Understanding intergenerational economic mobility by decomposing joint distributions

We propose a simple and generalizable decomposition method to evaluate intergenerational economic mobility. The method decomposes the difference between the empirical parent-offspring joint distribution of incomes and a hypothetical independent parent-offspring joint distribution of incomes. The difference is attributed to (1) a portion due to a link between parental income and offspring characteristics (a composition effect) and (2) a portion due to a link between parental income and the returns to characteristics (a structure effect). The method is based on the estimation of counterfactual joint distributions consistent with (actual and counterfactual) conditional distributions estimated via distributional regression and (actual and counterfactual) distributions of covariates. The counterfactual joint distributions are then caste into common measures of mobility found in the literature: intergenerational elasticities of incomes and their quantile regression counterparts, transition matrices, summary indices of transition matrices, and upward mobility probabilities. These counterfactual measures are used to assign portions of measured (im)mobility to composition and structure effects. We apply the method to U.S. intergenerational economic mobility of white males born between 1957 and 1964. Across multiple mobility measures and using two different counterfactuals, we find that the composition effect (i.e., differences in the distribution of characteristics) generally accounts for about 60-70% of the measured mobility gap. Further, we find evidence of a safety-net effect of parental income which appears to be primarily compositional in nature.