Split Sibling Correlations Aren’t What You Think: The Hidden Mean Differences

Research on social mobility has increasingly emphasized heterogeneity in mobility experiences. One popular approach examines heterogeneity in sibling correlations by splitting samples according to parental or ancestral characteristics. A prominent example is estimating sibling correlations in children’s earnings stratified by parental earnings. What remains underappreciated, however, is that splitting the sample by social background abstracts from mean differences in children’s earnings by social background—differences that constitute an important channel of intergenerational reproduction. Drawing on a decomposition by Karlson and In (2024) and a procedure used in Hällsten and Kolk (2023), I show how these mean differences can be incorporated. Replicating a newly published Swedish study (Forsberg et al., 2025) that reports attenuating sibling correlations in children’s earnings at higher levels of parental earnings, I show that accounting for mean differences substantially alters the conclusions, yielding sibling correlations that instead increase with parental earnings.