BIAS REDUCTION IN A DYNAMIC REGRESSION MODEL: A Comparison of Jackknifed and Bias Corrected Least Squares Estimators

Abstract

Employing small—sigma asymptotics we approximate the small—sample bias of the ordinary least—squares (OLS) estimator of the full coefficient vector in a linear regression model which includes a one period lagged dependent variable and an arbitrary number of fixed regressors. This bias term is used to construct a corrected ordinary least—squares (COLS) estimator which is unbiased to 0( cr2) . We also consider another technique for bias reduction, viz. jackknifing, and we present a simple expression for the JOLS(m) estimator: the m — delete jackknifed OLS estimator. Then we compare • the accuracy of the 0( cr2) approximation to the bias and the efficiency of OLS, COLS and JOLS(m) in a Monte Carlo study of artificial but realistic models. It is found that the bias is extremely sensitive to the value of a and that COLS can reduce it considerably without undue loss of efficiency if the standard deviation of the OLS lagged dependent variable coefficient estimate has a moderate value. (This abstract was borrowed from another version of this item.)