Abstract:
|
Estimation of the intergenerational income elasticity (IGE) with the short-run measures of earnings that are usually available is seriously affected by the "zeros problem," that is, the fact that those without earnings (due to unemployment or other forms of nonemployment) cannot be included in the analysis. This can be expected to generate a downward selection bias that should be particularly large when the earnings information comes from administrative records. We address this problem by relying on bivariate copula-based selection models to estimate the IGE of men’s earnings with data from the Statistics of Income Mobility Panel (a dataset built from tax and other administrative data). We estimate a very large number of selection models by maximum likelihood and use information criteria to select the best estimates. The selection models rely on various copulas and marginal distributions, including several distributions never used before in the copula-based selection literature. Our analyses confirm that the standard estimator badly underestimates the IGE, and indicate that, among U.S. men, about two-thirds of family economic advantages are transmitted through the labor market.
|