Referral Incentive Design

In the presented research, we study mechanisms for the environment where agents' network connections are their private information. A mechanism designer is directly connected to only some of the agents and the other agents can participate only through the connected agents’ referrals. The mechanism designer needs to incentivize agents to bring more agents, but agents may not fully refer their connected agents to avoid competition. The Vickrey-Clarke-Groves (VCG) mechanism provides the incentive, but it suffers a budget deficit. As a budget-feasible alternative, we introduce the multilevel mechanism, which concerns agents’ marginal externality in the network hierarchy, and study the conditions for individually rationality and incentive compatibility.In single-item auction, the multilevel mechanism outperforms the VCG mechanism and the second-price auction in terms of ex-post revenue.