On the asymptotic properties of some two-step procedures for estimating distributed lag models

It Is by now well known that in the presence of lagged dependent variables and serially correlated errors certain two-step procedures are not asymptotically as efficient as the method of Maximum Likelihood (ML). In the present paper we discuss three two-step procedures that have been suggested in the literature and show that not only are the two-step procedures inefficient when compared with the ML method, but in some cases may not be more efficient than the first step of the two-step procedure. The plan of the paper is as follows. In Section 2 we discuss a pair of two-step procedures for the estimation of the parameters of distributed lag models when the models contain lagged dependent variables. Section 3 contains a discussion of a two-step procedure which has been suggested for use when the distributed lag model is formulated in the distributed lag form, i.e., without lagged dependent variables. Section 4 presents the conclusions of the paper.