BIAS ASSESSMENT AND REDUCTION IN LINEAR ERROR-CORRECTION MODELS

The small-sample bias of the ordinary least-squares coefficient estima tor for dynamic regression models with innovation errors and lagged-de pendent and strongly-exogenous explanatory variables is approximated t hrough both small disturbance and large-sample asymptotics. Results fo r the standard ARMAX(p, 0, k) model are obtained and also for such mod els under linear parameter constraints and variable transformations. T hese approximations are then used to construct corrected estimators fo r the parameters of interest in higher-order dynamic models, including the empirically highly relevant linear error-correction model. By sim ulating two empirical cases the corrected estimators obtained via larg e-sample asymptotics are shown to have more attractive location and ef ficiency properties than ordinary least-squares.