Seasonal Adjustment of Economic Time Series and Multiple Regression
After demonstrating that any nontrivial technique for seasonally adjusting time series inevitably leads to certain distortions of the data, an eﬀort is made to provide explicit motivation for the process of seasonal adjustment for purposes of appraising current economic conditions. Inherent advantages in terms of certain consistency requirements of a least square procedure for seasonal adjustment are pointed out. Problems encountered by the econometrician when seasonally adjusted time series are to be employed in regression analysis are also explored. The dummy variable technique for dealing with seasonal fluctuations is generalized to encompass a flexible pattern of seasonal movement. It is argued that when seasonally adjusted data rather than the dummy variable procedure are employed, there is an inherent tendency to overstate the signiﬁcance of regression coeﬀicients; a correction procedure is suggested. Consideration is given to certain special problems created by autocorrelated residuals when seasonally adjusted data are utilized in regression analysis.
Lovell, Michael C., "Seasonal Adjustment of Economic Time Series and Multiple Regression" (1963). Cowles Foundation Discussion Papers. 380.