Do the phases of the business cycle die of old age?

Abstract

The paper re-examines the issue of duration dependence in the Australian classical and growth business cycles in light of the somewhat surprising results obtained recently by Cashin and Ouliaris (2004). In so doing the authors use the multinomial logit regime switching modelling approach of Layton and Smith (2003). The paper also represents an extension of the earlier work on the issue undertaken by Bodman (1998); the key extensions being that the issue is framed within an explicit established business cycle chronology, a leading index is also included within the analysis, and the growth cycle, in addition to the classical cycle, is considered. Strong evidence of duration dependence is found for periods of recession within the classical cycle and for both phases of the growth cycle. Moderate evidence of duration dependency is also found for periods of classical cycle expansion. However, the evidence in this regard is significantly reduced once movements in the leading index are included in the analysis with its movements exhibiting strong power in predicting the termination of classical business cycle expansions. For growth cycles, duration dependence symmetry is found across both phases of the cycle.