Six months of the year, which for convenience we call the spring and Christmas seasons, have a statistically higher number of troughs than the other six months of the year. In contrast, peaks do not exhibit any clustering. These results are drawn from the NBER chronology as well as alternative chronologies. As peaks are evenly distributed and troughs are not, it also appears that contraction lengths following peaks in the oﬀ-season are longer which is also an indication of the uneven propensity to switch regime throughout the year. This paper deliberately takes a “model-free” and “distribution-free” approach to test and document these phenomena. Markov chain models, their stochastic process theory and spectral representation appear in Ghysels (1991, 1992). Such models reveal more information and do not even require the use of the NBER chronology; yet, they require a fair number of auxiliary assumptions which are not imposed in the paper.
Ghysels, Eric, "Christmas, Spring and the Dawning of Economic Recovery" (1992). Cowles Foundation Discussion Papers. 1270.