A Unified Approach to Asymptotic Equivalence of Aitken and Feasible Aitken Instrumental Variables Estimators

Abstract

Asymptotic equivalence of Aitken and feasible Aitken estimators in linear models with nonscalar identity error covariance matrices is usually established in a tedious case-by-case manner. Some general sufficient conditions for this equivalence exist but there are problems with the extant conditions. These problems are discussed and new widely applicable sufficient conditions are presented and applied to a variety of error structures. Copyright 1994 by Economics Department of the University of Pennsylvania and the Osaka University Institute of Social and Economic Research Association.