Using Quantile Regression to Model Revisions Due To Late Reporting in the Current Employment Statistics Survey

John S. Dixon and Clyde Tucker

Abstract

This paper considers the possible effects that late responders and nonresponders to the Current Employment Statistics Survey (CES) have on bias in the employment estimates from the CES using data from the Quarterly Census of Employment and Wages (QCEW), which is a nominal census of US establishments based on unemployment insurance. Besides reporting on the level of bias produced by nonresponse and late reporting over time for all firms, the analysis also focuses on the relationship between size of firm and both nonresponse and late responding given that previous research found a relationship between firm size and nonresponse. In this latter case, quantile regression is used to estimate the relationship between size of firm and nonresponse and late reporting. Results are presented overall and by industry.