Social Networking and Individual Outcomes Beyond the Mean Field Case

This paper examines social interactions when social networking is endogenous. It employs a linear-quadratic model that accommodates contextual effects, and endogenous local inter- actions, that is where individuals react to the decisions of their neighbors, and endogenous global ones, where individuals react to the mean decision in the economy, both with a lag. Unlike the simple V AR(1) structural model of individual interactions, the planner's problem here involves intertemporal optimization and leads to a system of linear difference equations with expectations. It highlights an asset-like property of socially optimal outcomes in every period which helps characterize the shadow values of connections among agents. Endogenous networking is easiest to characterize when individuals choose weights of social attachment to other agents. It highlights a simultaneity between decisions and patterns of social at- tachment. The paper also poses the inverse social interactions problem, asking whether it is possible to design a social network whose agents' decisions will obey an arbitrarily specified variance covariance matrix.