Intergenerational Social Mobility in the United States: A Multivariate Analysis Using Distributional Regression

While previous studies have focused primarily on analyzing the spatial process of social mobility for the conditional mean, we model all parameters of a multivariate response distribution within a distributional regression framework and hereby extend the literature in several ways. First, we relax the assumption of independence between measures of absolute and relative mobility and study their interdependence by relating the strength of their association to economic factors. We find that indeed the dependence between the two measures is nonconstant and varies considerably, both with respect to economic variables and across the United States. Second, and following the recent trend of bringing the attention of economists and sociologists to analyzing the conditional variance, we investigate the within-district variability of social mobility via heteroscedastic regression using the distributional regression framework. Our results suggest a Great Gatsby curve for the variance, so that higher rates of income inequality increase the risk of being excluded from upward social mobility. Third, our analysis enables us to assess the ability of influential variables, as discussed in the literature, to capture the spatial patterns of social mobility across the United States. The results indicate that although the explanatory variables contribute to a better understanding of its underlying mechanisms, they leave spatial heterogeneity of social mobility mostly unexplained.