Concurrent Seasonal Adjustment of State and Metro Payroll Employment Series

Steven Mance

Abstract

Concurrent seasonal adjustment uses the most recent raw data in calculating seasonal adjustment factors, in contrast with methods where factors are estimated periodically and projected forward. The Current Employment Statistics - State and Area (CES-SA) program at the U.S. Bureau of Labor Statistics uses a unique two-step seasonal adjustment approach where the benchmarked universe data are adjusted separately from the survey-based data due to different seasonal patterns exhibited by data from the two sources. A history of survey-based estimates are used each January to provide projected factors for the coming year. Switching to a concurrent adjustment method was examined and is being considered. The concurrent adjustment method yielded factors that were more accurate (evidenced by smaller revisions.) Employment data were less volatile month-to-month under the concurrent method, and also more closely matched universe data were less volatile month-to-month under the concurrent method, and also more closely matched universe data, which are considered to be a more accurate gauge of economic reality. Some of the improvement was due to better estimation of a regression effect that is used to adjust for the varying number of weeks between survey reference periods.