Estimating Peer Effects Using Partial Network Data - Vincent Boucher Elysée Aristide Houndetoungan

Thank you also to the participants of the Applied/CDES seminar at Monash University, the Economic seminar at the Melbourne Business School, the Econometric workshop at the Chinese University of Hong Kong, and the Centre of Research in the Economics of Development workshop at the Université de Namur. [...] As such, for each group r = 1, ...,M , we have: yr = cr1r + Xrβ + αGryr + GrXrγ + εr, where cr is not observed, 1r is an Nr-dimensional vector of ones, Nr is the size of the group r, Gr is the subinteraction matrix in the group r, and εr is the vector of error terms in the group r. [...] This has the effect of changing the effective rank of the linking probabilities: increasing κ decreases the effective rank.22 We therefore expect the estimator proposed by Alidaee et al. [...] Indeed, the identification of the model rests on the a priori information of A. [...] In a classical setting, and under the usual assumptions, the estimation of (2) produces a estimator ρ̂ of ρ0 as well as an estimator of the asymptotic variance of ρ̂, i.e.