Who Is Multiracial? Assessing the Complexity of Lived Race Author(s): David R.

Harris and Jeremiah Joseph Sim Source: American Sociological Review, Vol. 67, No. 4 (Aug., 2002), pp. 614-627 Published by: American Sociological Association Stable URL: http://www.jstor.org/stable/3088948 Accessed: 08/11/2009 12:41
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=asa. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

American Sociological Association is collaborating with JSTOR to digitize, preserve and extend access to American Sociological Review.

http://www.jstor.org

WHO IS MULTIRACIAL?
ASSESSING THE COMPLEXITY OF LIVED RACE
DAVID R. HARRIS

Universityof Michigan

JEREMIAH JOSEPH SIM Universityof Michigan

Patternsof racial classification in the National LongitudinalStudyof Adolescent Health are examined.The survey'slarge sample size and multipleindicatorsof race permit generalizableclaims aboutpatterns and processes of social constructionin the racial categorizationof adolescents.About 12 percent of youthprovide inconsistent responsesto nearly identical questionsabout race, contextaffects one's choice of a single-race identity,and nearly all patterns and processes of racial classification dependon which racial groups are involved. The implicationsof thefindings are discussedfor users of data on race in general, andfor the new census data in particular.

he 2000 census marked a fundamental change in how we measure race in the United States. Rather than insist that every person identify with only one racial group, as all previous U.S. censuses have done, the 2000 census instructed household informants to "mark one or more races to indicate what this person considers himself/herself to be" (Nobles 2000). This small change in the wording of the race question will have important consequences for civil rights enforcement, social science research, and racial politics, as policies and procedures that implicitly assume the mutual exclusivity of racial groups seek to accommodate data that impose no such constraints. One of the first indications that our official racial classification system had changed occurred in March 2001 when the Census Bureau announcedthat 2.4 percent of Americans identified with two or more racial groups in the 2000 census. This announce426 to Directcorrespondence DavidR. Harris, ThompsonStreet,Universityof Michigan,Ann Arbor,MI 48106-1248 (drharris@umich. edu). We wish to thankYu Xie, PamelaSmock,Julia and of Malamud, members the Adams,Deborah and MacArthur FoundationSocial Interactions Economic Research Network helpfor Inequality ful comments earlierdraftsof this paper. on 614

ment received much attention (e.g., Schmitt 2001) and was interpretedby some as meaning that race is now less rigid and consequential than it was even a few decades ago (e.g., Clegg 2001). Less attention has focused on the possibility that, because race is a social construct, the 2000 census might well have yielded a different estimate of the size of the multiracial population had it selected an alternative, but equally plausible, measure. We address these concerns by examining the fluidity of race for a nationally representative sample of adolescents. Specifically, we (1) contrast counts of multiracial youth obtained from common schemes of racial classification, (2) describe individual-level patterns of consistency in racial reporting, and (3) examine how youth who self-identify as multiracial answer questions that insist upon single-race responses. Throughout, we allow for subgroup differences by conducting separate analyses for the three largest multiracial groups-white/black, white/ Asian, and white/American Indian.1 1 We focus on white/nonwhite multiracial populationsbecause they representthe largest multiracial "A/ groups(Suro1999).Throughout, B" refersto people who identifywith groupsA andB.
VOL. 67 (AUGUST:614-627)

AMERICAN SOCIOLOGICALREVIEW, 2002,

WHO IS MULTIRACIAL?

615

BACKGROUND During the nineteenth and early twentieth centuries, American racial ideology was dominated by the perspective that racial groups were biologically determinedand imbued with distinctive physical, mental, and moral abilities (Banton 1983; Gould 1996; Nobles 2000; Spickard 1992). This essentialist perspective on race later came under attack as growing numbers of social and physical scientists argued that there never were pure races, that on almost all traits there is greatervariation within racial groups than between them, and that the boundaries of racial groups vary both over time and across social contexts (Davis 1991; Espiritu 1992; Goodman 2000; Gould 1996; Nagel 1994; Spickard 1992). In arguing that there is no objective, biological basis for defining racial groups, advocates of this social constructionistperspective on race maintainthat the function of race is to reinforce and perpetuate social differences (Blauner 1972; Omi 2001). As Spickard (1992) observes, "The process of racial labeling starts with geography, culture, and family ties and runs through economics and politics to biology, and not the other way around"(p. 16). In addition to arguing that racial group boundaries are subjective, social constructionists also maintain that racial identities are fluid (Hahn, Mulinare, and Teutsch 1992; Nagel 1995; Snipp 1997). Echoing research on ethnic identity (Alba 1990; Lieberson and Waters 1993; Waters 1990, 1999), they argue that racial classifications can differ not only among nations and historical periods, but also in the day-to-day lives of individuals. From this perspective, each individual can be seen as having multiple context-specific racial identities. Racial identities can be distinguished along several dimensions. What an individual believes about his or her own race is an internal racial identity. Observers'beliefs about an individual are external racial identities. Moreover, there are expressed racial identities-words and actions that convey beliefs about an individual's race. Identities can be expressed from an individual to an observer, from an observer to an individual, or between observers. Although these identities need not be identical, they are not in-

dependent of one another. This point was long ago recognized by Cooley (1902), and has recently been revived by Nagel (1994), who observes that expressed ethnic identity "is the result of a dialectical process involving internal and external opinions and processes" (p. 154). As race is socially constructed, the outcome of this dialectical process can vary across contexts. In different settings, traits such as phenotype, ancestry, and culture are differently and differentially privileged as criteria for identifying one's race. Thus, as social composition, racial ideology, and knowledge about an individual can vary across contexts, there is the potential for variation in internal, external, and expressed identities, and their relationships to one another. Despite volumes of qualitative and anecdotal evidence attesting to the fluidity of racial identities (Davis 1991; G. Williams 1995), most empirical studies of race fail to take this point seriously. Rarely are multiple observations of an individual's race obtained, and the basis and perspective of racial classifications are rarely made explicit. The implications of this oversimplification of measured race are likely to be especially important for multiracial populations because of the relatively recent establishment of a socially recognized multiracial identity in the United States (Nobles 2000; Office of Management and Budget [OMB] 1997). As America slowly acknowledges that being a member of two or more racial groups is as legitimate as being a member of any single racial group, confusion arises about who is multiracial. In research, the multiracial population has variously been defined according to how people identify themselves, how they are identified by others, how their ancestors were identified, and combinations of these criteria (Corrin and Cook 1999; Davis 1991; Goldstein and Morning 2000; Hirschman, Alba, and Farley 2000; Kao 1999; Tafoya 2000; Twine 1997; Xie and Goyette 1998). However, if equally plausible racial classification schemes produce disparate estimates of the size and characteristicsof the multiracial population (Goldstein and Morning 2000; Harris forthcoming; Hirschman et al. 2000) and analysts are not aware of or sensitive to these effects, our ability to under-

6 6

AMERICAN SOCIOLOGICAL

REVIEW

Table 1. Selected Measures of Race in the National Longitudinal Study of Adolescent Health Measure School race Source of Data Interviewat school Description "Whatis your race? If you are of more thanone race, you may choose more thanone." Optionsare white, black or AfricanAmerican,Asian or Pacific Islander, AmericanIndianor Native American,and other." "Whatis your race? You may give more thanone answer."Same responsecategories as for school race. If more thanone race selected for home race: "Which one categorybest describesyour racial background?" A primarycaregiveris asked, "Whatis your race? You may give more thanone answer."Next, he/she is asked "Whatis the race of your current(spouse/partner)? You may give more thanone answer."Again, the aforementioned five race categories were the valid responses.

Home race Best single race

Interviewat home Interviewat home

Parent-based race Interviewwith parent/ caregiver

stand the burgeoning multiracial population will be severely compromised. Moreover, because individuals who are defined as multiracial in some schemes appear as monoracial in others, fluidity in multiracial populations can affect our understanding of single-race populations. Unfortunately,there have been few empirical assessments of the fluidity of race in the U.S. context (Goldstein and Morning 2000; Hahn et al. 1992). We know that one's racial classification can vary across contexts and observers, but we know little about the magnitude and patterns of racial fluidity in the United States, and even less about the circumstances that facilitate these shifts. This gap between the theoretical and empirical literaturesis especially troublingbecause of the Office of Managementand Budget's mandate that, by 2003, all federal data collection efforts must allow respondentsto identify with two or more racial groups (OMB 1997). This decision will produce volumes of data that identify multiracials based on a single racial classification scheme, but there is little understandingof how well race can be captured by any single measure (Harrison forthcoming; Root 1992). Here we examine this issue using data for a nationally representative sample of white/black, white/American Indian, and white/Asian youth. DATA AND METHODS We examine data from Wave 1 of the National Longitudinal Study of Adolescent

Health (Add Health). Add Health is a school-based, longitudinal study of health behaviors for youth in grades 7 through 12. Wave 1 was collected in 1994 and 1995. Initially, 83,135 in-school interviews were conducted with students from 80 high schools and 52 middle schools. Next, in-home interviews were conducted with 18,924 youth from the school sample.2 Last, in-home interviews were conducted with a primary caregiver of each of the youth interviewed at home. In over 70 percent of cases, the interviewed primarycaregiver was the youth's biological mother.3 For our purposes, the key features of Add Health are its large sample size and multiple measures of racial identity. The presence of multiple indicators of race distinguishes Add Health from most other surveys (Table 1). First, Add Health collects self-reported race data. However, rather than collect only one reportper person, Add Health allows adolescents to identify with up to five racial groups in each of two interviews. If more than one racial group is selected during the home interview, youth are then given the opportunity to indicate which single race best describes them. Next, Add Health collects data on
the survey commands in Stata to correct for oversamples in the home sample, clustering in the school sample, and survey nonresponse

2 Ouranalysesuse weighteddata,and we use

and (Chantala Tabor1999).

Add Health see Bearman, Jones, and Udry (1997).

3 For more information about the design of

WHO IS MULTIRACIAL?

617

coresidential biological parents that can be used to construct a limited measure of ancestral race. Taken together, these four measures-school race, home race, best single race, and parent-based race-provide rare, generalizable information about the fluidity of racial identity. The first three indicate expressed internal race, while the parent-based measure is a combination of parents' expressed races. In addition to collecting multiple reports on race, Add Health is also rare in that it supports an assessment of contextual effects on expressed internal race. School and home race items are nearly identical, but whereas the school survey is self-administered, the home survey is administered by an interviewer and is often observed by family members. As a result, youth enjoy greater anonymity when reporting their race at school than they do at home. Consistent with Cooley (1902) and Nagel (1994), we expect that home race will therefore be more reflective of the racial conceptions of older generations, while school race will gravitate toward contemporaryideals of multiraciality. Our analysis imposes two restrictions on the data. First, we exclude all youth who identify as Hispanic. This is necessary because Add Health follows the convention of asking separate questions about race and Hispanic origin. In treating Hispanicity as distinct from race, Add Health deviates from conventional academic uses of race that tend to contrast Hispanics with non-Hispanic whites, blacks, and Asians (Farley 1996), as well as understandings of race among Hispanics, many of whom treat Hispanic, white, black, Asian, and American Indian as comparable identifiers (Hirschman et al. 2000; OMB 1997). The two-question approach lowers the threshold for identifying as Hispanic (Hirschman et al. 2001) and leads to confusing responses. Comparisonscannot be made between Hispanic and non-Hispanic multiracials, because selecting two or more responses to the race question is very different from selecting a Hispanic origin in one question and a race in another. Moreover, it is not clear what people mean when they select a Hispanic origin and a race. Some are indicating mixed ancestry (e.g., mestizo mother from Mexico and a white father from Ireland), while others are indicating an an-

cestry and a nationality (e.g., Japanese from Peru, German from Argentina). Given the heightened ambiguity of racial responses for Hispanics, we focus on non-Hispanic youth and leave the examination of racial fluidity among Hispanics to future analyses of data that use a combined race and Hispanic origin question. Second, to facilitate comparisons of the two self-reported measures of race, we restrict the sample to youth who were interviewed both in school and at home. Twentyfour percent of non-Hispanic youth were only interviewed at home because some school administrators did not approve inschool interviews but allowed school rolls to be used for drawing the in-home sample. Overall, those who were interviewed only at home are quite similar to those interviewed in both contexts, although the former group is somewhat younger (15.3 versus 15.8), more female (50.7 percent versus 46.2 percent), and more likely to live in the South (42 percent versus 29 percent). However, the percentage of youth who identify as multiracial at home is nearly identical across the two samples. Finally, it is important to acknowledge that Add Health is a study of youth and so cannot be used to make inferences about the fluidity of race for other age groups. Nevertheless, we maintain that Add Health is an important source of information about the fluidity of race for at least two reasons. First, there is no other data set that contains multiple indicators of race for a large, representative sample. The 2000 census reports the race of individuals as well as the race of their coresidential parents, but because there is no way of knowing who reports their own race and who has a race assigned to them by the household member who completes the census form, census data are of limited use for studying patterns of racial classification. Second, there is a large and growing literature that relies on self-reported race from data sets such as Add Health, the National Educational Longitudinal Survey, and the National Longitudinal Study of Youth, to assess racial differences among adolescents (e.g., Jencks and Phillips 1998; Joyner and Kao 2000; Mouw and Xie 1999). This work implicitly assumes that the size and characteristics of racial populations are invariantto

618 10 9 8
0 ',
1)

AMERICAN SOCIOLOGICAL

REVIEW

* School Race

-1HomeRace

Parents' Race

7 6 5 4 3

f Ro0/

0)
1,

3.60/
9 AO/,

1) 0

2 1
0

.6% .6% 3% All Multiracial White/Black White/American Indian

.6% .4%

.9%

White/Asian

Figure 1. Percentage of Adolescents Classified as Multiracial, by Definition of Race: National Longitudinal Study of Adolescent Health

how race is classified. However, if this assumption does not hold, some of what we know about racial differences might be an artifact of how race is measured. PATTERNS OF RACIAL IDENTITY We begin by assessing levels of consistency across multiple indicators of race. If racial classification is independent of how it is measured, then all indicators should yield comparable estimates of the size of the multiracial population. However, as Figure 1 clearly shows, different measures of race provide significantly different estimates of the size of the multiracial population. Responses given by youth at school indicate that 6.8 percent are multiracial. By contrast, responses provided by these same youth at home reveal that only 3.6 percent are multiracial. Yet a third estimate emerges when, following the example of numerous previous studies (e.g., Eschbach 1995; Xie and Goyette 1998), we limit the sample to youth who live with both biological parents and define multiracial adolescents as those whose parents do not identify with the same monoracial group. This common definition suggests that 4.8 percent of youth are multiracial. A fourth estimate, not shown in Fig-

ure 1, is based on our analysis of Summary File 1 data from the 2000 census. It reveals that 2.5 percent of non-Hispanic youth between the ages of 12 and 18 were identified as multiracial. Figure 1 also shows great variation in patterns of expressed internal race for multiracial populations. At one extreme are white/ black and white/Asian youth: Both at home and at school, .6 percent of youth identify as white and black. Similarly, the share of youth identifying as white and Asian does not significantly differ between the school and home interviews. At the other extreme are white/American Indian youth. Although 2.4 percent of youth identify as white and American Indian at school, a significantly smaller share claim a white/American Indian identity at home (1.5 percent). These findings confirm that with respect to racial selfidentification there is not a single multiracial experience.

Table 2 extends our analysis of school and home race by contrasting individual-level responses. Column 1 reports patterns of school and home race for all adolescents. Our results provide strong supportfor a central tenet of the social constructionist perspective-that racial identities are fluid. In contrast to the assumptions implicit in most

WHO IS MULTIRACIAL?

619

of RacialIdentities, Distribution Adolescents Home-and School-Expressed Table2. Percentage by by Parents'Race

Parents'Race Missing Parent Does Not Live with Both Data Not Mixed Biological Parents Incomplete 91.3 .2 83.9 1.6 87.1 1.1

ExpressedRacial Identity ConsistentRacial Identity Same monoracialidentity in both contexts Same multiracialidentity in both contexts InconsistentRacial Identity Different monoracialidentity in each context Different multiracialidentity in each context Multiracialat home, monoracial at school Monoracialat home, multiracial at school "Refused"in either context Total UnweightedN Weightedpercentageof all adolescents

All 86.5 1.1

Mixed 36.5 11.9

2.8 .5 2.0 5.0 2.2 100.0

7.3 7.2 19.6 16.0 1.5 100.0 269 2.2

2.0 .1 1.1 3.6 1.7 100.0 4,695 44.7

3.3 .5 2.0 6.1 2.7 100.0 4,991 41.6

3.0 .5 2.1 4.4 2.0 100.0 1,576 11.4

11,531 100.0

Note: Weighteddata for non-Hispanicswith valid home and school race data.

empirical research, we find that only 87.6 percent of adolescents express identical racial identities across contexts.4 Table 2 further illustrates that as a result of the fluidity of racial identities, youth who identify with more than one racial group at home are not simply a subset of those who identify as multiracial at school. While 8.6 percent of youth report being multiracial at home or in school, only 1.6 percent identify themselves as multiracial in both contexts, and only 1.1 percent of youth select the same combination of two or more racial groups at home and in school. Consequently, 54 percent of the home multiracial population are not multiracial in school data, and 75 percent of the 4 We find no supportfor the hypothesisthat temporalgaps between school and home interviews explaininconsistencies expressedrace. in Unreportedanalyses show no relationshipbetweenthe probability shiftingexpressedracial of identityand the amountof time betweeninterviews.

school multiracial population are not multiracial in home data. The two multiracial populations are clearly overlapping, yet most youth who report being multiracial in one context identify as monoracial in the other context. Last, column 1 of Table 2 shows that 2.8 percent of youth express a multiracial identity by switching between single-race identities, rather than by selecting two or more racial groups in response to a particular race question. This group of multiracial youth is indistinguishable from monoracial youth in surveys that fail to collect multiple reports of race. The remainderof Table 2 summarizes patterns of school and home race by parents' race. These results show that having parents who claim to be from different racial groups is neither a necessary nor sufficient condition for expressing a multiracial identity. Column 2 summarizes racial identities for adolescents who are either the product of an interracial union, or have at least one bio-

620

AMERICAN SOCIOLOGICAL

REVIEW

Table3. Percentage Distribution Raceat Schoolfor Adolescents of WhoIdentifywithSelected Multiracial Groupsat Home
Race at Home Race at School White Black AmericanIndian Asian Other White/black 4.7 59.5 White/Black .9 20.8 White/American Indian 46.0 .1 20.0 .7 .7 21.8 2.4 White/Asian 13.4

White/American Indian White/Asian

Otherbiracial Threeor more racial groups No answer Total UnweightedN

1.2 2.5 7.4 2.9 100.0 108

| 24.1 45.91
1.0 7.3 7.3 9.1

100.0 159

100.0 82

Boxed cells indicateagreement Note:Cells areweightedcolumnpercentages. betweenschool andhomerace.

logical parentwho reports being multiracial. If race were simply hereditary, then we would expect these youth to self-identify as multiracial. However, the data show a marked deviation from this expectation. Although 54.7 percent of these adolescents do report being multiracial at least once, only 11.9 percent report the same multiracial identity in both contexts. An additional 7.3 percent of this group selects different singlerace identities in the two surveys, which means that no more than two-thirds of youth with known multiracial ancestry express a multiracial identity. This finding confirms that having parents of different races is not a sufficient condition for expressing a multiracial identity (Davis 1991). Although researchers have long known that some youth with multiracial ancestry identify as monoracial (DuBois [1899] 1996), previous work has largely assumed that having parents from different racial groups, or at least one parentwho is multiracial, is a necessary condition for expressing a multiracialidentity (Tafoya 2000). Column 3, which summarizes expressed internalrace responses for adolescents whose parents report being from the same monoracial group, suggests otherwise: Five percent of these youth report being multiracial at least once,

and another 2 percent offer different singlerace identities to the two race questions. Column 4 in Table 2 again contrasts responses to the school and home race questions, though only for youth who do not live with both biological parents and thus lack appropriate data for the parent-based race measure. This is the group that is missing in studies that rely upon parents' race to identify multiracial youth (e.g., Eschbach 1995; Xie and Goyette 1998). Although it has long been suspected that excluding such individuals from analyses of multiracial children might be problematic(Root 1992), the results in Table 2 provide an empirical basis for this concern. Comparedwith youth who live with both biological parents, adolescents who do not live with both biological parents are significantly less likely to give the same singlerace response to the school and home race questions (83.9 percent versus 88.6 percent) and are significantly more likely to ever report being multiracial (10.3 percent versus 7.4 percent). Our findings suggest that adolescents who do not live with both biological parents are more likely than other adolescents to express a multiracial identity, which implies that analyses of youth who live with both biological parentsfocus on a select subset of the multiracial population.

WHO IS MULTIRACIAL?

621

Although the summary statistics in Table 2 provide an overall portrait of patterns of agreement between school and home responses, they do not show specific transitions. This furtherinformation is revealed in Table 3, which reports school race for adolescents who identify with one of three multiracial groups at home. Again we see significant diversity in the consistency of identifying as white/black, white/American Indian, or white/Asian. A fairly high level of consistency is observed for white/black youth, though even for this group 40 percent of youth who identify as white/black at home identify differently in school. Twentyone percent identify as black, while 7.4 percent identify with three or more racial often white, black, and groups-most American Indian or "other";and 4.7 percent identify as "other"alone. White/Asian youth exhibit somewhat less consistency in racial identification, with less than one-half (45.9 percent) of those selecting this identity at home also selecting it in school. Youth who identify as white/Asian at home but not at school are most likely to identify as Asian at school, with "white" being the next most likely school race. Consistent with expectations (Snipp 1997), we find that the third multiracial group, white/American Indian, has the least stable racial identity (24 percent) and the highest probability of identifying as white in school (46 percent).5
FITTING MULTIRACIAL IDENTITIES INTO MONORACIAL BOXES

Thus far we have examined expressed internal racial identities in the context of "select 5 Table3 also supports contention inthat the consistenciesin reporting race are not the result of adolescents purposefullycreating fictional identitiesfor the survey.If thatwere the case, it is likely that not only would home and school race differ;they woulddifferdramatically (e.g., white/black Asian). However,Table 3 shows to thatwhenschool andhomeracediffer,inconsistencies are almostalways the result of a single racialgroupbeingaddedor dropped. Further supthat portfor the argument youthtookAddHealth questions seriously appearsin the observation thatmorethan99 percentof youthreported their countryof birth consistently in the two interviews.

all that apply" race questions. Although this is consistent with how race will soon be measured throughout the federal statistical system, American society has not always been tolerant of expressions of multiracial identity. Instead, popular and administrative discourses on race have assumed that racial groups are mutually exclusive (Nobles 2000). People who identify with more than one racial group have had a single race selected for them by state and federal agents (Davis 1991; Hahn et al. 1992; Hirschman et al. 2000; OMB 2000) and have been queried about their "real"race by acquaintances (T. Williams 1996). Given the slow pace of change of everyday understandings of race (Banton 1983), it is likely that changing how we measure race on official documents will not immediately alter the pure race assumption that is so central to conceptions of race in the United States. We assess how multiracial identities are converted to single-race identities by examining responses to the "best single race" item. In the Add Health home interview, youth who identify with more than one racial group are asked, "Which one category best describes your racial background?" Table 4 reports best single race separately for each multiracial group. White/American Indian adolescents offer the most predictable responses to the best single-race question, with nearly 86 percent selecting "white." This result is consistent with the assertion that many people regardAmerican Indian as a costless identity that can be adopted or abandonedat will (Harris 1994; Nagel 1995; Snipp 1997). As with white/American Indians, white/ black responses to the best single-race question are quite predictable, with almost 75 percent of white/black youth choosing "black." Given the enduring power of the one-drop rule, we suspect that many white/ black adolescents are socialized to understand that even if they identify as multiracial, they are "really" black (Davis 1991). So, when a girl who has just told an interviewer that she is white and black is asked what single race best describes her, it is likely that she knows what the "right" answer is. Given this, it is noteworthy that 17.1 percent of white/black adolescents select white as the single race that best describes

622

AMERICAN SOCIOLOGICAL

REVIEW

Table 4. Percentage Distribution of Selected Races at Home by Best Single Race Best Single Race White Black Indian Asian Other Refused Don't know Total UnweightedN 1.8 6.7 100.0 108 100.0 159 .1 .4 100.0 82 White/Black 17.1 74.5 14.1 52.0 Race at Home Indian White/American 85.9 White/Asian 47.4

Note: Cells are weightedcolumn percentages.

them. Further evidence of resistance to the one-drop rule is evident in the 8.5 percent of white/black adolescents who respond to the best single-race question by saying either that they do not know which single race best describes them or by simply refusing to give one. This suggests a commitment to being multiracial on the part of white/black youth that is not evident in the responses of white/ American Indian youth. Unlike the other two multiracial groups, white/Asian youth are equally likely to identify with their white and nonwhite heritages when asked to choose a single race. This finding is consistent with the results of Xie and Goyette (1998) and supportsthe hypothesis that the relatively small social distance between whites and Asians provides white/ Asian youth with the freedom to choose between monoracial identities in contexts where a multiracial identity is unacceptable. The logistic regression models in Table 5 present further detail about the monoracial identities of multiracialyouth. For each multiracial group we regress individual, family, and contextual factors on best single race. Our dependent variable is coded 1 if the best single race is white, and 0 otherwise. In response to previous work suggesting that gender and age affect racial identity (Corrinand Cook 1999; Jacobs 1992), we include indicators for whether the youth is female and whether he or she is at least 16 years old. We also include a measure of the educational attainment of the youth's coresidential parents, because prior work suggests that paren-

tal socioeconomic status is significantly related to racial identity (Xie and Goyette 1998). The final set of predictors addresses the hypothesis that context affects racial identity (Snipp 1997; Spickard 1992). The specific components of context we examine are region (a proxy for ideology, racial diversity, and the history of intergroup relations), neighborhoodracial composition, and whether family members were present during the home interview (i.e., private interview). All three models suggest that the individual and family-level factors we examine have little effect on the choice of a best single race. Sex and parents' education are not significant predictors for any of the multiracial groups, and age is only marginally significant (p < .10) for white/black youth. Among this group, youth who are at least 16 years old are about one-third as likely as younger adolescents to select white as the single race that best describes them. Instead, they are more likely to conform to the onedrop rule, maintaining that although they are white and black, they are more black than white. This finding is consistent with the expectations of Cooley's (1902) looking-glassself hypothesis and its reformulation by Nagel (1994). Both predict that older multiracial children will be more likely to adapt their identities to meet societal expectations, as years of being "corrected"about their racial identities take a toll. The finding of a racial difference in the age effect is also consistent with the assertion that American so-

WHO IS MULTIRACIAL? Table 5. Unstandardized Logistic Coefficients from the Regression of Best Single Race on Individual, Family, and Neighborhood Traits White/Blackat Home Variable Independent Female Age 16 or older Parentsat least some college a South Percentnon-Hispanic white in census tract Privateinterview
Constant

623

White/Indianat Home Coef. 1.16 (.77) .83 Odds Ratio 3.18 2.29 .83 1.66 1.61 .29
--.40

White/Asianat Home Coef. .22 (.78) -.85 Odds Ratio 1.24 .43 .69 .42 15.53 .16

Coef. .23 (.65) -1.11

Odds Ratio 1.26 .33 1.69 .20 7.00 4.78

(.68)
.52 (.77) -1.62* (.82) 1.95 (1.08) 1.56 (.83)
-3.32**

(.65)
-.18 (.67) .51 (.73) .48 (1.80) -1.23 (.68)
1.34

(.67)
-.38 (1.25) -.87 (1.33) 2.74* (1.21) -1.85* (.82) (.93) 6.16 6 .11 151 13.87 6 .16 81

(1.09) Model chi-square Degrees of freedom Pseudo R2 Numberof cases 9.87 6 .14 106

(1.60)

Note: Numbersin parenthesesare standarderrors.Dependentvariableequals 0 if best single race is not "white," 1 if best single race is "white."Data is weighted for non-Hispanicswho identify as white/black, white/AmericanIndian,or white/Asianat home and have valid responses for all independentvariables. a Measureof residentialparents'education;high school or less is the omittedcategory. *p < .05 **p< .01 (two-tailedtests)

ciety is especially sensitive to how white/ black youth identify (Davis 1991). Unlike age, sex, and socioeconomic status, our models suggest that context affects racial identification for all three multiracial groups. When white/black youth live in the South, they are significantly less likely to select white as their best single race. Given that the one-drop rule was born in the South and until fairly recently carried the weight of law in that region (Davis 1991), it is not surprising that so few southern white/black youth claim to be white. Contextual effects are also apparent in coefficients for neighborhood racial composition. For white/Asian youth, living in a neighborhood with more white neighbors significantly increases the probability that they identify white as their best single race. A similar, albeit marginally significant (p < .10), effect appears for white/black youth. These coefficients suggest that multiracialyouth consider their cul-

ture and social networks when selecting a single-race identity, although we cannot reject the hypothesis that neighborhood effects are a proxy for unobserved aspects of family culture and socialization (Duncan, Connell, and Klebanov 1997). Finally, Table 5 reveals that it is not only regional and neighborhood contexts that affect racial identification, but also the context of the interview. When no family members are present at the home interview, youth who initially identify as white/black are more likely to select white as their best single race, and white/American Indian and white/ Asian youth are more likely to choose American Indian and Asian, respectively. These effects are statistically significant (p < .05) for white/Asians, and approachstatistical significance for the other two multiracial groups (p < .10). Together, they suggest that in the presence of family members, which in Add Health usually means the

624

AMERICAN SOCIOLOGICAL

REVIEW

youth's mother, the choice of a best single race is more likely to follow racial norms that were dominant when parents and grandparentswere adolescents (Davis 1991; Snipp 1997) than the less restrictive norms that are increasingly common among today's youth (Senna 1998). DISCUSSION AND CONCLUSIONS

Our work has examined patterns and processes of racial classification for a nationally representative sample of youth. Analysis of Add Health data shows that for a substantial minority of adolescents, race is not only socially constructed, it is also fluid. Systematic inconsistencies in racial classification appear between self-reports, as well as between self-reports and parent-based measures. We find that two social factors are especially important for understanding patterns of racial fluidity. First, we observe that shifting racial regimes exert a significant influence on racial classification patterns. Whereas parents and Add Health interviewers came of age at a time when the one-drop rule dominated thinking about race, today's youth are being raised in a society that increasingly espouses the virtues of diversity and that has made real efforts to stress the legitimacy of multiracial identity (Nobles 2000; OMB 1997). Because expressed internal race is influenced by what we think about ourselves as well as by what others think about us, it is likely that changing perspectives on race and multiraciality partially explain patterns of multiracial reporting, the mismatch between parent's race and child's race, and the relationship between best single race and whether a parent is present during the home interview. In each case, greater anonymity leads to racial classifications that are more consistent with contemporary understandings of race. Comparedwith the intervieweradministeredhome survey, the self-administered school survey provides relatively little opportunity for youth's conceptions of race to be influenced by the norms of older cohorts. By contrast, contemporary conceptions of race figure less prominently in census race data, which in most cases reflects the beliefs of the household member who completes the census form. Similarly, we

suspect that some children of monoracial parents classify themselves as multiracial because parents and children disagree about how recently mixed one must be to report being multiracial. Also, we maintain that having family members present for the home interview affects best single race because it heightens the saliency of parental perspectives on race. Second, we observe that patterns of racial classification vary because multiracial groups comprise socially distinct monoracial groups. White/American Indian youth emerge as the largest, but least committed, multiracial group. This finding is consistent with other work that suggests white/American Indian identity is often expressed by whites who have little ancestral, phenotypical or cultural connection to American Indians, but who nevertheless wish to appeal to popular norms of multiculturalism by presenting a diverse portrait of themselves (Eschbach 1995; Harris 1994; Snipp 1997). Quite a different picture emerges for white/ black youth. Here we see evidence of the prominent, yet fading, influence of the onedrop rule on racial self-identification. Yet a third patternemerges for youth who identify as white and Asian. In an extension of Xie and Goyette (1998), we show that racial identification is not only optional for this group when it is provided by parents on the census, but that the strong social rules governing white/black and white/American Indian classification are also absent for white/ Asian self-identification. Although our work addresses some of the factors that affect the fluidity of race, still other factors are beyond its scope. Among these are the effects of age, period, and cohort. Add Health provides unique insights into racial classification patterns for adolescents, but contains no information about how these patternswill change as the sample ages, how this cohort compares to previous cohorts of adolescents, or the extent to which racial classifications have become more fluid for all cohorts in the contemporary period. Given the lack of comparable data for other cohorts and the absence of panel data on race for the Add Health cohort, it is not possible to assess the extent of age, period, or cohort effects. However, it would be surprising if the growing attention to

WHO IS MULTIRACIAL?

625

multiraciality in American society were not promoting greater flexibility, and therefore fluidity, in racial identity among all cohorts in the current period. It would also be surprising if racial fluidity were not greater for adolescents than for adults, because identities are generally less stable in adolescence (Demo 1992) and also because the identities of older cohorts were formed during a time of more restrictive racial regimes. It is less clear how time will affect the racial identities of today's adolescents. Research suggests that adulthood will bring more stability (Waters 1990), but one cannot predict how much less fluid racial identity will become, or whether identification with multiple racial groups will subside. As for the 2000 census, our work makes clear that it was a count of a multiracial population, not the multiracial population. Specifically, it captured the number of Americans who were identified as belonging to two or more racial groups by the person who completed their household's census form. This is probably not the same as the number who would have self-identified as multiracial, the number who self-identify as multiracial in everyday situations, the number who are identified as multiracial by strangers,or the number who have ancestors from more than one racial group. This does not mean that the Census Bureau erred in its enumeration of the multiracial population. Rather,we maintain that the census, like every other data set, captures an individual's "true"race for a particularpurpose, in a particular context, at a particular point in time (Telles and Lim 1998). The practical implication of this realization is that analysts must think critically about what they mean by race, design surveys that more precisely measure race, and be aware of the implications of mismatches between available and ideal racial data. Failing to do so will ensure that as the racial diversity of the United States grows, so too will our inability to understand that diversity.
David R. Harris is Assistant Professor of Sociology at the University of Michigan and Assistant Research Scientist at the Institutefor Social Research. He continues to explore the social construction of race through analyses of data from the National Longitudinal Study of Adolescent

Health and a web-based survey of University of Michigan freshmen. He is using insights drawn from this work to critique federal guidelines for collecting data on race, and to examine educational and mental health outcomes for monoracial and multiracial youth. Jeremiah Joseph Sim received his MA in Sociology from the University of Michigan. His sociological interests include multiracial identity and residential racial segregation. In September he will pursue a new set of interests as he begins graduate study in the Department of Industrial and Operations Engineering.

REFERENCES
Alba, Richard D. 1990. Ethnic Identity. New Ha-

Press. ven, CT:Yale University

Banton, Michael. 1983. Racial and Ethnic Com-

Unipetition.Cambridge, Cambridge England: versityPress. PeterS., Jo Jones,andJ. Richard Bearman, Udry. 1997. "The National LongitudinalStudy of Adolescent Health: Research Design." National Longitudinal Study of Adolescent Health. RetrievedFebruary21, 2001 (http:// www.cpc.unc.edu/projects/addhealth/design. html). America.New York:Harper Row. and Kim and Joyce Tabor. 1999. "StrateChantala, a gies to Perform Design-BasedAnalysisUsing the AddHealthData."NationalLongitudinal Study of Adolescent Health. Retrieved February21, 2001 (http://www.cpc.unc. edu/projects/addhealth/strategies.html). Clegg, Roger. 2001. "CensusSense."National Review Online. Retrieved March 17, 2001 (http://www.nationalreview.com/contributors/ clegg030701.shtml). Cooley, CharlesHorton. 1902. HumanNature
and the Social Order. New York: Charles Blauner, Robert. 1972. Racial Oppression in

Scribner's Sons. Corrin,WilliamJ. and ThomasD. Cook. 1999. "Spanning Racial Boundaries: Multiracial Adolescents and Their Families, Peers, Schools,andNeighborhoods." Working Paper No. 99-20, Institute for Policy Research, IL. Northwestern Evanston, University,
Davis, F. James. 1991. Who Is Black? One Nation's Definition. University Park, PA: Pennsylvania State University Press. Demo, David H. 1992. "The Self-Concept Over Time: Research Issues and Directions." Annual Review of Sociology 18:303-26. of Pennsylvania Press. Duncan, Greg J., James P. Connell, and Pamela

DuBois,W.E. B. [1899] 1996. ThePhiladelphia PA: University Negro. Reprint,Philadelphia,

626

AMERICAN SOCIOLOGICAL

REVIEW Homophily." Social Science Quarterly 81: 810-25. Kao, Grace. 1999. "Racial Identity and Academic Performance: An Examination of Biracial Asian and African Youth." Journal of Asian American Studies 2:223-49. Lieberson, Stanley and Mary C. Waters. 1993. "The Ethnic Responses of Whites: What Causes Their Instability, Simplification, and Inconsistency?" Social Forces 72:421-50. Mouw, Theodore and Yu Xie. 1999. "Bilingualism and the Academic Achievement of Asian Immigrants: Accommodation with or without Assimilation?" American Sociological Review 64:232-52. Nagel, Joane. 1994. "Constructing Ethnicity: Creating and Recreating Ethnic Identity and Culture."Social Problems 41:152-76. .1995. "AmericanIndian Ethnic Renewal: Politics and the Resurgence of Identity." American Sociological Review 60:947-65. Nobles, Melissa. 2000. Shades of Citizenship: Race and the Census in Modern Politics. Stanford, CA: Stanford University Press. Office of Management and Budget (OMB). 1997. "Revision to the Standards for the Classification of Federal Data on Race and Ethnicity." Federal Register, vol. 62, no. 210 (October 30). .2000. "Guidanceon Aggregation and Allocation of Multiple Race Responses for Use in Civil Rights Monitoring and Enforcement." Bulletin number00-02, Office of Management and Budget, Washington, DC. Omi, Michael A. 2001. "The Changing Meaning of Race." Pp. 243-63 in America Becoming: Racial Trendsand Their Consequences, vol. 1, edited by N. J. Smelser, W. J. Wilson, and F. Mitchell. Washington, DC: National Academy Press. Root, Maria P. P. 1992. "Back to the Drawing Board: Methodological Issues in Research on Multiracial People." Pp. 181-9 in Racially Mixed People in America, edited by M. P. P. Root. Newbury Park, CA: Sage. Schmitt, Eric. 2001. "For 7 Million People in Census, One Race Category Isn't Enough." New YorkTimes, March 13, pp. Al. Senna, Danzy. 1998. "The Mulatto Millennium: Waking Up in an Age of Racial Ambiguity." Utne Reader 89:31-34. Snipp, C. Matthew. 1997. "Some Observations about Racial Boundaries and the Experiences of American Indians."Ethnic and Racial Studies 20:667-89. Spickard, Paul R. 1992. "The Illogic of American Racial Categories." Pp. 12-23 in Racially Mixed People in America, edited by M. P. P. Root. Newbury Park, CA: Sage. Suro, Robert. 1999. "Mixed Doubles." American Demographics 21:56-62.

K. Klebanov. 1997. "Conceptual and Methodological Issues in Estimating Causal Effects of Neighborhoods and Family Conditions on Individual Development." Pp. 219-50 in Neighborhood Poverty: Context and Consequences for Children, edited by J. Brooks-Gunn, G. J. Duncan, and J. L. Aber. New York: Russell Sage Foundation. Eschbach, Karl. 1995. "The Enduring and Vanishing American Indian: American Indian Population Growth and Intermarriage in 1990." Ethnic and Racial Studies 18:89-108. Espiritu, Yen Le. 1992. Asian American Panethnicity: Bridging Institutions and Identities. Philadelphia, PA: Temple University Press. Farley, Reynolds. 1996. The New American Reality. New York: Russell Sage. Goldstein, Joshua R. and Ann J. Morning. 2000. "The Multiple-Race Population of the United States: Issues and Estimates." Proceedings of the National Academy of Sciences 97:6230-35. Goodman, Alan H. 2000. "Why Genes Don't Count (for Racial Differences in Health)." American Journal of Public Health 90:1699702. Gould, Stephen Jay. 1996. The Mismeasure of Man. New York: W. W. Norton. Hahn, Robert A., Joseph Mulinare, and Steven M. Teutsch. 1992. "Inconsistencies in Coding of Race and Ethnicity Between Birth and Death in US Infants: A New Look at Infant Mortality, 1983 through 1985." Journal of the American Medical Association 267:259-63. Harris, David R. 1994. "The 1990 Census Count of American Indians: What Do the Numbers Really Mean?" Social Science Quarterly 75: 580-93. . Forthcoming. "Does It Matter How We Measure? Racial Classification and the Characteristics of Multiracial Youth." In The New Race Question, edited by J. Perlmann and M. C. Waters. New York: Russell Sage. Harrison, Roderick J. Forthcoming. "Inadequacies of Multiple Response Race Data in the Federal Statistical System." In The New Race Question, edited by J. Perlmann and M. C. Waters. New York: Russell Sage. Hirschman, Charles, Richard Alba, and Reynolds Farley. 2000. "The Meaning and Measurement of Race in the U.S. Census: Glimpses into the Future."Demography 37:381-93. Jacobs, James H. 1992. "IdentityDevelopment in Biracial Children." Pp. 190-206 in Racially Mixed People in America, edited by M. P. P. Root. Newbury Park, CA: Sage. Jencks, Christopherand Meredith Phillips. 1998. The Black-White Test Score Gap. Washington, DC: Brookings Institution. Joyner, Kara and Grace Kao. 2000. "School Racial Composition and Adolescent Racial

WHO IS MULTIRACIAL? Tafoya, Sonya M. 2000. "Check One or More .. . Mixed Race and Ethnicity in California." California Counts 1:1-11. Telles, Edward E. and Nelson Lim. 1998. "Does It Matter Who Answers the Race Question? Racial Classification and Income Inequality in Brazil." Demography 35:465-74. Twine, France Winddance. 1997. "BrownSkinned White Girls: Class, Culture, and the Construction of White Identity in Suburban Communities." Pp. 214-43 in Displacing Whiteness: Essays in Social and Cultural Criticism, edited by R. Frankenberg.Durham, NC: Duke University Press. Waters, Mary C. 1990. Ethnic Options: Choosing Identities in America. Berkeley, CA: University of California Press.

627

. 1999. Black Identities: West Indian Immigrant Dreams and American Realities. New York: Russell Sage. Williams, Gregory Howard. 1995. Life on the Color Line: The True Story of a White Boy Who Discovered He Was Black. New York: Dutton. Williams, Teresa Kay. 1996. "Race as Process: Reassessing the 'What Are You?' Encounters of Biracial Individuals." Pp. 191-210 in The Multiracial Experience: Racial Borders as the New Frontier, edited by M. P. P. Root. Thousand Oaks, CA: Sage. Xie, Yu and Kim Goyette. 1998. "The Racial Identification of Biracial Children with One Asian Parent: Evidence from the 1990 Census." Social Forces 76:547-70.