The paper studies how social job finding networks affect firms' selection of employees and the setting of entry wages. Our point of departure is the Montgomery (1991) model of employee referrals which suggests that it is optimal for firms to hire new workers through referrals from their most productive existing employees, as these employees are more likely to know others with high unobserved productivity. Empirically, we identify the networks through coworker links within a rich matched employer-employee data set with cognitive and non-cognitive test scores serving as predetermined indicators of individual productivity. The results corroborate the Montgomery model's key predictions regarding employee selection patterns and entry wages into skill intensive jobs. Incumbent workers of high aptitude are more likely to be linked to entering workers. Firms also acquire entrants with higher ability scores but lower schooling when hiring linked workers supporting the notion that firms use referrals of productive employees in order to attract workers with better qualities in dimensions that would be difficult to observe at the formal market. Furthermore, the abilities of incumbent workers are reflected in the starting wages of linked entrants, suggesting that firms use the ability-density of social networks when setting entry wages. Overall the results suggest that firms use social networks as a signal of worker productivity, and that workers therefore benefit from the quality of their social ties.