FURTHER EVIDENCE AS TO THE RELATIVE EFFECTS OF MONETARY VERSUS FISCAL POLICY *

SINCE 1963 the debate over the relative impact of monetary versus fiscal policy has from time to time reappeared in the literature as further evidence has been offered. The debate initially was inaugurated by Friedman and Meiselman in 1963 only to abate with failure to reach agreement as to an acceptable autonomous expenditures variable to be used in the test.' Andersen and Jordan refuted the debate in 1968 with a persumably acceptable expenditures variable, full-employment expenditures, but with the monetary variable under attack by their respondents.2 The research in this study constitutes an additional contribution to the debate as well as a clarification of some of the issues involved. Specified in this study was a simple Keynesian model solved for its reduced form under three alternative sets of assumptions regarding the monetary variable to be used as exogenous to the model. Each of the models contained similar fiscal variables including government expenditures on goods and services, government transfer payments, and government tax receipts. The impact of fiscal action due to changes in the tax structure was obtained by use of the non-linear tax and transfer equations of the Federal Reserve-MIT econometric model of the United States.3 Inasmuch as the tax functions specified in the real sector of the model were linear, these non-linear functions were used to obtain an approximation of the changes in tax receipts and transfer payments due to discrete linear changes in the tax structure. Those taxes and transfers, as well as expenditures, for which an equation was not available in the Federal Reserve-MIT model, were assumed exogenous and were measured as first differences in the data. The three alternative monetary variables used included the money stock, the monetary base, and the rate of interest on 4-6 month commercial paper. In cognizance of the lagged response in behavior pending fiscal or monetary action, a polynomial distributed lag technique was used in estimation. The technique used was that of Shirley Almon, who suggested that the method of Lagrange for polynomial interpolation, in conjunction with ordinary least squares be used for estimation of a weighted lag structure. Using such a technique solution for the reduced forms of the model in this study reveals that the choice of a monetary indicator bears heavily upon the parameter composition of the multiplier. Comparison of money multipliers estimated using differing monetary indicators demonstrated clearly that grievous misinterpretation could result if due cognizance of these parameters is not taken. An example of this difficulty of parameter identification exists in the Andersen and Jordan article where two alternative monetary indicators yielded widely differing results. Solution of the reduced forms using these indicators render quite different implicit parameters in the structure of the two multipliers. Empirically, the results of estimation of the model confirm the importance of money as a determinant of the level of GNP. Stability of the estimates of the multipliers over four alternative observation periods support this conclusion. The role of the fiscal