A Simplified Approach to Administrative Record Linkage in the Quarterly Census of Employment and Wages

Justin J. McIllece and Vinod Kapani

Abstract

In the Quarterly Census of Employment and Wages program of the U.S. Bureau of Labor Statistics, quarterly establishment records are linked longitudinally by a software-based logistic regression system that assigns linkage probabilities to potential establishment matches. Each field in the administrative record is given a weight according to its non-missing rate, and records are linked if their estimated linkage probability exceeds a minimum threshold. In this paper, we consider a simplified matching alternative that relies on the principal data fields related to the administrative definitions of establishment linkage. A linkage system based on critical matching criteria, such as location and administrative business information, is constructed. The results of the alternative record linkage approach are compared to those generated from the logistic regression software.