In recent years, methods of 'optimal control' have been increasingly applied to macroeconomic models that incorporate forward-looking, rational expectations. In such models, individuals are assumed to make explicit and 'rational' calculations concerning future events, and in particular about what the government is likely to do. This is a complication which does not arise in engineering applications of control theory, where the system to be controlled is mechanistic; nor in the early applications of control theory to macroeconomic policy design, in which expectations were assumed to be formed in a backward-looking or adaptive fashion.

Forward-looking behaviour by the private sector may enlarge the range of policies available to the government. To see this, consider the case of a government that wishes to lower inflation through 'tighter' monetary policy. It may attempt this by raising short-term interest rates in order to reduce aggregate demand, or by inducing an appreciation of the exchange rate in an open economy. In the absence of forward-looking expectations by the private sector, such a policy will only begin to lower inflation when the policy is actually enacted. But with forward- looking expectations, the mere announcement of a future tightening of monetary policy is sufficient to lower inflation, if the announcement is believed by the private sector.

How does this occur? The expectation of a tighter monetary policy in the future will immediately raise longer-term real interest rates, thereby dampening aggregate demand. The interest rate increase will lead to an appreciation of the exchange rate, which will tend to reduce increases in domestic prices and wages. Thus the announcement of tight money in the future, if it is believed, has the effect of reducing inflationary pressures immediately. One example of such an 'announcement effect' is the UK Medium Term Financial Strategy. In announcing a phased reduction of monetary growth over a five-year horizon, the MTFS combined announcements of future tightening of monetary policy with an immediate policy action.

One question which has attracted considerable attention is whether such announcements of future policy action are generally credible. The standard answer has been negative. There will be, it is argued, an incentive for the government to renege on the announcement when the time comes to carry it out, even if no unforeseen event has occurred in the meantime. Such policies are said to be time-inconsistent. In the monetary policy example, if the announcement of tighter monetary policy in the future does indeed bring down the rate of inflation, then the government's incentive to tighten monetary policy in the future will no longer exist! The private sector, if it is sufficiently astute, can pursue this line of reasoning for itself and consequently the initial announcement will not be believed. Thus, it is argued, time-inconsistent policies which seek to exploit the power of policy announcements are not credible. Policy, in order to be credible, must be time-consistent and must not incorporate any future incentive to renege. Previous research has found that the requirement of credibility is costly: stabilization policies perform relatively poorly when they are obliged to be time- consistent in order that they be credible.

In Discussion Paper No. 94, Paul Levine and Programme Co-Director David Currie argue that issues of credibility and time consistency may be less important than was previously thought: announcements of future policy changes are generally credible in a variety of plausible circumstances. It is essential, Currie and Levine argue, to view policy-making as a 'repeated' game in which the government responds to a series of shocks to the economy. This contrasts sharply with the 'one-shot' policy game which Currie and Levine argue is implicit in many discussions of time consistency.

The prospect of continuing shocks to the economy affects the government's incentives to adhere to its policy announcements. Currie and Levine argue that the government's temptation to renege on policy announcements made in response to past disturbances is tempered by the fear that its credibility will be lost and that policy will be less able to cope with future disturbances. They demonstrate that if policy-makers attach sufficient importance to future disturbances, i.e. they do not discount the future too heavily, then the recurrent nature of the policy problem may well eliminate any incentive to renege. As the private sector will realize this as well, policy announcements should be rendered both credible and sustainable. In terms of the above example, the government will not renege if it expects the economy to be subjected to further inflationary disturbances to which policy must react. This argument will be strengthened if a government's reputation extends across a range of its activities, so that reneging on macroeconomic policy affects its credibility in other policy areas also.

Currie and Levine explore this argument further, carrying out a simulation exercise using a small open-economy model. They first calculate the fully optimal monetary and fiscal policy rules. These policy rules would be time-inconsistent in the conventional setting of 'one-shot' policy-making. In the repeated-game setting, however, their simulations confirm that the fully optimal policy is sustainable and credible under plausible conditions, provided that the government's rate of discounting for the future is not extremely high. Currie and Levine argue that the fully optimal rule may be of practical interest after all, even though it has tended to be ignored in discussions of policy design because of its time inconsistency. Since the benefits from adopting the fully optimal policy rule are typically much greater than for the optimal time-consistent rules, the sustainability of the fully optimal rule is of considerable importance for macroeconomic policy design.


Credibility and Time Inconsistency in a Stochastic World
David Currie and Paul Levine

Discussion Paper No. 94, February 1986 (IM)