about 2 million total), while the indigenous peoples of Hawaii and related Pacific Islander populations (NHOPI) number less than 400,000 (0.1 percent).
With the Hispanic population coded as a quasi-racial category, only half a million persons (0.2 percent) in the combined racial and ethnicity classification in Table 2 fail to report an OMB race, and about 5 million persons (1.8 percent) report multiple-race identities (white/black, white/AIAN, and white/Asian are the largest groups).
The revised racial and ethnic composition shows much greater diversity among children and young adults than among the elderly (Table 2). Age represents a stage in the life cycle as well as the succession of generations. Although racial and ethnic identities are generally stable across the life course, life cycle changes in identities may occur during the transition from adolescence to young adulthood. For example, part of the high percent multiracial among persons aged 0−14 (twice that of adults) may be explained by the sources of household reporting in the census. Multiracial persons are more likely to be identified as such when they are young, when their interracially married parents fill out the census form. There is some evidence of “simplification” of racial and ethnic identities as adolescents and young adults leave their childhood home and form new social networks, particularly when they marry and set up a new family (Lieberson and Waters 1993).
While life cycle changes may play a small role, generational replacement is the primary force behind the pronounced differences in the racial and ethnic composition between younger and older Americans. The trend toward greater racial and ethnic diversity reflects the effects of immigration and differential fertility. The share of non-Hispanic whites drops by more than 20 percentage points—from roughly 84 percent of Americans above age 65 to just 61 percent of Americans below age 30. The declining share of non-Hispanic whites from older to younger generations is primarily due to increases in the number of Hispanics and non-Hispanic blacks among younger cohorts. African Americans comprise only 8 percent of the elderly population, but over 13 percent of young adults and almost 15 percent of children below age 15. The proportion Asian almost doubles from a little over 2 percent of the elderly to over 4 percent of working-age Americans. American Indians have also increased their share from 0.4 percent of older Americans to just shy of 1 percent of young adults and children.
The largest generational shifts in Table 2 can be observed for Hispanics, who make up less than 5 percent of the population above age 65, but rise to the much higher level of 17 percent of young adults and children. The largest component of the Hispanic shift is Mexican Americans, who have risen from less than 50 percent to nearly 64 percent of all Hispanics. The much older Cuban American population has experienced a rapid decline in their share of the Hispanic and total populations across age groups, while Puerto Ricans have grown steadily, though at a slower pace than Mexicans. Another large and growing segment of the population is the “Other Hispanic” category. Nearly 3 in 10 Hispanics fall into this category, which includes descendants of the original Spanish settlers in the American Southwest as well as recent immigrants from Central and South America and the Spanish Caribbean. Among working-age adults, the number of Other Hispanics is roughly comparable to Asians in the United States. Among children below age 15, Other Hispanics greatly outnumber Asian Americans.
Immigration is the single most important factor behind the shifts in American diversity across generations (Edmonston and Passel 1994). This is illustrated in Table 3, which shows the percent foreign-born of each racial and ethnic group by major age groups.7 For nearly every racial and ethnic group, we see a common pattern of the percent foreign-born by age. The majority of immigrants arrive during their working years, so the 0−14 age group has very few foreign-born persons, even among the large immigrant populations. Of course, a significant share of the native-born are second-generation immigrants (the children of immigrants) who live as dependents in households with their immigrant parents.
For the three populations with the oldest historical roots in the United States—non-Hispanic whites, non-Hispanic blacks, and American Indians/Alaska Natives—immigration is relatively unimportant: less than 10 percent of these populations, in every age group, is foreign-born. The highest foreign-born figure is 9.3 percent for middle-age (30−44) blacks. On average, only 4 percent of non-Hispanic whites and 6 percent of non-Hispanic blacks are foreign-born.
For Hispanics, Asians, NHOPIs, and individuals who identify themselves as SOR or multiracial, immigrants are a substantial minority if not the majority. A little less than half (45 percent) of all Hispanics and over 70 percent of Asians are foreign-born. These figures are even higher for those in the working ages. About 3 out of 5 Hispanic Americans ages 30−65 are foreign-born, as are 8 of 10 working-age Asians.
The histories of the different Hispanic populations are reflected in the percent foreign-born by age group. For Mexicans, there is a majority native-born among the elderly, who comprise both the old stock of Mexican immigrants and portions of the long-resident Spanish-origin population in the American Southwest, whose presence predates the conquest and annexation of the region. By contrast, 63 percent of the prime working-age (30−44) Mexican population are foreign-born, reflecting the recent surge in immigration.
Cubans exhibit a different pattern, with the percent foreign-born rising with age. Among older Cubans, age 45 and above, over 95 percent are foreign-born. This pattern is consistent with the flight of Cuban refugees during the 1960s and 1970s, followed by a much more modest immigration stream since then. A substantial minority (about one-quarter) of Cuban Americans ages 30−44 are native-born, as are the majority of Cubans below age 30. Assuming the absence of a major new influx of Cuban immigrants, generational replacement will soon transform the largely foreign-born Cuban American community into a majority native-born population whose lives are less centered on the exodus from Cuba.
Similarly, the demographic center of gravity among Stateside Puerto Rican Americans has shifted from the island to the mainland. About 8 in 10 older Puerto Ricans (ages 45 and over) are island-born, compared to just 3 in 10 young adults (15−29). The remaining non-Hispanic populations (NHOPI, SOR, and multiple race) are mostly native-born, although immigration accounts for a significant share (20−35 percent) of these populations as well.
Even against the backdrop of an increasingly broad spectrum of racial and ethnic diversity, American race relations continue to pivot on the historical divide between white and black America. It was not just the twentieth century for which W.E.B. Du Bois famously noted that the color line would play a defining role (Du Bois 1999 : 5). The conflicts between blacks and whites have been a central issue throughout American history: the seventeenth-century project to equate African heritage with the mark of slavery, the eighteenth-century Declaration of Independence that left slavery intact, the nineteenth-century struggle to abolish slavery that led to the Civil War, and the twentieth-century Civil Rights Movement that demanded America live up to its ideals. Never has there been a point in three and a half centuries of American history in which the tensions and conflicts between blacks and whites have not been at or near center stage.
African Americans have always had a significant demographic presence in the United States. Shortly after the founding of the nation, one-fifth of the nearly 4 million persons enumerated in the first census were of African origin (Gibson and Jung 2005). More than two centuries later, there is remarkable demographic continuity. Of the 281 million Americans counted in 2000, over three-fourths identify themselves as white (including Hispanics) and about one-eighth identify themselves as black.
In Table 4, we compare the relative magnitude and characteristics of black and white Americans in two ways: those who report a single racial identity (labeled “white alone” or “black alone” in census terminology) and those who report a multiple-race identity (labeled “white in combination with other races” or “black in combination...” in census terminology). The sum of both groups is labeled here as “total whites” or “total blacks,” categories that include persons with any white or any black identity. Note that the sum of the total white and total black populations is a tally of responses that exceeds the number of persons because persons claiming both white and black identities are double counted.
The first two columns show the numbers of total whites and total blacks as percentages of the total US population. Some 77.1 percent of the American population claim to be white alone or in combination, and another 12.9 percent are black by the same inclusive definition. These figures include everyone who reports being only or partially white and only or partially black. Subsequent rows decompose the “total white” and “total black” populations by identity sub-type: those who report single versus those who report multiracial identities. Row three contains whites and blacks who report a single identity—labeled “white alone” and “black alone”—while row four includes whites and blacks who identify in combination with other groups (including one another). Subsequent rows break down specific mixed-race combinations.
What proportion of the total population is black or white? Counting only those with single-race identities, blacks and whites together account for more than 87 percent of the total American population. The number of responses by persons with any black or any white identity sums to about 90 percent, but this figure double counts persons with both black and white identities.
One of the common features of whites and blacks is their preference for a single-race identity. About 77 percent of Americans identify themselves as white, while 75 percent identify themselves as only white. Likewise, the percentage of Americans who report any black identity (12.9 percent) is just 0.7 percentage points larger than the percentage who report an exclusive black identity (12.2 percent). As shown in the subsequent panel (based on the universe of persons who report any white or black identity), blacks and whites are generally unwilling to acknowledge multiracial ancestry. More than 97 percent of those who claim any white ancestry claim only white ancestry, while the comparable figure for blacks is 95 percent. In other words, just a handful of whites and blacks report a multiracial identity.
The largest group of mixed-race whites, and the only combination that amounts to even 1 percent of the white total, is the White/SOR group. This largely Hispanic combination is 1.1 percent of all whites, followed by White/AIAN (0.6 percent), White/Asian (0.4 percent), and White/Black (0.4 percent). Among multiracial blacks, 2 percent report a shared white identity, with smaller shares reporting a part SOR, part American Indian, part Asian, or a combined white and American Indian identity.
The relative paucity of mixed-race reporting by whites and blacks is consistent with low levels of other forms of diversity—shown in the subsequent columns of Table 4. The shares of whites and blacks who identify themselves as Hispanic (3−9 percent) or foreign-born (7−8 percent) are fairly small, and are even smaller among the subsets who identify themselves as white or black alone. Among the whites and blacks who report multiple racial identities, however, we find much higher rates of Hispanicity and foreign birth. The peripheries of the white and black populations are quite small but very diverse.
The low levels of racial admixture reported by whites and blacks in the 2000 census represent an astounding loss of memory or at least a reluctance to acknowledge such memory in census responses. As noted earlier, researchers have estimated that about three-quarters of African Americans have some white ancestry, and a smaller but significant share of whites have some African ancestry (Hart 1921; Myrdal 1964 ; Shriver et al. 2003; Stuckert 1958, 1976). Persons of mixed black/white ancestry were considered to be an intermediate racial population in Charleston, New Orleans, and other Southern cities in the late nineteenth century (Davis 1991). The legacy of African ancestry among whites even entered into popular culture. The central theme of the very popular Jerome Kern and Oscar Hammerstein musical, Show Boat, is the stigmatization experienced by a woman of partial African ancestry who had passed as white (Breon 1995).
In early-twentieth-century censuses, the percentage of the population reported to be black and white (mulatto) declined (Frazier 1957: 185−187). In 1930, the Census Bureau dropped the mulatto category altogether. These patterns are consistent with the ideological triumph of the one-drop rule among most Americans—white and black. The division between blacks and whites was not merely ideological. For the first half of the twentieth century, discrimination was enforced by law and by extralegal violence in the South whenever racial hierarchies were challenged (Tolnay and Beck 1995). Intermarriage was illegal in many states. Black ghettoes were constructed in every large American city during the first half of the century (Massey and Denton 1993). In spite of the decline of de jure discrimination following the Civil Rights reforms of the 1960s, African Americans remain more highly segregated than any other minority in American society. Given the depth of the racial divide between whites and blacks, when multiracial ancestry returned to the census in 2000 (via the “check one or more” instruction on the race question), perhaps it is not surprising only 2 percent of blacks (and 0.4 percent of whites) reported overlapping ancestry.
Does the pattern of reporting Americanized racial identities extend to identities based on national or regional places of origin? Have whites and blacks lost the ties to their distinct geographic roots as they have adopted single-race Americanized identities? This question is addressed in Table 5 with responses to the Census 2000 ancestry question.
The census ancestry question, added to the long form in 1980, is open ended: individuals may report any ancestry or combination of ancestries they choose. The Census Bureau codes the exact responses for first and second write-in ancestries and publishes (or releases) tables with hundreds of detailed codes. Except for religious responses, no effort is made to mask (or correct) the reported ancestries of census respondents, and missing values are not imputed.
Originally hailed as a radical departure from traditional measures of nativity and lineage,8 the ancestry question was intended to provide richer and more comprehensive accounts of Americans’ diverse ethnic origins, particularly for the descendants of older immigrant waves whose descent was not captured in questions concerning nativity and parental nativity. As noted earlier, ancestry should, in theory, represent objective historical facts. Every American has ancestors who arrived from somewhere within the past several centuries, and with the exception of indigenous peoples and their descendants, all Americans have fairly recent lineage outside of North America.
In practice, however, the complexity of mixed ancestry, lost memories, and the subjective attractiveness or unattractiveness of different origins have yielded ancestry data of often dubious quality (Lieberson and Santi 1985; Alba 1990; Waters 1990, 1999; Nagel 1994). Some Americans see no need to report an ancestry, while others write in terms that are impossible to code. Farley (1991) finds that the population with English ancestry “grew” from 40 million to 50 million during the five-month window between the 1979 Current Population Survey and the 1980 census, only to plummet to 37 million by 1986. These fluctuations have led some analysts to question whether reported ancestral identities are of any value (Farley 1991).9
Ancestral reports of white and black Americans, however, provide revealing information about how these populations have come to see their identities in Americanized terms. Table 5 summarizes the first and second ancestry responses of whites and blacks in Census 2000,10 with detailed subgroups pooled within logical parent categories. Although the Bureau presents more than 200 ancestry categories (all those with 10,000 or more responses), we focus primarily on the distinction between Americanized or New World responses and geographic names of Old World countries or regions.
The most striking observation in Table 5 is that many whites and blacks no longer identify with a European or African country or specific region of origin. While a majority of whites do report a European ancestry of some kind, almost half report no ethnicity or give only a New World ancestry, including 9 percent who simply say “American.” To avoid stacking the deck in favor of our Americanization hypothesis, “European” or “Western European” responses are coded with specific European ancestries rather than Americanized racial identities such as “White” or “Caucasian.”
Many of the large European-origin groups reported, such as German, Irish, and Italian, reflect populations that are likely to have fairly recent ancestors—perhaps parents or grandparents—for whom the immigrant story has been heard first-hand. For many white Americans, however, ethnic identities have become primarily symbolic attachments to cuisine or expressions of the distinctive contributions of one's ancestral community (Gans 1979). Variations in European ancestry, which were strongly tied to socioeconomic status during the first half of the twentieth century, are no longer associated with educational or employment opportunities, residential patterns, or even marital choices (Duncan and Duncan 1968; Lieberson 1980; Alba and Nee 2003). Based upon in-depth interviews, Waters (1990) concludes that the contemporary ethnic claims of most whites are largely optional—symbolic, costless, or even trendy.
Although research based on the General Social Survey reports that 90 percent of whites claim ancestral origins from at least one European country (Hout and Goldstein 1994), the Census 2000 data suggest a more tenuous grip on European origins for most white Americans. Respondents interviewed in person may feel an obligation to name a specific ancestral country in response to a direct question. The much higher proportion who do not write in a response to the mail-back census questionnaire reveals a lack of deep ancestral attachments. Moreover, there appear to be strong secular declines in reports of some ancestries: the proportion of the total American population reporting English ancestry declined from 22 percent in 1980 to 13 percent in 1990 and to only 9 percent in 2000 (Brittingham and de la Cruz 2004: 4; Lieberson and Waters 1988: 34). Perhaps the lingering expressions of white ethnicity so attentively researched in the 1970s and 1980s represented the twilight, not the resurgence, of ethnicity (Alba 1990). Although more than half of white Americans still report a specific national origin, the journey from a deep remembrance of foreign roots to a generalized identity as white American is well underway (Jacobson 1998).
For black Americans, that journey may be nearing completion. Almost all African Americans have American roots that extend back to the seventeenth and eighteenth centuries. Except for a small minority of recent immigrants, ties to the African homeland have been lost to time. While the recent development of biogeographic markers has provided some African Americans with glimpses of their ancestral past, nearly 70 percent of blacks identified themselves simply as black or African American in the 2000 census. Under the weight of one of the most rigid systems of racial hierarchy in modern times, a system born in slavery, sustained by the legacy of the one-drop rule, and cemented with the passage of time, African Americans rarely acknowledge claims of ethnic heritage beyond race.
Although counts of persons reporting West Indian and African ancestries are nontrivial in absolute terms, the vast majority of the black population—nearly 90 percent—either report an African American or related ancestry (70 percent) or skip the ancestry question altogether (17 percent). Only about 8.3 percent of the black population report a West Indian or detailed African identity (even writing in “African” counts as an “ethnic” response under our coding scheme). More than one million blacks report various Caribbean ancestries in addition to 1.5 million who identify themselves with a continental African region (East African) or a specific country (Nigeria, Kenya, etc.). These reports probably reflect the recent waves of immigration to the United States, as well as the small number of African Americans who may have traced their ancestral lineage using genetic markers.
The dominance of Americanized identities is even more evident in the exact wording of the write-in responses. Almost two-thirds of blacks wrote in the specific term “African American,” a recently coined term that parallels other hyphenated American ethnic groups. First used in the 1980s and popularized by Jesse Jackson, the term “African American” symbolizes ancestral origins and also parity with white Americans (Baugh 1991; Fischer and Hout 2006: 34). Other terms were much less frequently mentioned. Less than 5 percent identify themselves as black, less than 2 percent as Afro-American, and fewer still as Negro or other terms that once commanded respect among black Americans but no longer do.
It is also of interest that almost no black Americans report any European ancestry. Just as very few white Americans acknowledge any African heritage, most black Americans do not consider their European ancestry to be noteworthy. In short, the results in Tables 4 and and55 show that blacks and whites share a limited and potentially diminishing interest in claiming identities beyond their race.
These trends are suggestive of an underlying process of increasing ethnic entropy—a generalized American identity with diminishing acknowledgment of ancestral complexity. Only about one-quarter of whites reported a second ancestry in 2000 and less than 2 percent of blacks. These figures are lower than in 1980 when the ancestry question was first asked in the census (Farley 1991). Having largely suppressed or lost the memories of their shared ancestry, blacks and whites are also well along the path of forgetting their ancestral places of origin in Africa and Europe. Skin color does matter, but beyond that, ancestral origins are no longer important for those whites and blacks who are far removed from the immigration experience.
The history of Asian settlement in the United States dates back to the mid-nineteenth century (Barringer, Gardner, and Levin 1993; Xie and Goyette 2005). Chinese immigrants first arrived in North America in substantial numbers in the 1850s. Nativist sentiment gradually arrested this process, however, and the influx of Chinese immigrants was essentially halted with the passage of the Chinese Exclusion Act of 1882 (Saxton 1971). Japanese began to arrive in the late nineteenth and early twentieth centuries following the bar to Chinese immigration. As with the Chinese, Japanese immigrants encountered discrimination and prejudice from white Americans, many of whom were immigrants themselves (Daniels 1977). In time, immigration from Asia was barred by federal policies, including international agreements, court orders, and restrictive legislation. With the passage of immigration laws during the 1920s, the national-origins quotas for Asians were set to zero. While the descendants of these early Asian immigrants remain an important part of Asian America, their numbers have been swamped by the much larger influx of Chinese, Filipinos, Asian Indians, Koreans, Vietnamese, and other national-origin groups following the liberalization of immigration in the 1960s (Min 2006).
As observed earlier, the major contemporary shift in racial and ethnic diversity in America is caused by increasing immigration from Asia and Latin America. Table 6 presents the ethnic composition and the prevalence of multiethnic and multiracial identities for the major Asian-origin groups in 2000. Table 7 presents comparable figures for Hispanic-origin populations.
OMB defines Asian peoples as those having origins in “the Far East, Southeast Asia, or the Indian subcontinent.” Although Asia encompasses dozens of countries and territories, and hundreds of ethnic groups, only six major Asian “races” or national-origin groups are listed as checkboxes on the census form: Asian Indian, Chinese, Filipino, Japanese, Korean, and Vietnamese (see Figure 1). There is also a checkbox for “Other Asian” where census respondents can write in a specific national origin. In Table 6, we include the six listed Asian categories plus the nine largest groups written in under “Other Asian.” We also include residual categories for respondents with different write-in responses (smaller Asian groups) and for those who did not write in a specific response.
The first panel in Table 6 shows the national-origin composition for several definitions of persons who identify themselves as Asian. The final column in the first panel, labeled “Total (any) Asian,” includes all 12.3 million persons who checked any Asian category. This total is composed of 10 million persons who reported a single Asian identity, 0.5 million who checked two or more Asian identities, and 1.8 million persons who reported both an Asian and a non-Asian identity. We refer to the first group as “Single Asian Alone” (only one Asian group identity), the second as “Multiethnic Asian” (e.g., Vietnamese and Chinese), and the third as “Multiracial Asian” (e.g., Korean and white). The second and third groups include persons who could be counted twice under different headings. For example, the sum of the tallies for different Asian groups in the second column exceeds the total number of multiethnic Asian persons.
The six largest Asian national-origin groups (those listed on the census form) account for 87 percent of the 12.3 million Total (any) Asians enumerated in 2000. One in four Asian Americans is Chinese (including Taiwanese). Filipinos and Asian Indians comprise another 19 percent and 16 percent, respectively. The other three major Asian populations—Koreans, Vietnamese, and Japanese—comprise about 10 percent each of the Total Asian population.
A diverse range of other Asian populations is represented among the write-in responses: Cambodian, Hmong, Laotian, Pakistani, Thai, Bangladeshi, Indonesian, Sri Lankan, Malaysian, and a catch-all category for smaller Asian groups. None of these groups comprises more than 2 percent of the Total Asian population, and most are much smaller. The vast majority of Asians checked a listed category or wrote in a specific Asian-origin population, but 3 percent simply claimed a panethnic Asian American identity and chose not to identify a specific Asian country or region of origin.
Among the some 450,000 multiethnic Asians, one in three is part-Chinese. The Chinese diaspora in many Asian countries has intermarried with other national-origin populations and is well represented in several immigrant streams from Southeast Asia, especially from the Philippines and Vietnam. We suspect that the Chinese/Japanese multiethnic population is a product of intermarriage among long-resident Asian populations in the United States. A significant share of Asian Americans (about 1.8 million, or 15 percent of Total Asians) reports that they are multiracial, primarily white and Asian.
The panels of Table 6 show the specific characteristics of each national-origin Asian population. The middle panel shows the prevalence of multiethnic and multiracial identities for each group (with percentages summed across rows). The last panel shows two additional characteristics of each national-origin group: the percent foreign-born and the percent Hispanic.
The rates of mixed ancestry among Asian Americans are higher than those reported by whites and blacks. Only 82 percent of Asians report themselves to be of a single national origin. More than 90 percent of all Vietnamese are only Vietnamese. This figure drops to the mid-80 percent range for most large Asian groups and even lower for Filipinos and Japanese. Reports of multiple Asian nationalities (i.e., Multiethnic Asian) are relatively low, only about 5 percent of most groups. The much higher report of 23 percent multiethnic composition among the small Malaysian American population undoubtedly reflects the tendency of many Malaysian Chinese to report their ethnicity (Chinese) and their country of origin (Malaysia). Many more Asians report multiple races (more than one OMB race category). Most multiracial Asians report having an Asian and a white identity.
Roughly 25 percent of Japanese Americans report multiracial ancestry. This high level of racial blending is due, at least in part, to the fact that most Japanese Americans are descendants of immigrants who arrived in the early twentieth century. Japanese Americans encountered widespread discrimination for much of the century, including internment in detention camps during the US participation in World War II. But in recent decades, Japanese Americans have become economically and spatially integrated with whites, including high rates of intermarriage (Espirtu 1992; Fu 2001; Xie and Goyette 2005).
Some 70 percent of Asian Americans in 2000 are foreign-born. The exception to this pattern is Japanese Americans—some 58 percent of Japanese Americans are native-born, mostly the grandchildren and great grandchildren of the immigrant wave of the early twentieth century. When the door to renewed immigration was opened in the late 1960s, Japan was a highly developed country with few factors encouraging emigration. The modest migration stream from Japan at present is more akin to the circulation of highly skilled professionals and business managers among industrial countries than the immigration influx from other Asian countries.
A small minority of Asians, about 370,000, checked the Asian American category but did not report a specific national origin. This Americanized panethnic population is largely native-born (55 percent) and has an unusually high proportion (55 percent) reporting multiracial ancestry. This group also has higher overlap with Hispanics (5 percent) than do other Asians (0.9 percent). Assuming continued high levels of intermarriage, the patterns observed for the Japanese and the emerging “Asian American” population might be a harbinger for other Asian groups. Filipinos, who account for 1 of every 5 Asians in the United States, are only slightly less multiracial (19 percent) than the Japanese (26 percent). Among the six major Asian populations, only the Vietnamese are less than 10 percent multiracial.
Hispanics share many parallels with Asian Americans. Immigration for both groups is the main force behind their rapid population growth in recent decades. In addition, national origins, rather than American racial categories, are the primary source of ethnic identity for Hispanics. As shown in Figure 1, the census question on Hispanic origin lists Mexican, Puerto Rican, and Cuban as separate checkboxes. There is also a space for respondents to check “Other Spanish/Hispanic/Latino” and to write in their specific Hispanic identity. Table 7 includes the three listed Hispanic groups and the 20 largest write-in groups, organized in terms of region: Central America, South America, and Other Hispanic.
The total number for each Hispanic population is reported in the first column in Table 7. The second and third columns show each Hispanic group as a percentage of the national US population and of the Total (any) Hispanic population. The census Hispanic-origin question does not allow multiple responses, so respondents are forced to select only one of the mutually exclusive categories. Because nativity is a defining characteristic of the Hispanic population, the next two columns show the national-origin composition of foreign-born and native-born Hispanics. The last two columns show two additional characteristics of each Hispanic-origin population: the percent foreign-born and the percent reporting “Some Other Race” in response to the census race question. These attributes are row percentages.
In 2000, nearly 13 percent of Americans (35.2 million) claimed Hispanic ancestry.11 With more than 20 million persons (7.4 percent of the US total), the Mexican-origin population outnumbers the sum of all nonblack minority groups (Asian, AIAN, and NHOPI) combined. The Mexican-origin population also outnumbers every white ancestry group except for German Americans (see Table 5).
Nearly three-fourths of all Hispanics identify themselves with the three major Hispanic communities: 60 percent as Mexican, 10 percent as Puerto Rican, and 4 percent as Cuban. The balance (27 percent of all Hispanics) represents a diverse group of national- and regional-origin populations. Some 5 percent have origins in various Central American countries, and another 4 percent have South American origins. The largest groups are Salvadorans (705,000) and Colombians (491,000). The largest group among other Hispanics are Dominicans (815,000). The only non–Latin American population listed under Hispanics is Spaniards, who number just 114,000.
Each of these Central and South American origin groups, however, is dwarfed by the 5.5 million Hispanics who claim what might be considered a new “Americanized” identity. Some 16 percent of Hispanics identify themselves simply as “Hispanic/Latino” and do not write in a specific national origin (listed in Table 7 as “Other Hispanic/Latino”). Second only to Mexicans in size, this large, panethnic group outnumbers Latinos from every South and Central American nation combined, and unlike most Latino-origin groups, panethnic Hispanics are predominantly (70 percent) native-born. While just 1 in 10 foreign-born Hispanics fails to list a specific Hispanic nationality, one-fifth of native-born Hispanics identify themselves simply as “Hispanic/Latino.” This group surely encompasses a diverse range of peoples.
Indeed, the term “Hispanic” applies to both the descendants of older immigrant flows and persons of Spanish origin whose ancestors lived in the Southwest (and parts of Mexico) long before they were annexed by the United States in the mid-nineteenth century, as well as the largest group of new immigrants (Bean and Tienda 1987; Tienda and Mitchell 2006). Although a majority of Hispanics, 55 percent, are native-born, this average masks the wide variation among the peoples of Hispanic origin. Regardless of birthplace, all Puerto Ricans are US citizens; moreover, about 3 in 5 are mainland-born. Even though the majority of Mexican Americans are native-born, the Mexican-born population is the largest component of foreign-born Hispanics (56 percent). On the other hand, Cubans, Dominicans, and those who report specific Central and South American national origins are predominantly—upwards of 70 percent—foreign-born. As noted earlier, the 5.5 million panethnic “Other Hispanics” are overwhelmingly native-born.
As evidenced by their responses to the race question, many Hispanics do not identify themselves with the standard categories of American racial statistics and see little need to report an identity beyond their Hispanic/Latino origin. Despite being given explicit instructions to answer both questions, many Hispanics either leave the race question blank or reiterate their Hispanic identity by writing in Mexican or some other Latin American country on the “Some Other Race” line (Logan 2004; Perez 2008; Rodriguez 2000). In 2000, nearly half of all Hispanics (48 percent) supplied an SOR response (alone or in combination) to the race question.
Of the three major groups, only Cubans show a small minority (1 in 10) selecting SOR (the majority chose white). Of the Central and South American groups, the latter are generally less likely to identify as SOR (38 percent on average). Statistics not shown in Table 7 indicate that most South American groups opt for white identity about 60 percent of the time, particularly Argentineans (81 percent), Chileans (70 percent), Uruguayans (81 percent), and Venezuelans (71 percent). Only trivial numbers of Hispanics select any racial identity other than SOR or white. Less than 2 percent of Hispanics claim a black racial identity, and with the exception of Panamanians (26 percent black), no group has a double-digit percentage of blacks.
If one considers the large proportion of “Other Hispanic/Latino/Spanish” responses to be an emerging New World identity, this could indicate signs of the Americanization of ethnicity among Hispanics (Itzigsohn 2004). Further, a significant number of persons with partial Hispanic ancestry may no longer claim Hispanic identity (Duncan and Trejo 2007, 2008). Among those who do identify themselves as Hispanic, their rejection of American racial categories in nearly half the cases poses a major challenge to the present system of classification. One interpretation is that Hispanic immigrants are simply unfamiliar with American racial typologies. Another possibility is that Hispanics perceive that the American racial system creates hierarchy and stigmatization, and they choose to ground their identities in national origins and cultures (Landale and Oropesa 2002).
American Indians and Alaska Natives (AIAN) and Native Hawaiians and Other Pacific Islanders (NHOPI) are the indigenous peoples of North America and the Pacific Islands, but at present they are at the demographic margins of the American population. Persons who identify as AIAN and NHOPI alone (reporting only a single race) together account for just one percent of the American population. If all AIAN and NHOPI respondents (including those with partial identities—AIAN and NHOPI in combination with other races) are counted, they comprise almost 2 percent of the total population. Because of space limitations and the complexity of their ancestry, our analysis of the racial and ethnic identities of AIAN and NHOPI populations is presented in a companion paper (available on request from the authors).
By the close of the nineteenth century, many Native American tribes had been driven to the brink of extinction by conquest, forced migration, and genocidal policies (Thornton 1987; Snipp 1989). A slow demographic revival among indigenous peoples has accelerated over the last half-century. In the 2000 census, 2.5 million Americans reported their race as AIAN alone, and an additional 1.7 million persons identified themselves as AIAN in combination with other race(s). This dramatic comeback, in part the result of natural increase, is primarily attributable to shifts in identities among persons with partial American Indian ancestry (Harris 1994; Eschbach 1993, 1995; Ogunwole 2006). “AIAN Alone” persons are more likely to report a specific tribal identity (75 percent) than are multiracial AIANS (63 percent), but about 30 percent of all Native Americans simply report a panethnic American Indian identity.
Similar to AIAN populations, NHOPIs are a diverse people who are highly blended with other groups—over 50 percent of NHOPIs are of mixed racial ancestry. The official OMB definition of NHOPI refers to persons “having origins in any of the original peoples of Hawaii, Guam, Samoa, or other Pacific Islands.” Native Hawaiians are the largest NHOPI population: about 45 percent of NHOPIs (alone or in combination with other races) identify themselves as Native Hawaiians. The other three major Pacific Islander populations are 1) Samoan, Tongan, and Other Polynesian, 2) Guamanian/Chamorro and other Micronesian, and 3) Fijian and other Melanesian. About 1 in 5 persons checked the Pacific Islander racial category but did not write in a specific ancestral population.
The disparate and at times jumbled patterns of racial and ethnic reporting in the 2000 US census result from a conceptual quagmire. The heart of the issue is a sharp divergence between the geographic definition of ancestry envisioned by the framers of the current classification system, and the de facto measures of race and ethnicity based on subjective, socially constructed identities. The expansion of the census race question to include many national-origin groups from Asia, but none from Europe, is an artifact of ideology and politics. So too is the use of a separate question on Hispanic origin, which has led to the creation of a quasi-racial Hispanic group under the SOR (“some other race”) heading. The addition of the ancestry question in 1980 was motivated, at least in part, by the political preferences of white Americans who wanted to measure their ethnic roots. The impact of politics on the American census, including the measurement of race and ethnicity, is not a recent phenomenon. From the first census in 1790 to the present, political forces have been closely intertwined with census taking (Anderson 1988).
However, the observations that the racial and ethnic categories in the census are shaped by political considerations and that the responses are not always accurate measures of ancestral origins do not mean that the data are without value. Racial and ethnic responses provide a good reading of contemporary social and political identities, even if the mix of influences stemming from ancestral origins and other factors may vary. Examination of racial and ethnic identities by age, birthplace, and other characteristics allows us to detect systematic patterns of how social divisions are constructed and perceived.
For example, the separate measurement of race and Hispanicity has created an artificial “some other race” category that is a proxy for about half of all Hispanics. Moreover, Hispanics who both checked “white” and wrote in a Hispanic identifier under “some other race” boosted the “two or more races” population by a sizable share. Both measurement problems are minimized by treating Hispanicity on par with the standard OMB racial categories. Our revised summary classification, presented in Tables 2 and and3,3, has the added virtue of consistency with the perceptions of many Hispanics who write an Hispanic identity (e.g. Mexican) in response to the census race question (Logan 2004; Perez 2008).
Immigration and generational succession are the primary forces changing the ethnic composition of the American population. The influx of newcomers from Latin America and Asia in recent decades, combined with modest differences in natural increase, has led to a much lower proportion of non-Hispanic whites, particularly among young adults and children. Not all diversity, however, is of foreign origins. Native Americans, and to a lesser extent African Americans, have the longest history of residence in North America. Moreover, many Mexican Americans and most of the panethnic Hispanic population have long historical roots in the country. The majority of Puerto Ricans residing in the 50 states are mainland-born. Unless there is a new immigration wave from Cuba, generational succession will lead to a predominately native-born Cuban American population in the coming decades.
The addition of the census ancestry question in 1980, conceived as an effort to highlight white ancestral origins, has in large part chronicled the opposite—partly in the form of blended ancestries, and more recently through the apparent and ongoing disappearance of ethnic consciousness from large segments of the American population. While these patterns may be interpreted as a failure of measurement, we suggest that the evidence is consistent with a model of Americanized panethnic identities that minimizes ancestral diversity.
The Americanization of racial and ethnic origins is evident for all long-resident groups. With only a small fraction of recent immigrants from Africa and the Caribbean, and ancestral roots dating back over 200 years, the geographic origins of most African Americans are not part of the popular consciousness. Most African Americans simply reiterate their Americanized racial identity to the open-ended ancestry question. The popularity of the recently coined “African American,” and the virtual disappearance of other terms among ancestry responses, including the once respectable “Negro,” reinforce the conclusion that identities reflect social affinities and political aspirations as much as historical roots.
European national origins are still common among whites—almost 3 of 5 whites name one or more European countries in response to the ancestry question. The ancestral roots of the grandchildren or great-grandchildren of immigrants are still part of living memories. However, a significant share of whites respond that they are simply “American” or leave the ancestry question blank on their census forms. Ethnicity is receding from the consciousness of many white Americans. Because national origins do not count for very much in contemporary America, many whites are content with a simplified Americanized racial identity. The loss of specific ancestral attachments among many white Americans also results from high patterns of intermarriage and ethnic blending among whites of different European stocks.
Ethnic blending between whites, blacks, and American Indians occurred in the centuries before and after the founding of the United States. Yet the racial responses to Census 2000 provide almost no acknowledgment of a substantial extent of shared ancestry between black and white Americans. Although there are no firm data on the “real” prevalence of multiracial descent, the creation of separate white and black populations is a product of political history, not strict genealogy. Only a powerful racial ideology, exemplified by the one-drop rule and backed by state-sanctioned discrimination, made it possible to create the fiction of separate origins.
The decline of state sanction and enforcement of policies based on the one-drop rule, moreover, did little to expose the myth of the “color line” or lay to rest any claims of racial “purity” among whites or blacks. Undoubtedly this pattern persists in part because biologically discrete races are not, and have never been, a precondition for racial prejudice or inequality. As a result, even as persons of mixed descent are increasingly represented in popular culture, politics, and advertising campaigns, there are few signs that these changes have led to a shift in the patterns of reporting on race and ethnicity among blacks and whites.
Levels of reported multiracial ancestry are much higher among AIANs, NHOPIs, Asians, and Latinos than among whites and blacks. While this difference reflects contemporary patterns of intermarriage, deeper forces are undoubtedly at work. Historically, antipathy and bigotry toward all American minorities were commonplace, including the savage wars to wrest control of the lands of American Indians, the campaigns to deny citizenship and other basic rights to Asians, and the economic and political exploitation of the peoples of Mexican origin in Texas and elsewhere. As these patterns have receded in recent decades, there have been increases in residential integration, economic mobility, and intermarriage among many American Indians, Asians, and Latinos with whites. There has been much less integration and intermarriage of blacks with whites.
As noted above, immigration from Latin America and Asia is the major force changing the racial and ethnic composition of the American population. Two patterns are evident from Census 2000. The first is that most Hispanic and Asian Americans identify themselves with specific national origins. This pattern is facilitated by the inclusion of Asian nationalities (Chinese, Japanese, Filipino, etc.) as explicit categories in the race question and by similar categories (Mexican, Cuban, and Puerto Rican) in the Hispanic-origin question. Both questions also include a write-in option allowing other national and regional groups to specify their exact origins. This pattern of response is reinforced by the high percentage of first- and second-generation immigrants among Asian and Hispanic Americans, for whom there are active memories of a specific homeland. In some ways, these homeland attachments are similar to those of Eastern and Southern Europeans a century ago. The descendants of these immigrants eventually blended into the mainstream white population, but this was not the outcome expected or desired by old-stock Americans who erected the immigration barriers in the 1920s to keep out undesirable Eastern and Southern Europeans (Higham 1988).
While national origins remain the primary mode of identification, there are tentative signs of emerging Americanized identities among Hispanics and Asians. The panethnic Hispanic population is the second largest Hispanic group and is largely native-born. The panethnic Asian American respondents are a proportionately smaller fraction of all Asians, but this probably reflects the recent timing of the immigration stream. Persons who claim to be simply Asian American are disproportionately native-born and of multiracial origin.
High levels of intermarriage among both Asians and Latinos, especially in the second generation, weaken ethnic ties and attachments (Stevens and Tyler 2002; Duncan and Trejo 2007, 2008). In most US censuses and surveys, an adult respondent, typically a parent, reports on the race and ancestry of children in the household. There is wide variation in the response patterns, with some evidence of a slight preference among some groups for assigning a white identity to children with one Asian parent and one white parent (Saenz, Hwang, and Anderson 1995; Xie and Goyette 1997; Waters 2002). The 2000 census allowed intermarried parents to report multiple identities for their mixed-race children, but as multiracial children grow to adulthood, they are likely to choose less complex identities than those reported by their families of origin. The choice of a marriage partner and the community of residence may have decisive influences on the identity choices of persons of mixed ancestry. They might well gravitate to one community or the other, or perhaps simply grow less interested in recording the complexity of their ancestry.
Racial and ethnic identities can shift, especially with the very high levels of intermarriage of Hispanics and Asians with whites. In addition to immigration, the future race and ethnic composition of the United States will be shaped by ethnic blending and patterns of identity choices. New categories may emerge for persons of mixed ancestry. The metaphor of the “melting pot,” the idea that intermarriage leads to the blending (or blending away) of distinct ancestries across generations, is as old as America. The concept was expressed in the eighteenth century by Jean de Crèvecoeur in his Letters from an American Farmer (1782), when he responded to the rhetorical question “What then is the American, this new man?” by observing that “He is an American, who, leaving behind him all his ancient prejudices and manners, receives new ones from the new mode of life he has embraced.... Here individuals of all nations are melted into a new race of men” (Alba and Nee 2003: 17−18). In the early twentieth century, the protagonist in Israel Zangwill's play, The Melting Pot, declares “A fig for your feuds and vendettas! Germans and Frenchmen, Irishmen and Englishmen, Jews and Russians—into the Crucible with you all! God is making the American.” These statements were formulated not so much as a description of a social process or even an inevitable trend, but rather as an ideological claim that the United States is (or should be) different from the Old World in which pedigrees of nation, language, and religion divided the population into separate societies.
The melting pot metaphor is not always viewed in such optimistic terms. Some see the concept as fanciful or incomplete, while others view the melting pot as a self-serving smoke screen for the stratified nature of American society. The persistence of racial and ethnic segregation and inequality belies any claims that race and ethnicity have disappeared into some crucible of American idealism. Yet considerable evidence suggests that the divisions between groups of European ancestry have been eroded through intermarriage over the course of the twentieth century (Alba and Nee 2003). The highly blended ancestries reported by many American Indians, Hawaiians, and Pacific Islanders are also more in line with the melting pot metaphor than with the one-drop ideology that divides white and black America.
In the current era of high immigration, the emergence of Americanized and admixed identities may be deferred because new arrivals outnumber the second and third generations for some groups. Yet it seems inevitable that ancestral ties will weaken with the passage of time. One possible outcome of this process is the emergence of panethnic Americanized categories of white, African American, American Indian, Asian American, Hawaiian/Pacific Islander, and Hispanic American. This pattern would be most likely if political, social, and economic sanctions or incentives reinforce group solidarity. But, should ethnic or racial identities lose their political and economic salience, or if intermarriage should continue to blur group boundaries, the reporting of ancestral roots will likely become more flexible, symbolic, and situational. Just as the forces of increasing ethnic entropy have led to the disappearance of ancestral roots within Americanized panethnic groups, the boundaries between and within other groups may also begin to erode.
An earlier version of this article was presented at the 2007 Annual Meeting of the Population Association of America in New York City. Contact Anthony Daniel Perez at ude.notgnihsaw.u@zerepda or Charles Hirschman at ude.notgnihsaw.u@selrahc. This research has been supported by a grant from the National Institute of Child Health and Human Development (HD47289). We are indebted to Matthew Snipp for his careful reading and comments on an earlier draft. We also gratefully acknowledge the support and helpful comments of Gunnar Almgren, Patty Glynn, Julie Miller, Geoffrey B. Palmer, Nikolas D. Pharris-Ciurej, Lynne Taguchi, and Irina Voloshin.
2A potential influence might have been the suggestion by Michael Novak, a prominent American neoconservative who wrote the influential 1972 book Rise of the Unmeltable Ethnics, at a White House seminar in 1976 that the Census Bureau collect data on the ethnic origins and identification of third- and later-generation Americans (Price 1980: 1033).
4IPUMS, or Integrated Public Use Microdata Samples, is an archive of individual-level records (stripped of personally identifying information) from a nationally representative sample of persons enumerated in decennial censuses from 1850 to 2000 (and from the American Community Survey for recent years). These data are produced and distributed by the Minnesota Population Center (Ruggles et al. 2008). All of the tables in this article are based on our analysis of data from the IPUMS archive, but we occasionally reference census data from other sources such as the Census 2000 summary files.
5Some forms of DNA (e.g., mitochondrial) are passed on without recombination and can be used to identify shared ancestry, although these genes are not reflected in a person's phenotype.
6Mixed-race categories such as mulatto and “part Hawaiian” were used sporadically in earlier periods, but 2000 was the first year in which respondents were permitted to check multiple races.
7The standard census definition of the native-born population includes all persons who are American citizens at birth; hence the children of American citizens born abroad and in American territories are considered native-born. In Table 3, we count everyone born outside the 50 states and the District of Columbia as foreign-born, even those who were American citizens already at birth, because they have had a migration-like experience when moving from Puerto Rico or American Samoa.
8Earlier censuses included questions on mother's and father's country of birth.
9The original questionnaire instructions emphasize the subjective reading by instructing respondents to “print the ancestry with which [they] identify” (italics in original). Although this emphasis was removed in 1990 and later censuses, it is unclear whether responses reflect objective information about ancestral origins or contemporary identity preferences, a confusion echoed by the census's vague and deliberately broad definition of the term (US Census Bureau 2004).
10Multiracial respondents are excluded, as they account for negligible shares of the total black and white populations, as shown in Table 4.
11As of July 2007, Hispanics had grown to some 45.5 million persons and made up more than 15 percent of the US population (US Census Bureau 2008a). These intercensal estimates are nearly identical to those from the 2007 American Community Survey (45.4 million Hispanics) and indicate both the rapid growth (roughly 4 percent per year) of the Hispanic population and a decisive shift in the composition of racial/ethnic minorities in recent years.