A growing number of empirical papers are investigating the effects of job referrals on labor market outcomes by using social network proxies. By com- bining very detailed data on social networks and actual job referrals for a large number of US firms between 2004 and 2008, this paper evaluates the use of such proxies. Because firms employ several workers and workers are employed in several firms in my dataset, I make use of worker and firm fixed effects to tackle endogeneity issues. I first estimate the effect of having a social tie with an employee on the probability to be hired and find a positive effect as in the literature - between 26.7 and 31.1 percentage points. I then relate the existence of social ties to the actual occurrence of job referrals and find that the important ties are the professional ones - professional ties in- creases the likelihood to be referred by 9.5 percentage points. I finally assess the value of using social networks and referrals for recruitment by estimating workers' ability and find that referred workers are significantly better than non-referred workers but connected workers do not significantly differ from unconnected workers suggesting that social network information might not be enough to proxy the information carried by job referrals.