SUR Estimation of Error Components Models With AR(1) Disturbances and Unobserved Endogenous Effects

This paper focusses on the estimation of error components models in the presence of a correlation of the disturbances across equations and AR(1) of the remainder disturbances for panel data with endogenous unobserved effects. Additionally, the set-up allows for unequally spaced panel data and differences in the autocorrelation parameters across equations. The derived procedure is a feasible generalized least squares (GLS) estimator, which provides estimates of the variance components in the spirit of Hausman & Taylor (1981).