A warning on the use of the Cochrane-Orcutt procedure based on a money demand equation

Abstract

Betancourt/Kelejian [1981] have recently warned against using the Cochrane-Orcutt procedure in models which include a lagged endogenous variable because this procedure can have more than onefixed point even asymptotically. FollowingSargan [1964], we argue instead that fixed points are not necessarily minima and that the question of practical importance is whether the residual sum of squares can have multipleminima. Within this formulation of the problem, we provide the firstreal example of multiple minima obtainable by the Cochrane-Orcutt procedure — with or without a lagged endogenous variable — and use it to caution against routine use of this procedure.