We bring together the strands of literature on the returns to education, its spillovers, and the role of the employer shaping the wage distribution. The aim is to analyze the labor market returns to education taking into account who the worker is (worker unobserved ability), what he does (the job title), with whom (the coworkers) and, also crucially, for whom (the employer). We combine data of remarkable quality – exhaustive longitudinal linked employer-employee data on Portugal – with innovative empirical methods, to address the homophily or reflection problem, selection issues, and common measurement errors and confounding factors. Our methodology combines the estimation of wage regressions in the spirit of Abowd, Kramarz, and Margolis (1999), Gelbach's (2016) unambiguous conditional decomposition of the impact of various omitted covariates on an estimated coefficient, and Arcidiacono et al.'s (2012) procedure to identify the impact of peer quality. We first uncover that peer effects are quite sizeable. A one standard deviation increase in the measure of peer quality leads to a wage increase of 2.1 log points. Next, we show that education grants access to better-paying firms and job titles: one fourth of the overall return to education operates through the firm channel and a third operates through the job-title channel, while the remainder is associated exclusively with the individual worker. Finally, we unveil that an additional year of average education of coworkers yields a 0.5 log points increase in a worker's wage, after we net out a 2.0 log points return due to homophily (similarity of own and peers' characteristics), and 3.3 log points associated with worker sorting across firms and job titles.