| Abstract: | A Markov‐switching model of postwar quarterly real GNP growth is used to examine the duration dependence of business cycles. It extends the Hamilton model and the duration‐dependent model of Durland and McCurdy, and compares quite favorably to simpler models in out‐of‐sample forecasting. When an expansion begins, the probability of the expansion ending is 0.2, but it gradually decreases as the expansion ages. When a contraction begins, the probability of the contraction terminating is 0.07, but it increases rapidly as the contraction ages. Output growth slows over the course of an expansion. The hypothesis of the 7–10‐year cycle is not supported. |
