Does Time-Inconsistency Really Matter?

Abstract

This paper considers the optimal policy design problem for an econometric model with forward-looking agents. It suggests a new approach to the question of how policymakers should react to new information. This is seen to have implications for the question of whether time-inconsistency is a practically important constraint on the policy design process.It is well established that the fully optimal control strategy in a deterministic environment will be time-inconsistent in the absence of full credibility of policymakers. This optimal strategy will comprise a feedback rule on the state variables of the model plus a feedforward term on future information. Moreover, in the face of additive stochastic errors, the fully optimal strategy will be certainty-equivalent when the problem is linear-quadratic. In practice, however, models are not linear, uncertainty does not conform to these error processes and most importantly, the information set available to policymakers is continually changing. Thus, in the face of evolving expectations about model specification and the future behaviour of exogenous variables, there will always be two separate incentives for policymakers to re-optimise; one on the usual time-inconsistency grounds, the other on the basis of new information.This paper demonstrates how policymakers might implement a policy which assimilates this new information in an optimal manner without reneging or losing credibility in the usual sense. Importantly, this does not require policymakers to announce, ex ante, their future reaction to all possible contingencies nor does it require pre-commitment to a sub-optimal feedback law. The paper concludes, first that previous work on the sustainability of different types of policy rules has been flawed because it has confused these two different incentives to renege, and second that in the face of the typical shocks which do impinge on the economy, the gains to be had from re-optimisation may be dominated by those arising from the reaction to new information, rather than through the exploitation of the time inconsistency property.