RDP 7903: Monetary Rules: A Preliminary Analysis 1. Introduction

It is often possible, with the benefit of hindsight, to see how different responses of economic policy could have produced a more favourable outcome for a particular historical episode. Thus, for example, Friedman and Schwartz (1963) suggest that maintaining steady and moderate growth of the US money supply during the 1930s would have meant more activity and less unemployment It can readily be accepted that avoiding very large increases or decreases in the money supply is sensible. Economists have, however, wondered if it is possible to do better than this. They are also concerned to test the general validity of simple policy rules based on the evidence of particular historical episodes.

One response is the literature on optimal control. In the light of severe doubts about the accuracy of the econometric models used to calculate the “optimal” responses, this highly complex procedure does not appeal to practical policy makers. A simpler approach is to calculate the effect of alternative rules, usually in a particular theoretical model. Thus, for example, Poole (1970) shows that with a simple LM-IS model it is sensible to fix the money supply when the shocks are “real” – that is when they consist of an exogenous change in a component of aggregate real demand – and to fix the interest rate when the shock is “financial” – that is when there is an exogenous change in the demand for money.

There are two practical problems with this approach. The first can be called the information problem. The economy is much more complicated than the static LM-IS model. Policy actions have lagged effects and in addition it is often difficult to know precisely where the disturbances are coming from, or, more generally, what is the mix of disturbances.

The second problem can be called the underwriting problem. One of the ways in which the economy is dynamic is that past policies influence the future. Thus, for example, many economists believe that over-emphasis on stability of the exchange rate and nominal interest rates in the 1950s and 1960s allowed the growth of money to accommodate various disturbances (such as large wage increases) and fostered a mentality in which such disturbances were more likely to occur. Grappling with these practical problems requires empirical analysis.

This paper tests the effects of three simple rules for monetary policy in a moderately complex econometric model of a real economy. The study by Cooper and Fischer (1972) using the M.P.S. model of the U.S. economy provides evidence most directly comparable with that of the present study. In contrast to the earlier study, the present one makes its main contribution by examining the effects of different rules for monetary policy following a series of exogenous shocks to the economy.[1] A secondary contribution is to the debate on the controllability of the money supply by variations in interest rates under official control.[2]

The model used in this paper, called RBA79, represents the Australian economy.[3] The most important difference between RBA79 and the M.P.S. model is perhaps the way in which the excess demand or supply of money has a direct impact on changes in expenditure and, more important, on inflation. The results of excluding the direct effect of money on inflation, and of making prices generally less variable, are examined in a sensitivity analysis.

As with any empirical study, the results must be treated carefully. The reactions illustrated in this paper represent the average for the 1960s and some of the 1970s. In any particular episode the results could be different. Also, it is conceivable that the exogenous shocks considered or the changes of policy regime allowed for could change the reactions of the economy in ways not allowed for in the model. Too much should not be made of this point, however, since the regimes and shocks considered in this paper produc reactions well within the range of recent experience. And, as argued by Clements and Jonson (1979), key equations of RBA79 are consistent with those used in analyzing the effects of rational expectations. However, the current results are perhaps best regarded as a contribut to a body of evidence which needs to be built up for a variety of models, just as the models themselves need to be tested rigorously.

The approach is taxonomic. A variety of inflationary impulses are simulated with three regimes for monetary policy. The impulses, which are probably best thought of as unanticipated shocks, consist of:

  • four shocks to components of real aggregate demand

    – a sustained increase in real government spending;

    – a sustained increase in real exports;

    – a sustained increase in the household saving ratio;

    – a sustained decrease in business fixed investment;

  • three supply side shocks

    – a sustained increase in minimum wages;

    – a sustained reduction in capacity output;

    – a sustained increase in world prices;

  • two financial shocks

    – a sustained decrease in the demand for money;

    – a sustained increase in capital inflow.

(Appendix B provides more precise definitions of the impulses.)

The regimes for monetary policy consist of:

  • a rigid bond rate. In this regime, the degree of monetary accommodation is usually high;
  • a bond rate which responds positively to monetary expansion, but allows some monetary accommodation;[4]
  • a bond rate which responds so strongly to monetary expansion or contraction that the degree of monetary accommodation is virtually zero except in the first three or four quarters.

(These regimes are discussed in more detail in Appendix B.)

Footnotes

The study by Cooper and Fischer examines the effects of different policy rules for a particular historical period. [1]

Cooper and Fischer examine the effect of different reaction equations for the growth of the money supply. [2]

Jonson and Taylor (1978) provide an account of the model and its use in the analysis of the inflationary surge of the 1970s. A fuller discussion of the model is in Norton (ed.), (1977), and the model specification is set out in Appendix A. [3]

In this case, the reaction function for the bond rate is that estimated in the model. (See equation 17 in [4]