This study compares the determinants of productivity and wages at both firm and worker level. In the firm-level analysis, we follow Hellerstein, Neumark and Troske (1999) and provide improved estimates based on an extended set of covariates including the intensity of firm-provided training. In the worker-level analysis we take a new turn and generate a proxy for unobserved worker productivity. Our results point to the presence of sizeable spillover effects from schooling and training as their impact is bigger on firm-level productivity equations than on the corresponding worker-level equations. In turn, our fully disaggregated model at worker level shows that, by using all possible combinations of worker attributes, we obtain that the wage differences across different worker groups are mostly productivity based and that the gap can be as high as 33%.