We assess the magnitude and mechanisms of workers’ productivity spillovers by estimating the peer effects among those working in the same occupation across firms using the setting of security analysts. The empirical design exploits one feature of social networks: the existence of partially overlapping peer groups. This refers to analysts who cover similar industries but not exactly the same group of industries, which generates peers of peers (excluded peers). This allows the identification of both peer characteristics and peer outcome effects. In addition, to deal with common group shocks, the exogenous characteristics of excluded peers are used as instruments. We find strong evidence of spillovers in terms of peer outcomes and characteristics. In particular, peer accuracy is positively related to analyst accuracy, while the number of industries followed by analysts' peers negatively impacts accuracy. In terms of the potential mechanisms that account for the spillover effects, we find that the effects are stronger when analysts see their peers performing well and that besides imitation, knowledge spillovers also help explain the results.