Rational Expectations
Credible promises

Governments face a credibility problem when economic agents form rational or model-consistent expectations, because the resulting policy is `time-inconsistent' in the sense that it is optimal when chosen but becomes sub-optimal thereafter, thus creating an incentive to renege. If the private sector has complete information, private decision-makers will anticipate the government's incentive to renege. A time-inconsistent policy can be made self-enforcing, however, if the private sector punishes the policy-maker in the event of reneging.
In an economy where output can deviate from its natural rate only as a result of inflation surprises, the `ideal policy' is zero inflation; but in any one period there is an incentive for the government to engage in surprise inflation. The all-knowing public can anticipate this, and the only credible policy is a time-consistent (or `discretionary') non-zero inflation rate. A lower rate of inflation can be sustained, however, if the private sector is prepared to believe that policy-makers' precommitment to the lower rate will continue once started, and to believe in the higher discretionary rate only for some given punishment period if the government does renege.
In Discussion Paper No. 409, Research Fellow Paul Levine and Joseph Pearlman note that this analysis, due to Barro and Gordon, entails a number of limitations. It applies only to an infinite time horizon, requires that the punishment period is exogenously given, and applies only to a specific model, which is essentially static.
They first note Cukierman and Meltzer's development of this analysis to include informational asymmetry. The private sector does not observe how highly the government values output and can only infer this by observations of money-supply growth. Further, monetary growth consists of not only a planned part but also a stochastic part as a result of either imperfect control or deliberate ambiguity. The optimal discretionary inflation rate is less than that for the time-consistent case. In other words, there are gains from governments' having private information. Also, if the policy-maker can choose the level of random noise in a policy instrument, then no noise is not necessarily the most desirable level.
Levine and Pearlman extend this analysis to a model that displays wage/price stickiness, which introduces a significant short-term trade-off between output and anticipated inflation. The role for surprise inflation is now less, and consequently the time-inconsistency problem is less acute. The roles of informational asymmetry and of ambiguous control of the money supply should therefore be correspondingly smaller, and Levine and Pearlman's results show that this is indeed the case.

Paul Levine and Joseph Pearlman
Credibility, Ambiguity and Asymmetric Information with Wage/Price Stickiness

Discussion Paper No. 409, April 1990 (IM)