Many workers believe that personal contacts are crucial for obtaining jobs in high-wage sectors. On the other hand, firms in high-wage sectors report using employee referrals because they help provide screening and monitoring of new employees. This paper develops a matching model that can explain the link between inter-industry wage differentials and use of employee referrals. Referrals lower monitoring costs because high-effort referees can exert peer pressure on co-workers, allowing firms to pay lower efficiency wages. On the other hand, informal search provides fewer job and applicant contacts than formal methods (e.g., newspaper ads). In equilibrium, the matching process generates segmentation in the labor market because of heterogeneity in the size of referral networks. Referrals match ‘good’ highpaying jobs to well-connected workers, while formal methods match less attractive jobs to less-connected workers. Industry-level data show a positive correlation between industry wage premia and use of employee referrals. Moreover, evidence using the NLSY shows similar positive and significant OLS and fixed-effects estimates of the ‘returns’ to employee referrals, but insignificant effects once sector of employment is controlled for. This evidence suggests referred workers earn higher wages not because of higher unobserved ability or better matches but rather because they are hired in high-wage sectors.