Skip Navigation

Demography's Race Problem

Skip sharing on social media links

David R. Harris
University of Michigan

Presented at the NICHD panel,
"Visions of the Future: A Town Meeting on New Directions in Population Research"
Annual Meeting of the Population Association of America

March, 2000

…the problem of the Twentieth Century is the problem of the color-line.

W. E. B. DuBois (1982 [1903], xi)

Last year I was approached by a graduate student who was having trouble writing a research proposal for my class. It was not that the student could not read the literature and generate interesting research questions, or even that he was unfamiliar with the appropriate methodologies; rather, the student was struggling with how to define one of his key variables. The student was interested in racial differences in age at first intercourse, and had proposed to use the National Longitudinal Study of Adolescent Health (Add Health) to test his hypotheses. Problems arose because Add Health contains numerous racial identifiers for each respondent. Youth are asked their race at home and in school; are allowed to identify with multiple racial groups, but then asked to select the monoracial group that best describes them; are assigned a race by interviewers; and can have their race surmised from the races reported by their biological parents. My student’s question was simply, "Which race variable should I use?" This question is one that will increasingly be asked by demographic researchers in the coming years, and I fear that in too many instances the response from the demographic community will be the same as what I offered my student—"I’m not sure." In hopes of averting this crisis in racial classification I submit that a key focus of the Demographic and Behavioral Sciences Branch (DBSB) of the National Institute of Child Health and Human Development (NICHD) should be encouraging a deeper understanding of what race is and how it affects life chances.

For several reasons now is a critical time for demographers to seriously debate racial identification. Over the past 35 years the United States population has become increasingly racially and ethnically diverse, and all indications are that this "browning" of America will continue for years to come (Hirschman 1994; Waters 1996). One obvious source of increasing diversity is immigration. Recent decades have seen substantial growth in the number of immigrants to the United States (Farley 1996; Massey and Singer 1995). This trend has implications for racial diversity because the majority of recent immigrants have been Asian or Hispanic rather than white. In addition to immigration, racial diversity in the United States has also been fueled by tremendous growth in interracial marriage and fertility. Sparked in large part by the civil rights movement of the 1950s and 1960s, as well as the 1967 Supreme Court ruling against state antimiscegenation laws, the share of all marriages that are between people of different races has more than tripled since 1970 (Besharov and Sullivan 1996; Farley 1999; Spickard 1989; United States Census Bureau 1998). The rate of interracial partnering is even higher among cohabitations, where young Asian women are more likely to cohabit with a white man than with an Asian man, and about one in eight black men who cohabit do so with a white woman (Harris and Ono 2000). As one would expect, these increases in interracial partnering have spawned a boom in the mixed-race population (Kalish 1992).

In addition to the increasing racial and ethnic diversity of the U.S. population, the complexity of recent race data also calls for a renewed focus on racial identification. I have already mentioned the myriad race questions in Add Health. This dataset provides a wealth of information for researchers interested in many of the core areas of demography (e.g., fertility, sexual activity, education, health), and so the question that my student faced about how to classify the race of respondents is one that has already confronted many demographers and will continue to be an issue for years to come. Even greater uncertainty about how to analyze race data has been created by the recent revision of Office of Management and Budget (OMB) Directive 15, which guides the collection of race data in the federal statistical system (Office of Management and Budget 1997). No longer will federal statistics contain an assumption of mutually exclusive racial groups. Instead, by 2003, all federal data collection must allow people to identify with more than one racial group. This change, which will be apparent when the 2000 Census data are released, has tremendous implications for the demographic community. Heretofore straightforward tasks, such as assessing levels and trends in racial inequality, will require new assumptions about race. Decisions will have to be made about whether to analyze all monoracial and mixed-race groups separately, or whether certain groups can be merged together. Again, there is a dearth of expertise in the demographic community to guide these decisions.

Last, a renewed focus on race is warranted because we are in a political and social climate that is challenging how much, and even if, race affects life chances. The President’s recent Commission on Race, legal challenges to affirmative action at places like the University of Michigan, and a national retreat from efforts to promote school integration (Orfield and Eaton 1996) are all evidence of this trend. In this context it is all the more important that demographers conduct thoughtful, thorough analyses of racial differences that are clear about not just how much race matters, but also why race matters.

Given the need to understand how race should be identified and how measurement affects analyses of racial differences, it is important to take stock of what we already know about these topics. Here I am afraid that the news is not good. Few surveys pose innovative race questions to large, nationally-representative populations, so much of what we know about how people answer race questions is highly speculative. However, thanks to new data sources, findings are beginning to emerge about the relationship between how race questions are asked and the answers people give. One important source of information derives from a study by Hirschman, Alba, and Farley (forthcoming). They examine responses to the 1996 Race and Ethnic Targeted Test (RAETT), which administered different race and ethnicity questions to five subsamples. The standard Current Population Survey race and Hispanic origin questions were varied by adding a multiracial category, allowing respondents to identify with multiple groups, and/or combining the race and Hispanic origin questions. Hirschman and his colleagues report that the size of racial and ethnic groups is largely insensitive to which race question is used, which suggests that we do not need to worry too much about racial identification issues.

In contrast to the RAETT findings, several other studies suggest that not only are racial identities subjective, but that how race is measured has large effects on several interesting outcomes. Hahn, Mulinare, and Teutsch (1992) compare data from matched birth and death certificates for infants who died during their first year of life. The overwhelming majority of children who are identified as black or white at birth are identified the same way at death. However, for other populations there are significant shifts in racial identity. As a result of inconsistencies in racial identification, it is possible to compute infant mortality rates for selected racial and ethnic groups that, depending on how race is defined, vary by as much as 50 percent.

Further evidence that how we measure race can affect our understanding of racial differences on important outcomes appears in the work of Collins and David (1993). Like many researchers before them, Collins and David are interested in racial differences in birthweight. What is different about their study is that it considers the race of both parents when stratifying the newborn population. This unusual step reveals complex race differences. Even after controlling for various characteristics of parents and neighborhoods, children born to white fathers and black mothers are still 40 percent more likely to be low birthweight than are children born to black fathers and white mothers. Collins and David further find that birthweight is similar among children born to white mothers, regardless of whether the father is white or black. Had Collins and David pursued either of the conventional approaches to identifying the race of newborns (i.e., used race of mother or treated as black anyone with at least one black parent), the finding that only the race of the mother affects birthweight would have been lost.

Finally, an even deeper analysis of the question of how measurement issues affect racial identity is possible with Add Health. Ongoing analyses of inconsistencies across the aforementioned Add Health race questions present a complex picture of racial identity (Harris and Sim 2000). These data show tremendous fluidity of racial identity Only 88 percent of Add Health respondents give the same response to the school and home race questions. More than one racial group is selected by seven percent of adolescents at school, four percent at home, and eight percent in at least one of the two contexts. Further discrepancies appear between the racial identities provided by youth and those assigned to them by interviewers. For example, only 67 percent of youth who identify as white and black, and then select white as the race that best describes them, are also identified as white by interviewers. By contrast, 95 percent of the white-black youth that most identify with blacks are identified as black by interviewers. The implication of this work is that how we measure race affects the racial identity of a nontrivial proportion of the population. Preliminary analyses further suggest that not only does membership in racial groups vary across race questions, but some characteristics of racial populations also depend on how race is measured.

The fact that membership in racial groups can be affected by how race is measured means that demographic analyses of racial differences must be supported by strong theories of why race matters. If we think all racial groups have unique experiences, then it may be preferable to treat all monoracial and mixed-race groups as distinct from one another. By contrast, if theory suggests that for particular outcomes all that matters is whether one is in any way identified with particular minority groups, then applications of the one-drop rule may be more appropriate (Davis 1991). Yet a third possibility is that outcomes are affected by what others believe an individual’s race to be (e.g., employment discrimination), not what the individual considers himself or herself to be. Finally, for some outcomes theory might dictate that what is important is one’s cultural, rather than phenotypic, race. This would imply that people who are generally seen as white, but speak "black" English, listen to hip-hop music, dress in urban styles, and exhibit other stereotypically black cultural traits may fare worse than comparable youth who adopt white culture. My point here is that in order to assess whether the color line persists as the central problem in American life we must first be able to determine where the color line is and who is on either side of it.

So what specifically could DBSB do to improve our understanding of race? I have several suggestions. First, it will be important to fund research that looks at the specific issues I have been discussing. This could include work that examines contextual effects on racial identity, the sensitivity of racial identity to changes in survey questions, and changes in the characteristics of racial populations that are attributable to variations in the way race is measured. Second, DBSB could encourage basic theoretical work on why race might affect demographic outcomes, and encourage all analyses that use race variables to be explicit about the advantages and disadvantages of available measures of race. Third, DBSB could encourage a greater emphasis on race in demography training programs. I am aware of demography of race courses at the University of Michigan and University of Pennsylvania, but I do not know whether similar classes are offered at the other population centers. Fourth, I would encourage DBSB to support the collection of innovative race data on the surveys it funds. Asking race in multiple waves of longitudinal surveys, recording interviewer observations of racial identity, asking "check all that apply" and "best race" questions, and asking race questions of every household member who is interviewed would go a long way towards providing the demographic community with the data it needs to assess measurement effects on racial identity and racial differences. Finally, I strongly urge DBSB to band with other governmental agencies to support a conference or series of workshops on how to use data collected under the new governmental standards.

As should be clear by now, racial identities are socially constructed and context-specific. Acknowledging this fact means that understanding the determinants of racial identity is a critical component of understanding population composition, and anyone who wants to use race as a variable must decide which measure of race represents the best match with theory. By encouraging a deeper understanding of race, DBSB will increase the odds that when some other student approaches some other faculty member with confusion over how to code complex race data, the response that he or she gets will be more informed than what I was able to offer.


Besahrov, Douglas J. and Timothy S. Sullivan. 1996. "One Flesh." The New Democrat (July/August) pp. 19-21.

Collins, James W. and Richard J. David. 1993. "Race and Birthweight in Biracial Infants." American Journal of Public Health 83: 1125-1129.

Davis, F. James. 1991. Who Is Black?: One Nation’s Definition. University Park, PA: Pennsylvania State University Press.

DuBois, W.E.B. 1982 [1903]. The Souls of Black Folks. New York: Penguin Books.

Farley, Reynolds. 1996. The New American Reality. New York: Russell Sage Foundation.

. 1999. "Racial Issues: Recent Trends in Residential Patterns and Intermarriage." Pp. 86-128 in Diversity and Its Discontents: Cultural Conflict and Common Ground in Contemporary American Society, edited by Neil J. Smelser and Jeffrey C. Alexander. Princeton, NJ: Princeton University Press.

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: 259-263.

Harris, David R. and Hiromi Ono. 2000. "Estimating the Extent of Intimate Contact between the Races: The Role of Metropolitan Area Factors and Union Type in Mate Selection." Presented at the Meetings of the Population Association of America, Los Angeles.

Harris, David R. and Jeremiah Joseph Sim. 2000. "Who Is Mixed Race?: Patterns and Determinants of Adolescent Racial Identity." Unpublished manuscript.

Hirschman, Charles. 1994. "The Meaning of Race and Ethnic Population Projections." Unpublished manuscript.

Hirschman, Charles, Richard Alba, and Reynolds Farley. Forthcoming. "The Meaning and Measurement of Race in the U.S. Census: Glimpses into the Future." Demography.

Kalish, Susan. 1992. "Interracial Baby Boomlet in Progress?" Population Today 20: 1-2, 9.

Massey, Douglas S. and Audrey Singer. 1995. "New Estimates of Undocumented Mexican Migration and the Probability of Apprehension." Demography 32: 203-213.

Office of Management and Budget. 1997. "Revision to the Standards for the Classification of Federal Data on Race and Ethnicity." Federal Register (October 30).

Orfield, Gary and Susan E. Eaton. 1996. Dismantling Desegregation. New York: The New Press.

Spickard, Paul R. 1989. Mixed Blood: Intermarriage and Ethnic Identity in Twentieth-Century America. Madison, WI: University of Wisconsin Press.

Telles, Edward E. and Nelson Lim. 1998. "Does it Matter Who Answers the Race Question? Racial Classification and Income Inequality in Brazil." Demography 35: 465-474.

United States Census Bureau. 1998. "Race of Wife by Race of Husband: 1960, 1970, 1980, 1991, and 1992."

Waters, Mary C. 1996. "The Social Construction of Race and Ethnicity: Some Examples from Demography." Presented at the Meetings of the American Sociological Association, New York.


Last Reviewed: 11/30/2012
Vision National Institutes of Health Home BOND National Institues of Health Home Home Storz Lab: Section on Environmental Gene Regulation Home Machner Lab: Unit on Microbial Pathogenesis Home Division of Intramural Population Health Research Home Bonifacino Lab: Section on Intracellular Protein Trafficking Home Lilly Lab: Section on Gamete Development Home Lippincott-Schwartz Lab: Section on Organelle Biology