RDP 9007: Operating Objectives for Monetary Policy 5. Conclusions

The paper essentially argues for four propositions.

1. It is possible to develop a coherent framework for monetary policy based on the short-term nominal interest rate as the policy instrument. In other words, one does not necessarily have to rely on control of either a monetary quantity or an exchange rate to anchor the system.
2. Within such a framework, it is essential that the policy rule be specified so as to respond to a variable that contains a nominal component. This requirement could be satisfied either by strict targeting of the price level or, more generally, by choosing a target variable which is some combination of nominal and real components, such as nominal income.
3. There are important practical reasons for favouring policy rules that are relatively simple. Simple rules help to focus expectations, and place less stringent information requirements on the authorities than complicated ones.
4. These arguments provide some support for the use of nominal income as a guide for policy. As a matter of formal modelling, the paper has argued that nominal income targeting provides relatively good short-run stabilisation properties, against a range of shocks, when compared with the obvious alternatives.

Putting aside the question of what variable might be chosen as a target, it is important to recognise both the similarities and differences between an approach based on simple targeting rules, and one which attempts to design a fully optimal policy. Both approaches can, in principle, require the authorities to “look at everything”: the former, because a range of imperfect signals must be used in estimating what is currently happening to the target variable; and the latter, because it is generally true that optimal policy responds to all shocks.

The essential difference between the approaches lies in the way that information is interpreted. On the one hand, to construct an optimal policy one would have to know the full dynamic structure of the macroeconomic system, including the variances of all unobservable shocks, and then use these to calculate the optimal response weights to all sources of information. It would be hard to argue that this is a feasible programme in practice and, in the absence of the required structural knowledge, there is no clear principle for determining the response weights to be used. A simple targeting rule, however, does provide such a principle. Policy responds to only one, possibly composite, variable, namely the current estimate of the chosen target. A wide range of signals is still looked at, but these are responded to only for the information they provide about the path of the target variable; in other words, policy responds to a signal to the extent that it causes revisions to the estimates of the variable being targeted. In this way, simple targeting rules would seem to gain important advantages in terms of clarity and transparency, although admittedly at the cost of giving up some degree of flexibility.