Simple rules derived from one model perform very well in the other three models; i.e., these rules are robust to model uncertainty within this class of models. For a given model, complicated rules perform only slightly better than simple ones, even when all observed state variables are incorporated in the rule. Furthermore, these rules are somewhat less robust to model uncertainty compared with well-chosen simple rules.

Thus, fme-tuning a complicated policy rule to one specific model may not be advisable, because policymakers are faced with substantial uncertainty about the true structure of the economy as well as with competing views about the quantitative effects of alternative policy actions.

Finally, rules that incorporate forecasts of the output gap and inflation rate generally do not outperform optimal rules based on current and lagged variables. This result is related to that regarding complicated rules: even in large models with hundreds of state variables, three variables (the current output gap, a moving average of current and lagged inflation rates, and the lagged funds rate) summarize nearly all the information relevant to setting the federal funds rate efficiently.

Simple rules derived from one model perform very well in the other three models; i.e., these rules are robust to model uncertainty within this class of models. For a given model, complicated rules perform only slightly better than simple ones, even when all observed state variables are incorporated in the rule. Furthermore, these rules are somewhat less robust to model uncertainty compared with well-chosen simple rules. Thus, fme-tuning a complicated policy rule to one specific model may not be advisable, because policymakers are faced with substantial uncertainty about the true structure of the economy as well as with competing views about the quantitative effects of alternative policy actions. Finally, rules that incorporate forecasts of the output gap and inflation rate