ABSTRACT
Adjusting an economic time series to take account of seasonal factors (and possibly also of the number of working days and bad weather) 1 eliminates significant disruptions in the cyclical information it provides. The irregular movements which remain in the seasonally adjusted time series nevertheless shed little light on the cyclical trend. In particular, they do not provide a satisfactory answer to the central question of when a cyclical turning point occurs. The alternative, which involves a direct estimation of the cyclical component, is also associated with a large degree of uncertainty towards the end of the time series. The usual estimation methods provide only a blurred picture of the cyclical component for the section of a time series that is of most interest for short-term economic analysis. It only acquires its final form once additional values are included. It is precisely at the time of an incipient turning point that the information provided by the cyclical component identified is very inaccurate. The reason for this estimation uncertainty lies in the fact that, alongside so-called systematic components such as trend, economic cycle and season, statistical time series also display more or less irregular movements. While the influence of systematic components on the estimation process can be neutralized to some degree, it is not possible to suppress the irregular components towards the end of the time series. The upshot is a lack of clarity of current cyclical information dependent on the relative dynamic (measured against the momentum of the cyclical component) of the irregular movements.
