Monte Carlo experiment on the pooled ols estimator in large mixed panels.

This paper investigates the behaviour of the pooled ols estimator in a panel model with stationary and nonstationary regressores as both N amd T go to infinity. the nonstationary regressor is assumed I(1) ,the stationary regressor is set i.i.d.The investigation is carried through four Monte Carlo experiments.The experiments show that in a model with no endogeniety the pooled ols estimator of the I(1) regressor is sqr(n)T consisten and asymptotically normal and the estimator of the I(0) regressor is sqr(NT) consistent and asymptotically normal.If a correlation between the I(1) regressor and the regression disturbance is introduced ,both estimators are inconsistent fro small n but sqr(nT) consistent and asymptotically normal for large n amd T.When an individual random effect is added to the basic model the results for the I(0) regressor coefficient do not alter ,whereas the estimator of the I(1) regressor loses in efficiency and becomes sqr(NT) consistent and asymptotically normal.