RDP 9007: Operating Objectives for Monetary Policy 4. The Information Requirements of Monetary Policy

The preceding analysis has ignored an important problem associated with the direct targeting of final objectives: this is that the target variables are not accurately observed as policy is being made. Rather, the policy instruments must be adjusted in response to imperfect signals. In principle, there would seem to be two ways of allowing for this in the design of simple policy rules. First, one could make the policy instrument respond to lagged, rather than current, values of the target variable. However, this can easily be seen to lead to unsatisfactory results. Suppose, for example, the targeting rule defined by equation (11) is modified to be of the form

Then the solution equation for prices (given by equation 12 above) becomes

from which it can be seen that the condition γ>0 no longer guarantees convergence of the price level. The policy rule fails to anchor price expectations because it is responding to information that is no longer relevant to the determination of current prices.

This indeterminacy result is of course a consequence of the rational expectations assumption, and presumably the problem can be overcome if there is sufficient price rigidity in the model. Even then, however, the results are not entirely satisfactory, as the following simple example shows. Suppose the inflation process is generated by

where y_t is excess demand, and excess demand is regulated by the interest rate

If the policy rule responds to current prices, we have R_t = γp_t, which leads to the solution p_t(1+αβγ) = p_t−1. As before, any arbitrary positive value for the policy parameter γ ensures long-run stability of the price level. This is not the case, however, if policy responds to lagged prices:

here the solution is

which is unstable for sufficiently large values of γ. Choice of the policy parameter thus requires a knowledge of the economy's structural parameters in order to ensure stability, and this would remove much of the attractiveness of operating policy according to simple rules of thumb.

An alternative approach to the problem of imperfect information is to retain the assumption that policy responds to current movements in the target variables, but to replace the actual values of those variables with estimates. Suppose that instead of observing the actual price level p_t,^[11] the authorities observe a signal , generated by

The optimal estimate of the price level conditional on the information in the signal is

where

An operating rule for policy can then be defined by

In terms of the model introduced in Section 2, this rule has equivalent long-run properties to those defined by equation (12), and thus ensures long-run price stabilisation for any positive value of γ. In the short run, however, such a rule introduces the equivalent of a stochastic shock to monetary policy, associated with measurement error of the final objective. This problem does not affect the fixed exchange rate case, where there is no important measurement error, but adds to the short-run variances of both prices and output under the two nominal targeting rules. Their attractiveness as alternatives to a fixed exchange rate is therefore diminished.

It might also seem that the imperfect information problem provides an argument for reverting to rules based on monetary aggregates; however, in the context of the model under discussion here, this is not the case. Consider the following three rules.

1. Fixed money supply.

   Assuming the money demand function is of the form

   
   View MathML

   a fixed money supply rule implies that the interest rate is determined by

   
   View MathML

2. Money supply target.

   Under this rule the interest rate is assumed to adjust according to

   
   View MathML

   Combining this with the money demand function, the equation for the interest rate is

   
   View MathML

3. Nominal income target. R_t = γE(p_t+y_t).

Expressed in this way, it can be seen that rule (ii) is in fact a restricted version of rule (iii), because the money supply can be interpreted as one possible component of a signal providing information about nominal income. Whatever the information problems associated with estimating current nominal income, the nominal income target.

It should further be noted that rule (i) is a restricted version of rule (ii), because the response parameter γ in rule (ii) can be chosen optimally by the authorities, rather than being fixed. The money supply target must therefore perform at least as well as a fixed money rule. This implies that the rules can be unambiguously ranked in the following order of preference: nominal income target, money target, fixed money supply. It can be concluded from this discussion that although measurement errors worsen the absolute performance of nominal income targeting, they do not worsen its relative performance compared with rules based on the quantity of money.

The imperfect information problem also has an important bearing on the optimal choice of the policy response parameter γ. This parameter can be thought of as representing the degree of “activism” in a policy regime: a high value indicates a high willingness to move interest rates in response to any given piece of information. If information about final objective variables was perfectly accurate and up-to-date, it would be theoretically possible to achieve perfect stabilisation of the targeted variable by making γ arbitrarily large. This is not optimal, however, under imperfect information, because the benefits of a more activist policy must be traded off against the additional variability introduced by making policy respond too much to inaccurate signals. This argument can be formalised (Edey, 1989) by saying that the optimal degree of policy activism decreases as the accuracy of information about the final objective variables decreases.

Footnote

This argument applies equally when the target variable is nominal income. [11]