Nonresponse Bias Analysis of Average Weekly Earnings in the Current Employment Statistics Survey October 2013

The nonresponse bias of average weekly earnings in the Bureau of Labor Statistics’ Current Employment Statistics (CES) survey is assessed. The impetus for this study is the low response rate for hours and earnings data in the CES survey, a longitudinal survey of business establishments, that provides monthly estimates of employment and average weekly earnings, among other statistics. Although we cannot produce a theoretical bias and do not have “true ” figures of average weekly earnings, we can assess the direction and relative magnitude of bias by comparing CES employment and earnings to employment and wages of the Quarterly Census of Employment and Wages (QCEW) program. The QCEW program collects employment and wages from employers covered under the States ’ Unemployment Insurance (UI) tax systems on a quarterly basis. Records of the QCEW are used as the frame for the CES survey and QCEW employment data are independent population controls for CES employment figures on an annual basis