RDP 2000-10: Monetary Policy-Making in the Presence of Knightian Uncertainty 1. Introduction

One of the stylised facts of monetary policy is that interest rates generally move in a succession of small steps in the same direction and can remain fixed at the same level for some time. However, models of optimal monetary policy behaviour tend to generate much more volatile paths of interest rates, in which policy reversals are frequent. Clearly, central bankers are not behaving the way the models predict they should. This indicates that there is a need to reconcile model-generated and observed behaviour.

Recently, there has been a growing body of literature that has focused on the role of uncertainty and its effects on monetary policy decisions to this end. One line of this research has shown that some of the volatility in interest rate paths generated by models can be eliminated by considering uncertainty arising from the estimation of the model's parameters. In this case, the distribution of possible outcomes is determined by the probability distribution of the parameter estimates. Consequently, the standard assumptions underlying expected utility theory apply, and the methods of solving these decision problems are essentially the same as those that are used when parameter uncertainty is not a part of the decision-making environment.

A second line of research considers the consequences for interest rates of model uncertainty, and finds that the policy-maker is likely to be more responsive to changes in the economy rather than less in this environment. Another line of research has found that policy-makers, faced with uncertainty about the data they have available, either because of measurement issues or the possibility of future revisions, will tend to act more cautiously. The feature that distinguishes data and model uncertainty from parameter uncertainty is that there is no straightforward way of characterising these types of uncertainty with a unique probability distribution. Following the literature, we label this Knightian uncertainty.

An important implication of allowing for the possibility that future outcomes for the economy cannot be characterised by a single probability distribution is that the policy-maker can no longer be characterised as an expected utility maximiser. This paper discusses two alternative ways of approaching decision-making in the presence of Knightian uncertainty and the likely consequences they would each have for the observed path of interest rates. We conclude that an appropriate formulation of the monetary policy decision-making process in the presence of Knightian uncertainty has the potential to explain the inertia in the level of the interest rate observed in actual interest rate paths.

The rest of this paper is structured as follows. In Section 2, we review the current literature that seeks to reconcile observed and model-generated interest rate paths. In particular, we discuss the different forms of uncertainty that have been investigated, and use the results of this literature to motivate our interest in Knightian uncertainty. In Section 3 we make the distinction between risk, which can be thought of as uncertainty characterised by a unique probability distribution, and uncertainty, where such a characterisation is not possible. This is followed in Section 4 by a discussion of the two existing approaches to modelling decision-making under Knightian uncertainty. In Section 5 we consider the applicability of each of these approaches to the monetary policy decision-making problem and in Section 6, we discuss how the standard tools of optimal control can be adapted to allow us to operationalise the form of Knightian uncertainty we argue is most appropriate. We also show simulations of a simple closed-economy model to formalise the intuition presented in Section 5. We conclude in Section 7 by assessing the extent to which incorporating Knightian uncertainty into optimal monetary policy models has allowed us to reconcile model-generated interest rate paths with those observed in the real world.