A0394
Title: Causal inference and interpretation of linear social interaction models with endogenous networks
Authors: Tadao Hoshino - Waseda University (Japan) [presenting]
Abstract: Causal inference is studied for linear social interaction models in the presence of endogeneity in network formation under a fully heterogeneous treatment effects framework. An experimental setting is considered where individuals are randomly assigned to treatments while no interventions are made on the network structure. It is shown that running a linear regression ignoring the network endogeneity is not problematic for estimating the average direct treatment effect of own treatment, but it leads to both a sample selection bias and a negative weight problem for the estimation of the average spillover effect is proposed to use. To overcome these issues, a potential peer's treatment as an instrumental variable (IV), which is a valid IV for the actual treatment exposure by experimental design. With this IV, several IV-based estimands and demonstrate are examined that they do not suffer from selection bias and have a local average treatment effect-type causal interpretation for the spillover effect.