There is robust field data showing that a frequent and successful way of looking for a job is via the intermediation of friends and relatives. Here we want to explore this experimentally. Participants first play a simple public good game with two interaction partners ("friends"), and share whatever they earn this way with two different sharing partners ("cousins") who in turn have different friends. Thus a participant's social network contains two "friends" and two "cousins". In the second phase of the experiment participants learn about a job opportunity for themselves and one additional vacancy and decide whom of their network they want to recommend and, if so, in which order. In case of coemployment, both employees compete for a bonus. Will others be recommend for the additional job in spite of this competition, will "friends" or "cousins" be preferred and how does this depend on contributions (of "friends") or shared profits (with "cousins")? Our findings are partly puzzling. Most participants, for instance, recommend quite actively but compete very fiercely for the bonus.