What Do Group-Specific Sibling Correlations Really Measure?

The sibling correlation, or intraclass correlation coefficient (ICC), is a widely used omnibus measure of the total impact of family background on socioeconomic outcomes in social stratification research. A growing literature compares these correlations across subgroups defined by family-level characteristics (e.g., race, parental education) to test theories of social reproduction, such as parental compensation or reinforcement. In this note, I highlight the ambiguities involved in meaningfully interpreting such subgroup variation. I first define the theoretical estimand of intergenerational transmission that researchers are arguably most interested in, and then formally demonstrate why group-specific ICCs do not identify this estimand. In fact, the estimand of interest cannot be recovered from the variance-components models underlying sibling correlations unless researchers are willing to make assumptions that are not only untestable but also likely unrealistic.