A Monte Carlo Investigation of Econometric Models with Fixed and Stochastic Regressors where the Error Terms are AR(1) and MA(1).

摘要

A Monte Carlo study of four specialized regression procedures is reported. Two of the procedures are designed for a simple linear econometric model while the other two procedures are designed for a single period lagged endogenous variable and single exogenous variable econometric model. For the simple linear model, the small sample properties of the Pesaran procedure, designed for an MA(1) error term, and the Beach-MacKinnon procedure, designed for an AR(1) error term, are compared against the small sample properties of OLS and against those of the Prais-Winston procedure. For the lagged endogenous model, the small sample properties of the Zellner-Geisel procedure, designed for an MA(1) error term, and the Wallis procedure, designed for an AR(1) error term, are likewise compared against the small sample properties of OLS and of Paris-Winston. In addition, the power of the Durbin-Watson d test is analyzed for both models and the Durbin h and the McNown tests are analyzed for the lagged endogenous model.