The Regularity of Business Cycles
NBER Working Paper No. 2381
Do business cycles have predictable periodicities or are they random walks without past regularities or predictive value? Arguments in support of either position are found in the literature, with no apparent convergence to an agreement. This paper first examines the implications of the NBER chronologies and other findings for the question of the regularity of business cycles. It discusses hypotheses and presents evidence concerning the incidence and coexistence of cycles with different periods. An extension of the analysis covers growth cycles in the United States and other major countries. The paper then considers different models -- linear, nonlinear, endogenous, and exogenous -- for what they have to say about the problem. The regularity of investment cycles and the possible asymmetries in cyclical behavior receive particular attention, and some related data and tests are provided. Our results suggest that business cycles defy simple characterizations: they show a strong tendency to recur and at times even near periodicity, along with great diversity and evolution of phase durations. The age of a phase is not of much help in predicting the date of its end; the regularities are mainly in the dynamics of the developing business conditions.