Empirical Comparison of Two Methods for Non-Gaussian Seasonal Adjustment

This study compares two new seasonal adjustment methods designed to handle outliers and structural changes: X-IZARIMA and GAUSUM-STM. X12-ARIMA is a successor to the X-ll-ARIMA seasonal adjustment method, and is being developed at the U.S. Bureau of the Census (Findley et al. (1988)). GAUSUM-STM is a non-Gaussian method using time series structural models, and was developed for this study based on methodology proposed by Kitagawa (1990). , The procedures are compared using 29 macroeconomic time series from the U.S. Bureau of the Census. These series have both outliers and structural changes, providing a good testbed for comparing non-Gaussian methods. For these series, the X12-ARIMA decomposition consistently leads to smoother seasonal factors which are as or more “flexible” than the GAUSUM-STM seasonal component. On the other hand, with some significant exceptions, GAUSUM-STM generally handles outliers and level shifts better than X12-ARIMA. The differences between GAUSUM-STM and X-12-ARIMA in handling outliers and structural changes are swamped by the fundamental differences in the nature of the seasonal decompositions. Recognizing that seasonal adjustment is a subjective enterprise, we feel the X-12-ARIMA procedure yields more appealing seasonal adjustments for most of the series examined. However, GAUSUM-STM potentially offers some important advantages. This study gives guidance on what problems need to be tackled to improve STM-based seasonal adjustments.