Decomposing economic mobility transition matrices

The intergenerational mobility literature has consistently found that the distribution of adult economic outcomes differ markedly depending on parental economic status, yet much remains to be understood about the drivers or determinants of this relationship. Existing literature on potential drivers focuses primarily on mean effects. To help provide a more complete picture of the potential forces driving economic persistence, we propose a method to decompose transition matrices and related indices. Specifically, we decompose differences between an estimated transition matrix and a benchmark transition matrix into portions attributable to differences in characteristics between individuals from different households (a composition effect) and portions attributable to differing returns to these characteristics between individuals from different households (a structure effect). We also incorporate a detailed decomposition, based on copula theory, that decomposes the composition effect into portions attributable to specific covariates and their interactions. We illustrate our method using data on white men from the 1979 National Longitudinal Survey of Youth. Estimation is based on an extended Mincer equation that includes cognitive and non-cognitive measures. To address the potential endogeneity of education, we implement an IV strategy that allows us to estimate causal effects and investigate the role of unobserved ability.