Abstract

Thomas D. Evans "Differences in Performance between Concurrent and Projected Seasonal Factors for Weekly UI Claims Data"

The U.S. Bureau of Labor Statistics (BLS) seasonally adjusts weekly Unemployment Insurance (UI) data for the Employment and Training Administration (ETA) of the Department of Labor. Although concurrent seasonal adjustment is quite common for monthly and quarterly data, projected factors continue to be the norm for weekly data. Many studies, such as Pierce and McKenzie (1987), suggest that concurrent adjustment will lead to greater accuracy and smaller revisions. However, the assumption that weekly data are likely to be more subject to calendar effects, and also that the program would need to be run 52 or 53 times a year instead of 12 or 4, has led to the continued use of projected factors. Comparisons are made between concurrent estimates with ones projected out 13, 26, or 52 weeks utilizing week-to-week changes and revisions.