Propriétés spectrales des filtres usuels en économie : désaisonnalisation par CENSUS XII, et différences d'ordre d

This study concerns the efficiency of usual statistical filters in economics to reduce the tendencial and seasonal components : the moving average processes of CENSUS Xll seasonal adjustment method and the difference sof d order early proposed by BOX-JENKINS. We are herein using the spectral and cospectral analyses to know whether the other components keep stable in the filtering process and whether the leads and lags found on the basis of the initial original data are kept after these filtering processes. One application illustrates the problems by means of monthly indicators that are highly representative of French economic evolution from 1963 up to 1982.This theoretical and empirical research verifies the better performance of the moving averages in eliminating the seasonal component as well as in maintaining the stability of leads and lags between economic indicators.The use of spectral techniques leads to the determination of their limits.Finally this study which certainly confirms and makes predictable results more precise is also proposing one methodology, to question the efficiency of different seasonal adjustment techniques as CENSUS, ASA, etc. and to compare them precisely.