Estimating the Intergenerational Correlation of Incomes : An Errors in Variables Framework

Because the permanent incomes of parents and children are typically unobservable, the estimation of the intergenerational correlation of incomes is usually carried out via averaging methods or instrumentation. In this paper we take the permanent income of the parent family to be unobserved, but we assume that a model for its determinants is known to the researcher. In turn, this leads us to propose two related estimators for the intergenerational correlation: a two-stage least squares procedure and a more efficient MIMIC estimator. The MIMIC framework also provides estimates for the determinants of permanent income and the variance parameters required to evaluate the bias of the OLS estimator. Using a US sample of parents and children we provide estimates for the intergenerational correlation ranging between 0.30 and 0.78. The bias of the OLS estimator is estimated to be in the order of 40%.