Partial Credibility and Policy Announcements: The Problem of Time Inconsistency in Macroeconomics Revisited
Reeves, Silke Fabian, Atlantic Economic Journal
Credibility problems are of concern in many areas of macroeconomic policymaking. Common to all applications is a two-period sequential move structure. In the first period, market participants must make irreversible economic decisions based only on expectations about future central bank or government policies. Policymakers only act in the second period, while they may announce future policy intentions during the initial period. They do not, however, commit themselves. Policymakers may, thus, renege on these announcements if there exist incentives to do so. Rational economic agents will recognize such incentives. As a result, policy signals will not be accepted at face value.(1)
Perhaps the most prominent application appears to be the design of monetary policy in light of an expectations-augmented Phillips curve. The studies launched by Barro and Gordon's [1983a, 1983b] analysis focus on the central bank's temptation to create surprise inflation in order to raise employment. However, this being anticipated by the market, the discretionary equilibrium exhibits a high rate of inflation without any gains in employment. Welfare would be higher, could the central bank commit to a lower (ideally zero) inflation rate.
In contrast, the credibility problem may be resolved in a repeated game context [Kreps and Wilson, 1982b; Milgrom and Roberts, 1982]. In a reputational model, the policymaker can mimic a "forthright type" who always honors policy announcements. Except for the final periods of the game, policy signals are always credible, thus removing the credibility problem. This result has been applied to a vast array of credibility problems in macroeconomics (for surveys see Persson and Tabellini  and Rogoff ).
The analysis of this paper builds on the work by Backus and Driffill , Barro , and Cukierman and Liviatan . These studies develop a reputational equilibrium in a situation where a central bank faces incentives to create surprise inflation but cannot commit to a zero inflation rate. If the duration of play is long enough and the policymaker sufficiently patient, the central bank implements and the market expects a zero inflation rate from the beginning. In the final periods the game reverts to the discretionary equilibrium.(2) Central bank announcements of zero inflation are immediately and fully anticipated by the market and in this sense perfectly credible.(3)
This does not, however, characterize most real-world situations and, as is shown here, game theory also predicts perfect credibility only as a special case. This paper argues that credibility is partial in the sense that the public attaches only a positive probability p [less than] 1 to the event that the policymaker will carry out preannounced economic policies. Perfect credibility refers to the case of p = 1.
In particular, in a reputational model the paper shows that, under very general assumptions, partial credibility is a sequential equilibrium of the repeated game.4 Interestingly, credibility is only partial even if forthright policies, where announcements are followed through with policy changes, are implemented in all periods. Moreover, reputation gradually improves over time but credibility remains partial throughout.
With partial credibility, policy surprises cannot be avoided, even with forthright policymaking. These surprises may be costly. The policymaker will, therefore, attempt to minimize the resultant welfare losses. This is exemplified by studying the case of the expectations-augmented Phillips curve discussed above. There, low inflation rates are not fully credible and, when implemented, lead to a loss of employment. Contrasting with the case of perfect credibility, the optimal policy calls for a higher rate of inflation to reduce the resultant unemployment. The inflation rate is gradually lowered as credibility rises over time.
The plan of the paper is as follows. …