Optimal Policy Design
I have my reputation to think of!

The rational expectations revolution in macroeconomics has emphasized the way that the current behaviour of economic agents such as households and firms depends on their expectations of future events, including future government policy actions. The essential feature of the rational expectations hypothesis is the argument that expectations should be modelled as if they are formed making optimal use of all available information. Knowledge of how the economy works should therefore be used in calculations involving expected future events. So, for example, if inflation were known to be related to the money supply, rational agents would not forecast inflation by extrapolating past inflation rates, but would instead forecast the future path of the money supply and compute the future inflation implied by it, using the known relationship between the two.

This aspect of rational expectations raises questions concerning optimal government policy, first highlighted by Kydland and Prescott in their 1977 discussion of the time inconsistency of optimal plans. In the field of macroeconomics, the most straightforward example of time inconsistency concerns monetary policy and inflation. Suppose that in the short run money wages are fixed, but that in the medium and long run the rate of wage inflation depends, indirectly, on the expected rate of monetary expansion. The government has an incentive to announce an expansionary monetary policy initially, since with fixed money wages a higher money stock will cause higher employment and output. The policy should be much tighter subsequently, since in the long run faster money growth merely induces more rapid growth of wages and prices but no change in output or employment. Thus the government may want to have a monetary policy which is 'loose' now but 'tight' in the medium and long run. This may be the optimal policy for the government and one to which it would commit itself if it could.

Without such a precommitment, however, this policy would be difficult for the government to sustain because there would always be a temptation to introduce an unexpected temporary loosening of monetary policy to raise output and employment in the short run.

In Discussion Paper No. 124, David Backus and Research Fellow John Driffill consider whether a government's concern for its 'reputation' may persuade it to stick to the optimal policy of tight money. The private sector may interpret a deviation from this policy as a sign of loose money in the future, thereby inducing expectations of future inflation, high wage increases and high interest rates. Once lost, a government's 'reputation' for sticking to an announced tight money policy may be difficult to recover, and a temporary relaxation of policy, although attractive if the private sector is convinced that it is merely temporary, may become very unattractive if it leads to a loss of reputation and a long period of high inflation or higher unemployment in the future.

Backus and Driffill present a formal analysis of this argument for a very general linear model, going beyond the illustrative examples explored in earlier work (see Discussion Paper No. 63). They show that a government's concern to preserve its reputation can impose some discipline on its policy choices.

The authors demonstrate that the effects of reputation emerge most clearly in models where the economy is subjected to a sequence of random shocks, such as OPEC price increases, unexpected financial innovations, or changes in foreign fiscal and monetary policies. They also find that the government must attach sufficient importance to the long-run success of policy if reputation effects are to influence its policy choices. Similar results have been obtained by Currie and Levine (Discussion Paper Nos. 94 and 102), using a continuous-time model. Backus and Driffill's analysis uses discrete time methods, and so is complementary to the work of Currie and Levine.


The Consistency of Optimal Policy in Stochastic Rational Expectations Models
David Backus and John Driffill

Discussion Paper No. 124, August 1986 (IM)