Estimating the Union Earnings Effect Using a Sample of Displaced Workers

Abstract

This paper improves on past longitudinal estimates of the union earnings effect by using a sample of workers for whom the error in measuring changes in union status is minimized. I use a sample of workers displaced by plant closings from the 1994 Current Population Survey Displaced Workers file to estimate the effects of union membership on weekly earnings and the probability of membership in a group health plan. Several tests for selection bias in cross-sectional estimates are evaluated. In contrast to past research, I find very little evidence of selection bias. In pre-post displacement difference equations, estimates of the union earnings effect range from 15 to 21 percent and are between roughly 70 and 90 percent of the cross-sectional differentials. Moreover, the hypothesis of equal cross-section and longitudinal point estimates cannot be rejected in any of the specifications tested. Similar results are found for coverage in a group health insurance plan.