Abstract
The official NBER chronology of business cycle peaks and troughs is the basis both for studies on asymmetries in business cycle behavior and for the hypothesis that the business cycle has systematically changed in the period following W.W.II, with contractions decreasing and expansions increasing in length. The accuracy of such findings hinges upon the historical consistency of the NBER-dated cyclical peaks and troughs, with such consistency being defined in part by the uniform application of a single procedure for dating business cycle turning points. Our focus is the development and application of such a procedure and the subsequent remeasurement of U.S. business cycles.
In developing our dating procedure, we introduce time-varying transition probabilities into a multivariate algorithm for identifying turning points in the comovements of series. In doing so, we utilize a regime switching model in which the evolution of the transition probabilities is conditioned on an additional economic variable whose value is inferred from the number of periods the economy has been in the current phase. It is the introduction of such duration-dependent transition probabilities which enables us to detect asymmetries in the duration dependence of expansions and contractions without reference to an existing business cycle chronology.
In applying our dating algorithm to U.S. business cycle indicators, the outcome is twofold. First, we obtain results characterizing the duration dependence of post-W.W.II business cycles, concluding that post-W.W.II contractions exhibit marked positive duration dependence while post-W.W.II expansions exhibit no duration dependence, positive or negative. More specifically, our dating algorithm predicts that while the postwar economy moves out of a recession after only 12 months it has better than a 95% probability of remaining in an expansion even after two years. Second, we derive a business cycle chronology extending from 1885 to 1996. While our postwar chronology approximates that of the NBER, our interwar chronology omits two contractions while our pre-W.W.I chronology omits four contractions identified by Burns and Mitchell. In both cases, estimates of the growth rate of a common factor component would seem to suggest that these omitted episodes are merely periods of decelerated growth and not periods of recession.