Using Current Population Survey data for 1983-93, this article analyzes whether there is a union membership wage premium among full-time, private sector employees covered by union contracts. Ordinary least squares estimates of the membership wage premium are 12%-14%, and allowing membership to be endogenous yields larger estimates. Differences in job tenure, unobservable characteristics, and measurement error cannot fully explain the estimated premium. Significant differences in this premium, as well as in membership rates conditional upon coverage, across various demographic subgroups are also documented. In general, "free riders" do not appear to be free riding.