Robust Least Squares Dummy Variable Estimation Of Dynamic Panel Models In The Presence Of Outliers

Okeke Joseph Uchenna, Okeke Evelyn Nkiruka, Obi Jude Chukwura

This research is focused on the consistent, robust least squares dummy variable (LSDV_R) estimator which is predicated on the correction of the bias of the inconsistency of the least squares dummy variable estimator of the parameters of the dynamic panel data model, as an extension of earlier results. We compared the results of the bias corrected least squares dummy variable estimator of the dynamic panel data models in the presence of outliers, at stated specifications of the model with the consistent instrumental variable (IV) and the generalized method of moments (GMM) estimators of Anderson and Hsiao (AH), Arellano and Bond (AB) and Blundell and Bond (BB) to validate the claims or otherwise of the estimators. We observe at  and B=0.2 that the robust least squares dummy variable estimator (LSDV_R) performs better than the IV- GMM in finite and large samples in terms of predictive powers and in the estimation of the autoregressive coefficient in large samples followed by the LSDV, though, with maximum RMSE property while the Blundell and Bond (BB) performs better than the other contending models in estimation of the autoregressive coefficient in finite samples showing that the presence of an outlier does not affect the predictive power of the robust least squares dummy variable (LSDV_R) estimator.