Using a large linked employer-employee data set, this paper presents new evidence on the collective bargaining wage effect in western and eastern Germany. The novel feature of our analysis is that we use a longitudinal data set. Thus, in contrast to previous studies, we seek to assess the extent to which differences in wages between workers in covered and uncovered firms really represent an effect of collective bargaining coverage, rather than a consequence of the non-random selection of workers and firms into the different regimes. By measuring the relative wage gains or losses of workers employed in firms that change contract status, we obtain estimates that depart considerably from previous results relying on cross-sectional data. Industrylevel contracts in western Germany and firm-level contracts in eastern Germany are associated with a small, but statistically significant average wage premium of about 2 per cent. Moreover, wage decompositions indicate that the overall effect of collective bargaining coverage on the returns to observable attributes appears to be negligible once the selection into the regimes is accounted for.