Parameter selection in the Almon lag method: theory and practice

Abstract

The Almon lag method for estimating the weights in a distributed lag model has often been applied to the Monetary versus Fiscal Policy debate. Unfortunately, empirical results have differed greatly, largely on the basis of the parameters used with this method. It appears that with commonly used techniques of parameter selection, a researcher may easily locate evidence supporting his theoretical point of view. In this treatise we develop statistical criteria for parameter selection that eliminate the ambiguity surrounding the use of Almon's method. We consider in detail three areas of particular difficulty: (1) selecting the length of lag; (2) selecting the degree of polynomial; and (3) deciding whether to constrain the endpoints of the weights to be zero. Additionally, we apply our developed techniques to the Monetary versus Fiscal Policy debate to eliminate several of the contradictions that have arisen due to the use of the Almon lag method. ...