Assortative Matching in Risk-sharing Networks: Evidence from Thai Villages

Abstract: We look into the formation of risk-sharing groups in social networks. In particular, we are interested in risk-sharing link formation among agents of different risk types. Previous empirical studies almost consistently find that informal mutual insurance links are largely formed based on homophily. In this study, we point out that the evidence of homophily, established in the previous literature, is not robust if we use panel data documenting people’s network connections instead of using snapshot self-reported social links, and if we introduce degree heterogeneity in the econometric model for testing. We retrieve networks in 16 Thai villages based on records of economic in- teractions over 84 months. Our results show negative assortative matching with respect to risk types in kinship networks, which is consistent to the pre- diction of a standard risk-sharing matching game. However, links beyond kinship tend to follow a positive assortative matching process, even though the tendency is trivial. We apply the model of risk sharing in social networks (Ambrus et al. (2014)) to further explore the mechanism of the pattern of sorting. Our theoretical deduction indicates that the benefit of risk diversification through informal risk-sharing increases with the aggregated social value of network connections in the risk-sharing network.