Estimation of Seemingly Unrelated Regressions with Autoregressive Disturbances

Abstract Joint estimation of a system of linear regression equations with mutually correlated disturbances leads, in general, to more efficient estimates than estimation of each equation separately. A method of joint estimation developed by Arnold Zeller on the assumption that the disturbances are non-autocorrelated has recently been adapted by Richard Parks for the case in which this assumption does not hold. In the present article we develop several alternative estimators and compare their small-sample efficiency. The comparison is done with the help of a sampling experiment applied to various model specifications.