The cyclical sensitivity of seasonality in U.S. employment

There is a growing recognition in the literature on business cycles that production technologies may give rise to complicated interactions between seasonal and cyclical movements in economic time series, which can distort business cycle inference based on seasonally adjusted data. For the most part, however, the empirical research in this area has relied on standard univariate seasonal adjustment techniques that provide only a partial description of such interactions. In this paper, we develop an unobserved components model that explicitly accounts for the effects of business cycles on industry-level seasonality and for the potential feedback from seasonality to the aggregate business cycle. In particular, the model extracts an aggregate "common cycle" from industry-level data, allows formal statistical testing of seasonal differences in the comovement of an industry with the common cycle, and identifies economy-wide and industry-specific contributions to the seasonal and non-seasonal variation in the data. Applying the model to quarterly US payroll employment data, we frequently find evidence of statistically significant differences across seasons in the comovement between sectoral employment and the common cycle. On the other hand, we also find that seasonal fluctuations in employment at the industry level are largely idiosyncratic and that the proportion of the total variance of the common cycle accounted for by seasonality is much less than for aggregate employment. This suggests that seasonal shocks may have less of a business cycle element to them than one might infer from the seasonal movements in aggregate variables.(This abstract was borrowed from another version of this item.)