Comparative analysis of the effect of nonspherical disturbances on the OLS and OLSA estimators using Monte Carlo

Abstract

This paper undertakes a comparative analysis of estimation efficiency of OLS and OLS-Adjusted (OLSA) in a small sample when the economic model contains a spherical disturbance. A covariance matrix estimator that can consistently estimate the covariance of the model parameters which have been receiving attention in the econometric literature in recent time is employed. But so much attention has been paid only to the asymptotic property ,while pure OLS has continued to be the dominating estimation technique in small samples even in the face of nonsperical errors. This paper examines the small sample properties of OLS-Adjusted for moving average (MA), autoregression (AR), autorregresivemoving average (ARMA), when the sample size is small and there is a nonsperical error term. Since monte carlo experiment remains one of the best approaches for empirical extermination of finite sample properties it is herein employed in examining the small sample properties of OLSA compared OLS. It is found that when the sample size was deliberately made small and the nonsperical error fixed at 0.4, 0.6, and 0.8 at lower error OLS dominates and as the error increases OLSA dominates.