Department of Labor Logo United States Department of Labor
Dot gov

The .gov means it's official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you're on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Evaluation of X‐11 and Model‐based Seasonal Adjustment Methods for U.S. Economic Time Series

Stuart Scott

Abstract

The U.S. Bureau of Labor Statistics (BLS) is undertaking an evaluation of X-11 and model-based seasonal adjustment methods. While comparisons at BLS date back 20 years, the current project is the most substantial to date, both in terms of number of series and depth of analysis. Augmented X-11, ARIMA-based, and structural model-based adjustment are the methods being compared. Software employed are an experimental version of X-13-SEATS, TRAMO-SEATS, STAMP, and an internal program handling structural models. The first phase will involve about 75 series from three major BLS programs: industry employment, consumer prices, and producer prices. An effort has been made to identify diagnostics that go across methods. Such diagnostics include examination of spectra and descriptive "F statistics" for presence or absence of seasonality and sliding spans statistics for stability of seasonal adjustment. Supplementary programs have been prepared to capture and summarize diagnostics and to produce useful graphs for analysis. The evaluation is being carried out by a Bureau-wide team with 11 members from four major offices. Results will be presented for "automatic" runs from the first phase of the study. Even for this case, interventions and calendar effects used in production have been accounted for in the specifications. Additional results may be given for determining a suitable model: automatic choices from TRAMO and from X-12 (X-13), plus judgmental choices based on AIC and other statistics. A second phase of the project will focus on special features offered with modeling approaches. Including a sampling error component in the model can limit the amount of correlated sampling error entering the seasonally adjusted and trend series. There are additional statistical tests, such as those of Findley and McElroy (2005) for misestimation, available with models but not for X-11. While no empirical results from the second phase will be available at this time, some plans and earlier results will be mentioned.