This paper estimates a wage equation that includes worker- and firm fixed effects simultaneously, using a longitudinal matched employer-employee dataset covering virtually all Portuguese employees over a little more than two-decades. The exercise is performed under optimal conditions by using (a) data covering the whole population of employees and (b) adequate econometric methods and algorithms. The variation in log real hourly wages is then decomposed into six different components related to worker and firm characteristics (either observed or unobserved) and a residual component. It is found that worker heterogeneity is the most important source of wage variation (46.2 percent), due in roughly equal parts to the unobserved component (24.2 percent) and the observed component (22 percent). Firm effects are less important overall (29.6%), although firms’ observed characteristics do play an important role (14.8) in explaining wage differentials.