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.