All Times ET
Program is Subject to Change
A Better Benchmark Process for the US Current Employment Statistics Survey (308070)*Kenneth W Robertson, U.S. Bureau of Labor Statistics
Keywords: Current employment statistics, benchmarking, alignment
The U.S. Bureau of Labor Statistics’ Current Employment Statistics (CES) survey is a highly watched indicator the of U.S. economy. The survey sample of 689,000 worksites provides data published in about 50,000 time series each month. On an annual basis, two substantially different procedures are used to align these time series with population values. The procedure used for national series aligns the prior March value with population levels and wedges the adjustment back one year. The procedure used for state and area series replaces the CES estimates with population values. Neither of these procedures is optimal. The national procedure ignores information contained in quarterly population reports; therefore the wedging of the annual difference back one year may not always provide the most accurate historical data. The state and area procedure ignores the substantially different seasonality of the monthly population data, creating a historical series that is seasonally different from current estimates. Research has identified a candidate procedure which makes better use of the population data to solve both of these problems. This paper describes the background leading to this research, the new procedure, and work still to be done to fully develop this improved process.