The orthodox approach to modeling the Fed's response to economic conditions is to employ a monetary policy reaction function. Most reaction functions are estimated by regressing a policy indicator, possibly the federal funds rate or a monetary aggregate, on variables that describe the state of the economy, such as unemployment, inflation, and growth in output.' Under certain conditions which we discuss below, estimated coefficients from a reaction function provide information about the Fed's monetary policy priorities. Moreover, extending the model to include election and partisan dummy variables may provide information about the Fed's response to political pressures. There are two problems with interpreting the coefficients in a standard reaction function as the weights that the Fed attaches to its policy objectives. First, the monetary policy decision may not be represented accurately by an aggregate reaction function (one that models the Fed as a whole) because there are twelve voting members on the FOMC. Chappell, Havrilesky, and McGregor [10] have recently addressed this issue by employing a disaggregated set of reaction functions. The second problem with interpreting reaction function coefficients as the weights the Fed places on its policy objectives is that this interpretation implicitly assumes structural stability of the underlying macroeconomy. Abrams, Froyen, and Waud [1, 31] acknowledge this shortcoming when they state, . .. coefficients from estimated reaction functions ... do not provide direct information on policymaker utility functions. Rather than being the solution to an unconstrained optimization dependent only on policymaker preferences, reaction functions are the output of a constrained maximization, where the constraints are the reduced-form equations that characterize
Read full abstract