The Bureau of Labor Statistics (BLS) generates monthly payroll estimates through the Establishment Survey, which is a component of the Current Employment Statistics (CES) program. This survey gathers input from approximately 119,000 businesses and government agencies, covering around 622,000 individual worksites across the United States.
Since the payroll figures are based on a sample rather than a complete count, they are subject to statistical estimation and subsequent revisions. The Quarterly Census of Employment and Wages (QCEW) provides a more comprehensive approximation of employment, as it compiles administrative records from unemployment insurance filings, encompassing about 97% of U.S. employment. Consequently, the QCEW serves as a benchmark for revising nonfarm payroll figures.
Each year, the BLS aligns CES data with QCEW estimates through a benchmarking process. This involves comparing the payroll level reported for the previous March with the QCEW estimate, then distributing any differences evenly across the preceding 12 months through a method known as linear “wedging” adjustment. This process ensures that nonfarm payroll data are accurately adjusted to reflect QCEW information for March of the benchmark year.
In recent years, benchmark revisions have consistently revealed significant downward adjustments. Over the past three years, adjustments have reduced previously reported payroll employment figures by a total of 1.75 million jobs, highlighting potential discrepancies in earlier employment estimates.
Why this story matters: Insights into employment data accuracy impact economic policies and labor market understanding.
Key takeaway: Benchmark revisions have notably decreased payroll estimates in recent years.
Opposing viewpoint: Some may argue that these revisions highlight necessary adjustments rather than flaws in the methodology.
