We investigate the effect of union membership on job satisfaction. Using linked employer–employee data from the 1998 British Workplace Employee Relations Survey, we analyse the relationship between the membership decision and overall job satisfaction and satisfaction with pay. In this paper we account for the endogenous selection induced by the sorting of workers into unionized jobs. Controlling for both individual and establishment heterogeneity and explicitly modelling the effect of union status, we find that the marked difference in job satisfaction between unionized and non-unionized workers disappears, suggesting that a selection effect, rather than a causal effect, characterizes the relationship.