Table 6 reports several summary statistics about the extent to which rules A and В are robust to model uncertainty. The fourth column reports the value of the objective function using the ‘true” value of A (assumed to be 0.25 for rule A and 0.75 for rule B, which are the values used to determine these rules in the FRB model). The fifth column reports the absolute loss in terms of the objective function implied by following the specified rule (A or B) instead of the three-parameter optimal rule (with the same amount of interest rate volatility) for the specified value of A. The last column reports the value of A which would be consistent with the choice of rule A or В for the particular model; e.g., in the first row, the implicit value of A = 0.1 means that policymakers with preferences described by A = 0.1 would choose rule A in the FM model.

These results indicate that if a policymaker were to use the FRB model to choose a policy rule, but the real world were actually described by one of the other three models, the policy rule would generate slightly greater output volatility and slightly less inflation volatility compared with the preferences of the policymaker.

In particular, the implicit values of A associated with rules A and В are all smaller in the non-FRB models than the “true” values used in choosing rules A and В in the FRB model. However, while rules A and В are suboptimal, the loss in terms of the objective function, measured either in absolute or percentage terms, is quite small in the MSR model and in TAYMCM. The somewhat larger loss in the FM model occurs because the optimal 3-parameter rule for that model uses a coefficient below unity on the lagged funds rate.

Now we consider the extent to which complicated rules are robust to model uncertainty.

Table 6 reports several summary statistics about the extent to which rules A and В are robust to model uncertainty. The fourth column reports the value of the objective function using the ‘true” value of A (assumed to be 0.25 for rule A and 0.75 for rule B, which are the values used to determine these rules in the FRB model). The fifth column reports the absolute loss in terms of the objective function implied by following the specified rule (A or B) instead of the three-parameter optimal rule (with the same amount of interest rate volatility) for the specified value of A. The last column reports the value of A which would be consistent with the choice of rule