In this paper, we test whether a regime shift from expansion to recession and vice versa is, ceteris paribus, equally likely throughout the year. If not, then it may, for instance, be less likely to get out of a recession in the middle of the winter than it is, say, in the spring or summer. We make use of Markov switching regime models to test the hypothesis of interest. The evidence is based on the conventional NBER business cycle chronology as well as alternatives to it. We find that recessions exhibit a periodic pattern in their switching regime transition probability structure. It is particularly the months associated with Christmas and spring that appear to have higher switching probabilities from recession to expansion. Our results also imply that a recession and an expansion are, on average, longer or shorter depending on what time of the year they start. Such results suggest the presence of seasonal patterns in business cycle durations. One should note though that such a notion of seasonality is quite different from the common one based on unobserved component linear time series models. Our paper investigates issues which go beyond linear dependence between seasonality and business cycles.
