Kohmei Makihara, KU IFRO

"Robust Mechanism Design on Networks with Externalities"

Abstract

Decision makers often seek to target the most deserving, or the most productive, people within a community, while lacking perfect information. This paper examines the allocation problem of a good with a positive externality without monetary payments. Agents, embedded in a network, know both their own and their neighbors' valuations. The principal's goal is to allocate the good to the agent with the highest valuation by proposing a mechanism which asks each agent to report their own and their neighbors' valuations and allocates the good based on these reports. This requires to correctly incentivize agents to report their truthful information. Due to the positive externalities, agents' incentives are partially aligned with the principal's objective—an agent not only wants to receive the good but also prefers that the agent with the highest valuation receives it if they do not. The paper identifies the network structures in which an efficient mechanism exists without assuming any common knowledge on the distribution from which the valuation is drawn, regardless of their beliefs about agents they are not directly connected to. We show that such mechanism exists if and only if at least two agents are connected to everyone. If agents do not use weakly dominated strategies, such mechanism exists if there is at least one agent connected to everyone. This insight guides decision makers in structuring agent networks when they have control over connections.

Contact person: Egor Starkov