How severe is the time-inconsistency problem in monetary policy?

Introduction The history of inflation in the United States and other countries has occasionally been quite bad. Are the bad experiences the consequence of policy errors? Or does the problem lie with the nature of monetary institutions? The second possibility has beene xplored ina long literature, which starts at least with Kydland and Prescott (1977) and Barro and Gordon (1983). This paper seeks to make a contribution to that literature. The Kydland-Prescott and Barro-Gordon literature focuses on the extent to which monetary institutions allow policymakers to commit to future policies. A key result is that if policymakers cannot commit to future policies, inflation rates are higher than if they can commit. That is, there is a time-inconsistency problem that introduces a systematic inflation bias. This paper investigates the magnitude of the inflation bias in two standard general equilibrium models. One is the cash-credit good model of Lucas and Stokey (1983). The other is the limited-participation model of money described in Christiano, Eichenbaum, and Evans (1997). We find that, for a large range of parameter values, there is no inflation bias. In the Kydland-Prescott and Barro-Gordon literature, equilibrium inflationin the absence of commitment is the outcome of anin terplay between the benefits and costs of inflation. For the most part, this literature consists of reduced-form models. Our general equilibrium models in corporate the kinds of benefits and costs that seem to motivate the reduced-form specifications. To understand these benefits and costs, we must first explain why money is not neutral inour models.