摘要

This paper finds that observed monthly United States stock index returns are consistent with an underlying mechanism of shifts in regimes amongst multiple states with differing means and volatility. An issue of especial interest is whether long-term clustering of regimes gives rise to stock market cycles. The paper therefore introduces a likelihood ratio test for long-term clustering of regimes. Clustering of regime presence tends to involve much longer term cycles than the bull and bear market cycles identified by Pagan and Sossounov (2003), thus extending the research issues that are associated with the analysis of mean returns using multiple state regime-switching models.