Identification and Estimation of the Linear-in-Means Model of Social Interactions
1 We would like to thank Gary Chamberlain, Enrico Moretti and participants at UCLA Work-shop in Economic Applications and the Harvard development and econometric lunches for many helpful comments. A very special thanks goes to Ariel Pakes, who appeared as a co-author on an initial version of this paper, and whose insight and discussions have proved invaluable. We are much indebted to his intellectual generosity. Abstract This paper studies identification and estimation of the linear-in-means model of social interactions. The paper emphasizes the different information contained in the between-social-group and within-social-group variation of the data. Different identifying assump-tions can then be used to deliver exogenous sources of between-and within-group vari-ation. We provide specific identification and estimation results for grouped cross-section (GCS), repeated grouped cross-section (RGCS), grouped panel (GP), and time series data. Our unifying framework can be used to reinterpret several well-known papers in the social interactions literature and should also facilitate future research.