CONSISTENT EMPIRICAL RESULTS WITH ALMON'S METHOD: IMPLICATIONS FOR THE MONETARY VERSUS FISCAL POLICY DEBATE

Abstract

IN 1968, ANDERSEN AND JORDAN reported that the response of economic activity to monetary actions is much larger and faster than its response to fiscal actions. Their conclusions triggered the extensive and still on-going monetary versus fiscal policy debate. (See [2, 3, 4, 5, 9, 10, 22, 24]). The Almon method for estimating distributed lags (See [1]), which serves as the estimating technique in the debate, is the focus of the recent contributions. Even when common variables and data sets are employed, often different empirical results follow from different choices of the degree of the polynomial and/or lag length. This sensitivity of results to parameter selection has posed a dilemma, for the profession has disagreed upon how to appropriately select these parameters. The lack of agreement has seriously impaired confidence in results obtained with Almon's method and has been instrumental in delaying satisfactory resolution of the monetary versus fiscal policy debate. Schmidt and Waud [22] first demonstrated the problems encountered with the Almon technique, arguing that an incorrect choice of either parameter is a specification error which leads to biased and inconsistent estimates. They conclude that when a priori knowledge of the parameters is not available... use of the Almon lag technique should be avoided. ([22, p. 13]). Only as a second-best solution did they advocate choosing Almon parameters by minimum standard error (MSE), which is the strategy they employed in their contribution to the debate. Frost [12], however, found that if one selects Almon parameters on the basis of MSE, the estimated lag weights and their variances are biased, and the distributions of the estimators deviate significantly from the normal distribution.' Most recently, Harper [16] has shown that even without a priori information reliable estimates are theoretically attainable with the Almon approach. The purpose here is to demonstrate how to rigorously apply these theoretical findings and to obtain statistically reliable estimates of the paths and magnitudes of the impacts of monetary and fiscal policy actions on economic activity. Section I outlines a procedure for choosing on the basis of six specification error tests the proper length of lag and the degree of polynomial in the Almon technique, so that the ensuing estimates are unbiased, consistent, and efficient. In Section II, these tests are applied to the original models of previous investigators, and the tests