The Union Wage Premium: An Analysis Using Propensity Score Matching

This paper estimates the size of the union membership wage premium by comparing wage outcomes for unionised workers with 'matched' non-unionised workers. The method assumes selection on observables. For this identifying assumption to be plausible, one must be able to control for all characteristics affecting both union status and wages. This requires very informative data. We illustrate the value of the rich data offered by the linked employer-employee Workplace Employee Relations Survey (WERS) 1998 in implementing this methodology. We estimate the union membership premium for the whole private sector, among workers in workplaces where at least some workers are covered by collective bargaining, and in occupations with pay set by collective bargaining. We find a raw 17-25% union premium in gross hourly wages for the private sector in Britain, depending on the sub-group used. However, post-matching this difference falls to between 3% and 6%. This indicates that the higher pay of unionised workers is largely accounted for by their better underlying earnings capacity, which is associated with their individual characteristics, the jobs they do and the workplaces they find themselves in