Revealed Social Networks

Abstract The linear-in-means model is the standard empirical model of peer effects and asks that an agent’s choice or outcome is a combination of their ideal point and the mean outcome of their group. Using choice data and exogenous group variation, we develop a revealed preference style test for the linear-in-means model. This test is formulated as a linear program and can be interpreted as a condition about differentiating the behaviour of each agent in a consistent manner. We then study the identification properties of the linear-in-means model. A key takeaway from our analysis is the close relationship between the dimension of the outcome variable and identification. When the outcome variable is one-dimensional, failures of identification are generic. When the outcome variable is multidimensional, we provide natural conditions under which identification is generic.