“Informal Risk Sharing with Local Information”

This paper considers the effect of local information constraints in risk-sharing networks. We assume individuals only observe the endowment realizations of their neighbors, and bilateral risk-sharing arrangements can only depend on commonly observed information. We derive necessary and sufficient conditions for Pareto efficiency under these constraints, and provide an explicit characterization of Pareto efficient arrangements under CARA utilities and normally distributed endowments. With independent endowments, a local equal sharing rule is shown as optimal. For correlated endowments, the optimal sharing rule is characterized in closed form, relating to a novel measure of network centrality. Contrary to other models in the literature, more central individuals are likely to become quasi insurance providers to more peripheral individuals, attaining more volatile consumption. This framework has important implications for empirical tests of risk sharing. We show that standard estimates exhibit heterogeneity in insurance opportunities that correspond to differences in consumption volatility