Identifying and Estimating Social Connections from Outcome Data

Abstract

Knowledge of the relevant linkages between individuals is usually necessary for the estimation of social interaction models. We obtain results that allow for the estimation of parameters of interest in a model with endogenous, exogenous and correlated effects without information on the relevant linkages. Our identification analysis relies on usual assumptions on the nature of interactions found in, e.g., Bramoullé et al. [2009]. To obtain identification we further impose conditions on the density of links and repeated observation of outcomes for a given group of individuals. We provide an estimation strategy which we investigate via simulations and an empirical illustration. (with Imran Rasul and Pedro Souza)