RDP 2006-02: Term Structure Rules for Monetary Policy 5. Conclusion

In this paper I have studied the implications of using long-term nominal interest rates in two types of monetary policy rules. Under the first, type-1 rules, the monetary authority adjusts the short rate in response to movements in some long-term yields as well as output and inflation. There are plausible regions of the policy-parameter space for which a unique stationary REE arises under such policy rules. However, normative analysis reveals that in the context of a simple New Keynesian model there are no significant gains from using type-1 rules in terms of reducing the value of the loss function. Surprisingly, the optimal parameter value of the reaction to long-term rates turns out to be negative for a range of plausible preferences, contradicting the initial intuition that recommends such use.,

Under the second use, the monetary authority conducts policy according to a type-2 rule, which is like a Taylor rule but with the short rate replaced by a longer-term rate. Mathematically this is equivalent to a more complicated rule for the short-term rate constructed such that the long-term rate would move in accordance with a Taylor rule. There are a number of surprising aspects of this proposal. First, significant regions of the policy-parameter space exist where a unique stationary REE obtains. Those policy parameters that yield a unique REE can be characterised as satisfying a generalised version of the Taylor principle – namely, that the long-run reaction of the instrument to movements in inflation should exceed one.

Second, type-2 rules can be shown to be better under certain central bank preferences than the standard Taylor rule. In particular, when the relative concern for output variability is relatively low, medium or long-term interest rate rules turn out to yield a better outcome. That is, the choice of maturity length for the rule is sensitive to the preferences of the central bank.

Third, even when preferences are such that optimal use of a Taylor rule outperforms type-2 rules, the latter seem to be more forgiving of ‘mistakes’ in setting the parameter values of the rule.

It is worth noting that these results hold under the pure expectations hypothesis (PEH) of the yield curve. This is not to say, however, that type-2 rules would not be preferable to Taylor rules when the PEH does not hold. Indeed, it may be that both type-1 and type-2 rules provide a useful way to respond to important additional information in the yield curve in a way that is not achieved by standard Taylor rules when the PEH fails. Such a possibility is left for future research.