Estimating heterogeneous peer effects with partial population experiments

The standard linear-in-means model of peer effects assumes that the endogenous peer effect parameter is homogeneous. In the context of group interactions, we relax this assumption by allowing individuals to respond differently to the outcomes of other group members depending on the identity of these members. We propose a simple methodology to identify and estimate the model using partial population experiments (i.e. designs in which only some individuals in a group are eligible for treatment) with variation in the share of eligible individuals across groups. We discuss two cases: randomized experiments and differences-in-differences. The estimation procedure builds on the Generalized Method of Moments. We provide a package to implement our method in R.