Testing for business cycle asymmetries based on autoregressions with a Markov-switching intercept

In this paper, we investigate the implications of the two concepts of asymmetry defined by Sichel (1993) - deepness and steepness - for first-order autoregressive processes with a Markov-switching intercept. In order to do so, we derive the two required formulas determining the coefficient of skewness of first-order autoregressive processes with a Markov-switching intercept and the coefficient of skewness of the first differences of these processes. For the special case of two states, we present the parameter restrictions leading to non-deepness and non-steepness. We show that these restrictions imply that the conclusions of Clements & Krolzig (2003) with respect to asymmetries of processes with a Markov-switching intercept are not correct. Finally, we apply the results to U.S. GDP which is found to exhibit strongly significant deepness and steepness.