Fourier Methods for Evolving Seasonal Patterns

Abstract The use of Fourier or spectral methods for the seasonal adjustment of economic and other time series has recently been suggested; see for example Hannan 3, 4. The case in which the seasonal pattern does not change appreciably from year to year has been covered in some detail but methods appropriate to an evolving seasonal pattern have been given only a little attention. The present paper discusses the nature of evolving seasonal patterns which may arise in economic series and examines a method due to Hannan which may be used when the pattern is changing slowly over time. Although the suggested method is not readily mechanized for routine application on a large scale since some personal judgments are called for, it is found useful in detailed studies and might well be applied to any series for which traditional methods are not adequate; it is also applicable to the problem of prediction.