Online Program

Modeling Aggregates for Reliability and Confidentiality of Output with Application to the QCEW
David Hiles, U.S. Bureau of Labor Statistics 
*Santanu Pramanik, NORC at the University of Chicago 
Fritz Scheuren, NORC at the University of Chicago 
Avi Singh, NORC 
Michael Yang, NORC at the University of Chicago 

Keywords: cell suppression, random noise method, output treatment, small area estimation

The Quarterly Census of Employment and Wages (QCEW) program of the Bureau of Labor Statistics (BLS) publishes tabulations of employment and wages by industry and geography. The BLS has been concerned for some time about the current cell suppression method for disclosure limitation because it results in substantial data suppression that compromises the quality and utility of the QCEW data. To address such concerns, BLS has been conducting research on the application of the random noise method (input treatment) to QCEW as an alternative to cell suppression. The goal is to release significantly more data and to respond to new disclosure vulnerabilities. In this paper, we explore another alternative based on the application of small area modeling techniques to disclosure limitation for the QCEW by modeling aggregates. In this new application of small area modeling we exploit the built-in perturbation of direct estimates (here, true totals) by the synthetic component. We will present results to evaluate the performance of the output treatment of disclosure limitation.