抄録
In this paper, we examine small sample properties of a pre-test estimator of the disturbance variance when we have extraneous information on the disturbance variance and when there is a possible specification error in a specified model. It is shown that the pre-test estimator dominates the usual (OLS) estimator if the level of the pre-test is chosen appropriately, when the alternative hypothesis in the pre-test is one-sided.
