Abstract

This note describes gmmcovearn a user-written Stata package that performs GMM estimation of the covariance structure of earnings for a variety of models. The program decomposes the variance of earnings into a permanent and transitory component using the GMM estimator. The program incorporates both time factor loadings and cohort factor loadings on both the transitory and permanent component, allows the transitory component to follow either an AR or an ARMA process and allows for random heterogeneous growth in the permanent component. The program is used in recent papers by Doris et al (2010a, 2010b).