Using a matched employer-employee data set for Norway, we exploit rare info
rmation on the union status of both individual employees and their workplac
es. We establish two key results. First, we find a positive effect of workp
lace trade union density on the level of the individual's pay in establishm
ents covered by collective agreements. Second, we find that, conditioning o
n coverage, the individual union membership differential disappears after c
ontrolling for establishment-level union density. The union wage effect is
therefore a pure public good, with individual membership conveying a positi
ve wage externality.