Non-cooperative formation of agent-event networks
thesisposted on 23.02.2017, 00:50 by So, Chiu Ki
We study non-cooperative network formation models with two primitives -- a set of agents and a set of exogenously given events. An event is conceived as any entity around which joint activities can be organized and plays no role beyond serving as a platform to mediate connections between agents. Agents incur costs in subscribing to events but derive benefits from being connected with other agents. This framework differs from most existing models of network formation where agents decide which agents to link with. We first study a benchmark environment and identify the set of strongly efficient and Nash stable architectures. The tension between efficiency and stability in this environment is quite weak because there is no conflict of interest between the agents. Conflict of interest is absent because the underlying preference orderings of agents over the set of feasible networks are such that there is at least one network which is best from the perspective of every agent. The tension that does exist reflects coordination failure among agents. We then consider two extensions which help further clarify the two sources of tension between efficiency and stability. In a model where each event may fail with some exogenously given positive probability, the tension between efficiency and stability arises due to coordination failures. In a model where event subscription costs depend on the identity of agents and events, the tension is either due to conflict of interest or due to coordination failure depending on event subscription costs.