A Bivariate Distribution Regression Model For Intergenerational Mobility

Abstract
This paper investigates intergenerational mobility by modeling the joint distribution of parents' and children's earnings. Using a bivariate distribution regression framework, we (i) flexibly account for the effect of observables on the distribution and (ii) assess the role of the remaining local correlation. The latter indicates how strongly earnings correlate at a specific location in the distribution once controlled for covariates. Comparing the observed distribution with a potential counterpart with no remaining correlation, we address the effect of unobservables throughout the distribution. Exploiting administrative data from Switzerland, we find that the distribution of fathers' and sons' earnings considerably differ from the father's and daughters. In particular, the local correlation's magnitude and shape vary substantially.