Why the Referential Treatment? Evidence from Field Experiments on Referrals

This paper presents the results of three \u85eld experiments in an online labor market designed to determine whether referred workers perform better than non-referred work-ers and, if so, why. We hired workers and asked them to refer others; we then hired all referred and non-referred applicants. Referred workers performed better and had less turnover than non-referred workers. We \u85nd that this is partially due to selection: refer-rals outperformed non-referred workers even without any on-the-job interactions with their referrers. We \u85nd also that team production is important: referrals performed particularly well when working with their referrers