Capturing complexity in the United States: which aspects of race matter and when?
It is tempting to assume there are clear distinctions that identify a person as being a particular race or ethnicity. Though the characteristics that define racial or ethnic difference vary across societies, it is nevertheless common for people to maintain that their country’s ‘others’ are easily singled out e.g., by face, accent, name, or dress. Indeed, in recent years, governments around the world have begun to mandate the collection of data to monitor racial/ethnic discrimination as if the information needed were obvious. In the United States, official racial data has been collected since at least 1790, and how it should be gathered was rarely questioned because, according to commonsense belief, racial differences were ‘unmistakable’.1 Today, the assumption of measurement agreement can take on a different tone as racial data is put to different purposes: why quibble over
conceptual definitions when the inequalities are so obvious and so much larger than any possible ‘errors’ in the data? Surveys in the United States have long taken an all-measures-are-
interchangeable approach to capturing race and ethnicity. Many, including the census, have shifted over time from employing the enumerator’s racial classification to encouraging self-identification; yet the interpretation of results remains the same. Other surveys have combined both types of information into one variable as when interviewers are instructed to ask for the respondents’ self-identification but record their race by observation if they refuse to answer. When calculating race-specific vital statistics, such as birth or death rates, the numerators come from an unknown combination of selfidentification and the observations of nurses or funeral directors while the denominators draw on census self-(or household head) identifications. Recent research casts doubt on the validity of this assumption that
all measures of race are created equal raising the question of how data on race and ethnicity should be collected in the United States. Most of this work has focused on the extent to which the census and other surveys force people to select ill-fitting categories that do not accurately capture their preferred, often complex identities (e.g., Rodriguez 2000; Rockquemore and Brunsma 2002). A growing literature also suggests that efforts aimed at trying to get one ‘right’ answer to the race question might be misdirected if different measures of race measure different but perhaps equally useful things (Saperstein 2006; Roth 2010). This study demonstrates the added benefit of including multiple
measures of race by examining racial disparities in adult outcomes using two measures of race: how the person self-identified and how they were classified by the survey interviewer. Drawing on data from the 1988 National Survey of Family Growth, I show that incorporating both measures of race reveals more complex patterns of racial advantage and disadvantage in the United States than can be seen using standard single-measure methods. Further, by comparing disparities across two domains family income and health care I show that the salience of each aspect of race varies depending on the outcome in question. Though identifying the specific mechanisms that perpetuate racial inequality in each domain is beyond the scope of this study, the results make a compelling case for reconsidering standard methods of racial data collection to allow for such studies in the future. Similarly, while these findings may not generalize to the present day or for all Americans, they raise important questions about the processes that maintain racial inequality in the United States, as well as how to design effective policy interventions.