Estimates of the Effectiveness of Monetary Policy
Fair, Ray C., Journal of Money, Credit & Banking
MANY OF the recent studies examining monetary policy effects have used macroeconomic models with a particular feature. In these models, an increase in inflation with the nominal interest rate held constant is expansionary, and the economy is not stable unless the coefficient on inflation in the nominal interest rate rule is greater than one. (1) Although these models, which will be called "modern-view" models, have been widely used, it is not clear that they have adequately captured the effects of inflation shocks. The results in Fair (2002), which are based on a structural macroeconometric model (discussed below), suggest that an increase in inflation is contractionary even when the nominal interest rate is held constant. Essentially the same results are reached in Giordani (2003) from analyzing VAR models. The results from these two quite different approaches thus cast doubt on a key property of modern-view models. They suggest that the coefficient on inflation in the interest rate rule does not have to be greater than one for the economy to be stable, which, as will be seen, has important consequences for monetary policy.
This paper uses the multicountry econometric (MC) model in Fair (1994) to examine monetary policy effects. The MC model has been extensively tested, including tests for rational expectations, and it appears to be a good approximation of the economy. These tests are in Fair (1994) and on the website mentioned in the introductory footnote. The MC model is briefly outlined in the Appendix. The Appendix includes an explanation of the property of the model that a positive inflation shock with the nominal interest rate held constant is contractionary, which, as mentioned above, is a key difference from modern-view models. In the MC model, a positive inflation shock lowers real wage income (because wages lag prices) and real wealth (because nominal wealth lags prices), both of which have a negative effect on real consumer expenditures. In addition, the estimation results suggest that households respond to nominal interest rates rather than real interest rates. There is thus no estimated positive household response to lower real interest rates when there is a positive inflation shock, with nominal interest rates held constant.
Section 1 examines the stabilization features of four interest rate rules for the United States. The first is the estimated rule in the MC model, which has an estimated long run coefficient on inflation of 1.0. The other three rules are modifications of the estimated rule, with imposed long run coefficients on inflation of 0.0, 1.5, and 2.5, respectively. It will be seen that as the inflation coefficient increases, there is a reduction in price variability at a cost of an increase in interest rate variability. Even the rule with a zero inflation coefficient is stabilizing, which is contrary to what would be obtained using modern-view models, since in these models, the economy is not stable if the inflation coefficient is less than one.
Section 2 then computes optimal rules for particular loss functions. These solutions require a combination of stochastic simulation and solving deterministic optimal control problems, and this is the first time that such solutions have been obtained for a large-scale model. It will be seen that the optimal control results are similar to those obtained using the estimated rule mentioned above for a loss function with a much higher weight on inflation than on output.
Another feature of the results in Sections 1 and 2 is that considerable variance of the endogenous variables is left using even the best interest rate rule. Section 3 then adds a fiscal policy rule--a tax rate rule--to see how much help it can be to monetary policy in trying to stabilize the economy. The results show that the tax rate rule provides some help. This is also the first time that such a rule has been analyzed using a large-scale model.