Duration dependence in the US stock market cycle: a parametric approach

Abstract

This study uses parametric hazard models to investigate duration dependence in US stock market cycles over the January 1885 to July 1992 period. The results show that duration dependence exists in pre-World War II expansions and in post-World War II contractions. Pre-war contractions and post-war expansions, however, do not exhibit duration dependence. Additionally, the evidence suggests that duration dependence in market expansions has reduced over time, while duration dependence in contractions has increased. Duration dependence is consistent with predictable stock price behaviour. Although not formally addressed in this study, predictability may arise from either temporary ‘fads’ in the pricing of securities or from time-varying required returns. Regardless of the source, the shift in the pattern of duration dependence suggests that a change occurred in the cyclical behaviour of the stock market during the period studied. Discrete shifts and trends in mean phase duration do not exist, indicating that the average duration of both market expansions and contractions have not changed over the sample period. This result is consistent with the finding of Shapiro that the variance of stock returns in the post-World War II period is essentially the same as in previous periods