How hard is it to count? Reconciling monthly survey and administrative data: experiences from the CES program

Steven Mance


The Current Employment Statistics (CES) program at the U.S. Bureau of Labor Statistics produces monthly nonfarm payroll employment, hours, and earnings estimates at the national, state, and metropolitan level from a fast-turnaround survey of about 146,000 businesses and government agencies representing approximately 623,000 establishments. These data are subject to many common sources of error such as sampling and nonresponse, and the employment data are benchmarked (reconciled) using the Quarterly Census of Employment and Wages (QCEW), which is based on administrative records from the Unemployment Insurance system, available 6-8 months after the sample-based estimates. QCEW employment snapshots are available at a monthly frequency; however, noneconomic changes and differences in coverage, response, and seasonality complicate its use in benchmarking. Each year, National CES data are reconciled to a QCEW-based March level, while state and metro series are replaced using adjusted QCEW values for every month. The respective benefits, complications, and challenges of both methods is recounted. The recent use of QCEW data to reconstruct 23 years of metropolitan employment data due to definitional changes is detailed. Future directions in the use of administrative data for the CES program is explored.