Principals Matching Agents : Applications to Organizations and Auctions

We develop a principal/agent model for matching agents in two-sided assignments. A principal has preferences over all agents' assignments, and agents have privately-known preferences about their own match (but are indifferent about others'). Unhappy agents can quit, but the principal can stop agents from trading assignments and can pay targeted transfers to agents to induce retention. We show that the principal values flexibility from agents to accommodate other players' interests. The principal therefore pays each side information rent over all their productive match partners (including for pairs that are ultimately not chosen). This induces truthful revelation on all productive dimensions (on both sides), and the principal can choose her preferred allocation (given constraints). The mechanism can be interpreted as a scoring auction. We conclude by discussing applications to job assignments within organizations, and extensions to other multi-sided mechanism design problems