Is race in the eye of the beholder? A longitudinal, comprehensive review of measures of racial classification in Add Health


Larimore, Savannah; Esposito, Michael; Rafferty, Jane; Hicken, Margaret T.; Hargrove, Taylor Woodland; & Lee, Hedwig (2018). Is race in the eye of the beholder? A longitudinal, comprehensive review of measures of racial classification in Add Health. 2018 Add Health Users Conference. Bethesda, MD.


Recently, health scholars have attempted to move beyond unidimensional measures of race in their analyses to look within as well as across racial and ethnic groups for meaningful sources of variation. Add Health, which includes measures of not only self-classified race, but observed race, skin tone, and reflected race, has been a commonly used data set in these endeavors. While Add Health also includes multiple measures of racial classification across waves, few studies using Add Health data consider the fluctuations in these measures within individuals across waves. Given the theoretical relationship between experiences of racial misidentification, the centrality of racial identity, experiences with racial discrimination, and health, a failure to consider the longitudinal and multidimensional nature of race may lead to incorrect estimation of effects. For example, preliminary analyses find that individuals who are otherwise race-concordant (i.e., their self-classified race and observed race are the same) in Waves I & III may be race-discordant in Wave IV. A respondent may self-classify and be observed as "White” in Wave I, self-classify and be observed as "Asian/Pacific Islander” in Wave III, and be observed as "White” in Wave IV. These individuals likely have a different set of racialized experiences than their peers who are never recorded as race-discordant in Add Health. Additionally, since self-classified race is not explicitly asked in Wave IV, rather it is assumed to remain unchanged between Waves III and IV, race-discordance in Wave IV represents a theoretically different process of racial misidentification than race-discordance in Waves I and III. As such, the goal of this analysis is two-fold: we first present a comprehensive and descriptive summary of the variations in survey design and racial classifications across waves, making specific note of how the prior likely informs the latter and implications for these variations in health research. In this summary, we will also provide a descriptive analysis of the sociodemographic and contextual characteristics of respondents who experience racial fluctuation and/or discordance across waves. Second, we outline strategies for dealing with this variation, both conceptually and methodologically, and will test the utility of these strategies for self-rated health. In doing so, we demonstrate ways that researchers can fully account for various expressions of race in Add Health.

Reference Type

Conference proceeding

Book Title

2018 Add Health Users Conference


Larimore, Savannah
Esposito, Michael
Rafferty, Jane
Hicken, Margaret T.
Hargrove, Taylor Woodland
Lee, Hedwig

Year Published


City of Publication

Bethesda, MD

Reference ID