Union earnings gaps and membership dynamics
Henry S. Farber and Stephen Machin
This paper revisits estimation of the union-nonunion earnings gap based on the experience of job losers in the 1994-2024 Displaced Workers Survey (DWS). A central focus is placed on earnings changes associated with union status changes for full-time private sector job losers who find a new full-time private sector job. An important feature of the analysis is that we directly account for classification error in union status in order to correct the attenuation of estimates of the earnings gap. This is particularly serious when measuring earnings changes associated with union status changes in longitudinal data. Two results are highlighted. First, in the low union environment in the U.S., losers of union jobs are surprisingly likely to be union members in their new job (on the order of 60 percent). Second, estimated difference-in-difference union earnings gaps that account for classification error are substantial but asymmetric among union joiners and leavers. Union leavers suffer a substantial earnings loss relative to union stayers of about 28 percent, while union joiners enjoy a 19 percent earnings gain relative to those who remain nonunion.
19 November 2026 Paper Number CEPDP2233
Download PDF - Union earnings gaps and membership dynamics
This CEP discussion paper is published under the centre's Labour programme.