Peer effects with sample selection: an application in online job training

We study peer effects in social interactions models where individual participation in actual groups is endogenous. By exploiting instruments in participation choices, we identify and estimate the peer and contextual effects using both individual- and group-level correction terms. We study peer effects in an online job training program in China and document significant peer effects and selection bias in the duration of lecture attendance among the trainees. Our estimates suggest that ignoring sample selection would over-estimate the peer effects.