Estimating social effects in a multilayered Linear-in-Means model with network data

Abstract

This paper studies the identification and estimation of social parameters in a general version of the Linear-in-Means model commonly fitted in the Social Sciences with multilayered network data. A Monte Carlo exercise showcases its good small-sample properties while an empirical application to Canadian consumers’ credit usage demonstrates its applicability. Our estimates show that one’s credit-card balance increases by $0.31 for an extra $1 owed by surrounding neighbors.