"Using unique employer-employee linked full population data from Danish administrative registers, this paper offers the first empirical study to differentiate between three different mechanisms of the union wage premium that are often conflated in the literature: bargaining coverage, individual union membership and workplace union density. The full population longitudinal nature of the data enables us to exploit within-worker variation in wages over time, controlling for unobservable individual-level heterogeneity. Our results suggest that all three mechanisms produce wage premiums. However, the premiums depend on the degree of centralization in wage setting. The bargaining coverage mechanism is more important when collective agreements centrally set wages, while individual union membership is more important in contexts where wage-setting is more decentralized. The findings deepen our understanding regarding what unions do for workers and how they do it."

Support: Digital