Several recent studies show that seasonal variation and cyclical variation
in unemployment are correlated. A common finding is that seasonality tends
to differ across the business cycle stages of recessions and expansions. Si
nce seasonal adjustment methods assume that the two sources of variation ca
n somehow be separated, the present study examines the impact of seasonal a
djustment on the analysis of cyclical patterns. Seasonally adjusted quarter
ly unemployment data for five G-7 countries are modeled by a Smooth Transit
ion Autoregression (STAR), whereas the corresponding unadjusted data are mo
deled by a so-called Seasonal STAR (SEASTAR). A comparison of the implied e
stimated peaks and troughs shows that there is substantial agreement on the
business cycle chronologies, albeit that for seasonally adjusted data, rec
essionary periods tend to last longer.