To evaluate the extent to which simple policy rules are robust to model uncertainty, we take first-difference rules from the policy frontier of one model, and evaluate the performance of these rules in each of the other three models. In particular, we consider the first-difference rules A and В from the corresponding policy frontier of the FRB model; the parameter values for these rules are given in Table 3. In the FRB model, both rules generate the same standard deviation of the first-difference of the funds rate as that generated by the estimated rule given in equation (1).

For each of the other models, we calculate the funds rate volatility associated with each of the two rules, and then we compute a separate policy frontier for each upper bound on funds rate volatility, using the class of 3-parameter rules given by equation (4). For example, the “Rule A” policy frontier for the MSR model is the 3-parameter policy frontier for rules with the same funds rate volatility that rule A generates in this model. For the FM model, rules A and В produce virtually identical amounts of funds rate volatility.

The results of this analysis are depicted in Figure 9, which shows that rules A and В provide reasonably efficient performance in stabilizing output and inflation in all four models. Conditional on the level of interest rate volatility implied by these rules, the coefficients of the rules are such that inflation and output volatility lie very near the 3-parameter policy frontiers of all models. Evidently, in terms of efficiently reducing the volatility of output and inflation, well-chosen simple rules are very robust to the type of model uncertainty encompassed by these four models. Nevertheless, while these results show that rules A and В are reasonably efficient, this figure does not indicate how well the rules perform in terms of specific values of A.

To evaluate the extent to which simple policy rules are robust to model uncertainty, we take first-difference rules from the policy frontier of one model, and evaluate the performance of these rules in each of the other three models. In particular, we consider the first-difference rules A and В from the corresponding policy frontier of the FRB model; the parameter values for these rules are given in Table 3. In the FRB model, both rules generate the same standard deviation of the first-difference of the funds rate as that generated by the estimated rule given in equation (1). For each of the other models, we calculate the funds rate volatility associated with each of the two rules, and then we