| Περίληψη : | Fixed effects estimator is a widely known estimator for panel data models whichcan be severely biased. The cause of this shortcoming is the so called incidentalparameter problem first stated by Neyman and Scott [14] for fixed T panels. Alvarezand Arellano [2] have shown that the asymptotic bias is reducing as T grows toinfinity and derived its asymptotic distribution. In practice growing T to infinityis not realistic assumption since most microeconomic panels have moderately smallT. Hence, a new approach was introduced by Hahn and Kuersteiner [6][5]whoconsidered alternative assumptions on the asymptotics, i.e that N, T grow at thesame rate, as an approximation which assist us on studying the bias term properties.As a result the bias reduction estimator is centered at the truth, whereas fixed effectsestimator is not. Due to the fact that bias reduction techniques require fixed effectsestimator to be already calculated, they are computationally heavy. While thismay be true, Monte Carlo studies have shown that they outperform many otherestimators while they keep the asymptotic efficiency.
|
|---|
Copyright © 2013 Βιβλιοθήκη Ο.Π.Α. Με την επιφύλαξη παντός δικαιώματος.