Modern research in macroeconomics provides many reasons why monetary policy should be evaluated and conducted as a policy rule - or contingency plan for policy - rather than as a one-time change in policy. First, the time-inconsistency literature shows that without commitment to a rule policymakers will be tempted to choose a suboptimal inflation policy—one that has a higher average inflation rate and no lower unemployment than a policy with a lower average inflation rate. Second, one needs to stipulate future as well as current policy actions in order to evaluate the effects of policy. (This is a positive statement of the Lucas critique of policy evaluation.) It is why virtually all policy evaluation research on monetary policy in recent years has focused on policy rules. Third, credibility about monetary policy appears to improve its performance; sticking to a policy rule will increase credibility about future policy action. Fourth, policy rules that give market participants a way to forecast future policy decisions would reduce uncertainty. Fifth, policy rules are a way to teach new policymakers, students, and the public in general about the operations of the central bank. Finally, policy rules increase accountability, potentially requiring policymakers to account for differences between their actions and policy rules.

In arguing in favor of policy rules I recognize that certain events may require that the rule be changed or departed from; that is, some discretion is required in operating the rule. But there is still a big difference between a policy approach that places emphasis on rules and one that does not. With a policy rule in mind the analysis of policy—including questions about whether a deviation from the rule is warranted—will tend to focus more on the rule rather than pure discretion. But to be more specific about rules versus discretion one needs to be more specific about the policy rule. What should the rule be? . . .

A survey of simulations of econometric models with rational expectations suggests to me that monetary policy should respond in the following way. First, the policy should respond to changes in both real GDP and inflation. Second, the policy should not try to stabilize the exchange rate, an action which frequently interferes with the domestic goals of inflation and output stability. Third, the interest rate rather than the money supply should be the key instrument that is adjusted. Because of the nature of the trade-off between inflation stability and output stability, the weights on these two measures of stability appear to matter relatively little for these general conclusions.

In order to investigate the practical application of such a policy rule, several years ago I proposed a specific formula for policy that had these characteristics. According to this policy rule the federal funds rate is increased or decreased according to what is happening to both real GDP and inflation. In particular, if real GDP rises one percent above potential GDP the federal funds rate should be raised, relative to the current inflation rate, by .5 percent. And if inflation rises by one percent above its target of 2 percent, then the federal funds rate should be raised by .5 percent relative to the inflation rate. When real GDP is equal to potential GDP and inflation is equal to its target of 2 percent, then the federal funds rate should remain at about 4 percent, which would imply a real interest rate of 2 percent on average.

The policy rule was purposely chosen to be simple. Clearly, the equal weights on inflation and the GDP gap are an approximation reflecting the finding that neither variable should be given a negligible weight.