RDP 2006-05: Optimal Monetary Policy with Real-time Signal Extraction from the Bond Market 4. Conclusions

This paper has presented a general equilibrium model of monetary policy where the central bank operates in an uncertain environment and uses information contained in the term structure to estimate the underlying state of the economy more efficiently. This set-up creates a link between the term structure and the macroeconomy that is novel to the literature. A movement in the term structure signals that a change in the short-term interest rate set by the central bank may be desirable, which when implemented, in turn affects aggregate demand. Söderlind and Svensson (1997) warn that ‘central banks should not react mechanically to [market expectations]’ since this may lead to a situation of ‘the central bank chasing the market, and the market simultaneously chasing the central bank’. This argument is formalised in Bernanke and Woodford (1997) where the authors show that if central banks react to market expectations, a situation with a multiplicity of equilibria or where no equilibrium exists may arise. In this paper we have argued that there may be benefits from systematic reactions to market expectations, but with some important qualifications. The non-existence of equilibria arises in the model of Bernanke and Woodford because the central bank can extract the underlying state perfectly from observing the expectations of the private sector. Inflation will thus always be on target. But if inflation is always on target and private agents only care about accurate inflation forecasts, there is no incentive for the private sector to pay a cost to be informed about the underlying shock, and observing expectations will not reveal any information. The model here differs because, to the extent that there is noise in the bond market, the central bank cannot extract the underlying shock perfectly. Thus there will always exist a cost of information-gathering that is small enough to make it profitable for the private sector to acquire information about the underlying shock, even if private agents only cared about having accurate inflation forecasts. Additionally, in this model the forecasting problem of private agents involves more than accurately forecasting inflation, since bond prices depend on real factors through the stochastic discount factor as well as the price level. Insofar as the real discount factor is affected by the underlying state, agents will have an incentive to collect information about it, regardless of the behaviour of inflation.

Ultimately, the informational content of the term structure is an empirical question. The model presented here provides a coherent framework within which any information about the state of the economy that is contained in the term structure can be quantified in a general equilibrium setting. The model explicitly takes into account that the central bank may use the information in the term structure to set policy and therefore influences what it observes. The model was estimated on US and Australian data using Bayesian methods. The empirical exercise suggests that there is some information in the US term structure that allows the Federal Reserve to respond to shocks in a timely manner, while the Australian term structure appears to be more noisy and less informative for the monetary policy process. This difference may be explained by the fact that Australia is a small and relatively open economy, and hence difficult to represent using a closed-economy model.