ABSTRACT
Most economic indicators have fluctuations around a growth trend. Different detrending methods lead to competing perspectives in business-cycle theory. Two detrending methods are tested by time-frequency analysis: the first differencing (FD) and the Hodrick-Prescott (HP) smoothing filter. We find HP is much better than FD in revealing deterministic patterns from economic time series (Chen 1996a). From a wide range of aggregate data, we find the existence of persistent cycles, in addition to color noise. Spectral analysis not only provides complementary evidence of “co-movements” of business fluctuations (Lucas 1981; Kydland and Prescott 1990), but also reveals distinctive patterns of frequency evolution. It is found that characteristic frequencies of business indicators are remarkably stable. Only minor changes occurred under such events, for example, the oil price shocks in 1973 and the stock market crash in 1987. Surprisingly, more significant changes happened during the Vietnam War and the Reagan administration. The time lag between frequency responses of different indicators provides important information about the propagation mechanism in the real economy. A new approach of economic diagnostic and policy evaluation can be developed quantitatively. The new perspective of time-frequency analysis indicates fundamental barriers for Friedman’s rational arbitrageurs against market disequilibrium. The role of time scale, observation reference, dynamical instability, and information ambiguity in studies of business-cycle theory is discussed.
