“An Econometric Investigation of Frequency Dependence in the Fed’s Behavior: A Re-analysis of the Taylor Rule” by Byron TSANG
Randal J. Verbrugge
We estimate a nonlinear Taylor Rule allowing for possible frequency dependence in the relationship i.e., we model the Fed's interest rate setting behavior as depending on the recent persistence in fluctuations in the real-time unemployment and inflation rates over the sample period 1960:03 to 2008:08. Each of these time series is partitioned into frequency components using a set of moving one-sided band-pass filters in such a way that the components add up to the original series. We find that the Fed did not act as if it believed in a Phillips-Curve trade-off during this period, as it did not react to the natural rate of unemployment (i.e., the trend or zero-frequency component of unemployment) and did respond to changes in long-run expected inflation (i.e. to the zero-frequency component of inflation). At positive frequencies, corresponding to fluctuations of 3 to 36 months in length, the Fed did react in a frequency dependent way to fluctuations in the unemployment rate, but not the inflation rate. And the economic and statistical significance of these reactions varies with sample sub-period: During the 1960s and 1970s, the Fed did not respond much to both the unemployment rate and inflation, and for the Volcker-and-after period the Fed was aggressive towards inflation. For the Great Moderation period the Fed was less hawkish towards inflation and was more concerned with the unemployment gap. Our results cast doubt on the conventional pre-Volcker and Volcker-Greenspan division of the sample period.