Second Order Bias Corrected Efficient GMM Estimator

B F Chakalabbi, Sagar Matur and Sanmati Neregal

Department of Statistics, Karnatak University’s Karnatak Arts College, Dharwad – 580001, India

Abstract: In this paper, the three conventional GMM estimators First-difference, Level and System GMM estimators with respective efficient initial weight matrices are considered to estimate the autoregressive panel data model. It is observed that the bias of first-difference GMM estimator is heigher and the bias of system GMM estimator is lesser among the above mentioned estimators but as variance ratio in-creases and as autoregressive parameter approaches to one the bias of all the afore-mentioned estimators increase. Hence to reduce such bias, second order bias correction method is considered. Through Monte-Carlo simulation it is observed that the considered second order bias correction method works well for first-difference and system GMM estimators, specially when the variance ratio is greater than one.

Keywords: First-difference GMM estimator, Level GMM estimator, System GMM estimator, Second order bias.

Pages: 54 – 75 | Full PDF Paper