Working Paper

Abstract

The authors consider a production function model that transforms worker inputs into outputs through peer effect networks. The distinguishing features of this production model are that the network is formal and observable through worker scheduling, and selection into the network is done by a manager. The authors discuss identification and suggest a variety of estimation techniques. In particular, they tackle endogeneity issues arising from selection into groups and exposure to common group factors by employing a polychotomous Heckman-type selection correction. They illustrate their method using data from the Syracuse University Men’s Basketball team, where at any point in time the coach selects a lineup and the players interact strategically to win games.