When labor markets are imperfectly competitive, firms may be willing to finance general training if the wage structure is compressed, that is, if the increase of productivity after training is greater than the increase in pay. We propose a novel way of testing this proposition, which exploits the variation in training incidence and in the training wage premium within the European Union. Our results unambiguously show that (general) training incidence is higher in clusters – defined by country, sector, occupation and educational attainment – with a lower training wage premium, measured as the differential between the median wage growth of trained and untrained employees.