Estimators for a simultaneous equation model with lagged endogenous variables and autocorrelated errors

Two full information estimators and a limited information estimator for the simultaneous equation model with autocorrelated errors are studies by the Monte Carlo method.The estimators share features of the two-step Gauss-Newton procedure and of Aitken generalized least squares.One full information method generates the estimated endogenous variables used in the later stages of computation from the unrestricted reduced form while the other uses a restricted reduced form.The observed small sample behavior was close to that suggested by asymptotic theory.