Counting by Race and Sex: Categorical Fluidity and Survey Weights in the American Community Survey


Evidence continues to mount that reports of an individual’s race, ethnicity and sex are subject to change over time. Many survey weights, however, implicitly make the assumption of fixed race, ethnicity, and sex categories. I use data from the American Community Survey and Census intercensal and postcensal estimates to construct two sets of survey weights: one under assumptions of fixed sex, race, and ethnicity categories following Census Bureau methodology, and one without assumptions of fixed categories. I then estimate net categorical fluidity by comparing components of population change under the two weighting assumptions. In a subsequent regression analysis, I show that there is significantly patterned categorical fluidity among race and ethnicity groups, and categorical fluidity is less substantially patterned within sex groups. I also compare estimates of unemployment rates, poverty rates, and diversity indices under each of the weighting assumptions and find mixed results. Unemployment and poverty rates are not substantially affected by weighting assumptions, but racial composition and diversity can change substantially. Researchers applying survey weights should be aware of the assumptions implicit in their calculations, and consider the sensitivity of their analyses to these assumptions.

Seattle, WA