Networks and Workouts: Treatment Size and Status Specific Peer Effects in a Randomized Field Experiment

This paper estimates treatment size and status specific peer effects that are not detected by widely-used approaches to the estimation of spillovers. In a field experiment using university students, we find that subjects who have been incentivized to exercise increase gym usage more if they have more treated friends. However, control subjects are not influenced by their peers. Findings demonstrate that fraction treated has a large influence on outcomes in this environment, and spillovers vary greatly by treatment status. Results highlight subtle effects of randomization and document a low-cost method for improving the generalizability of controlled interventions in networked environments.