Abstract

Greg Erkens "Using Data Mining to Explore Seasonal Differences Between the U.S. Current Employment Statistics Survey and the Quarterly Census of Employment and Wages"

The Current Employment Statistics survey is a major economic indicator published by the Bureau of Labor Statistics to provide a timely measure of total payroll employment in the Nation. The Quarterly Census of Employment and Wages is a census of total employment that is available about 9 months after the CES estimate. While these two BLS programs measure the same information, they don’t show the same seasonal patterns. BLS recently conducted a response analysis survey (RAS) to ascertain why the patterns from the two programs differ, and this research employed classification tree algorithms to explore the data, using trees to select variables that exhibit a large influence on the seasonal difference. The results were more informative than previous searches through multiple sets of tabs, and several interesting and unconsidered relationships emerged.