曹志刚教授，北京交通大学

We investigate informal risk sharing using a dynamic network game model. In each round, a randomly selected agent experiences a negative shock, and the agent's friends decide whether to provide assistance. Assuming that agents have concave utility functions, we prove a version of the Folk Theorem. Our analysis shows that a group of agents will help each other in all relevant rounds of a subgame perfect Nash equilibrium if and only if there is a subgraph of these agents, in which each agent has a number of friends that is neither too low nor too high. Although optimization problems related to this model are generally NP-hard, we are able to perform several natural comparative statics.