Observing Racial Hierarchies in the Response Variability of Multiracial Persons

Victor Thompson, Stanford University
Matthew Snipp, Stanford University

In this paper, we analyze data from the 2002 census Quality Survey. Preliminary results show a high degree of stability in the responses of Black and White persons, and for persons who report only one race. However, a high level of response variability has been documented for persons who identify themselves with more than one race. We estimate baseline models of independence and quasi-independence in the association of racial identification over time. We attempt to model variability in the responses of multiracial persons by fitting log-linear models that parameterize identity shifts consistent with known hypotheses about racial hierarchies in the United States. For example, given the principle of hypodescent for determining membership in the African-American population, we predict the persons who identify themselves as African-American and White at one point in time, would be more likely to identify themselves as Black rather than White at some later point in time.