Successive Incentives

Abstract

We study the design of optimal incentives in sequential processes.  An agent initiates a value-creating sequential process through costly investment with random success. If unsuccessful, the process stops. If successful, another agent thereafter faces a new investment decision, and so forth. For any realization of the process, the total value is distributed among the agents using a reward rule, which induces a game among them. We guarantee equilibrium existence for a rich domain of rules and we characterize rules that yield the highest welfare created in equilibrium. We also characterize rules that yield the highest possible payoff for the initiator in equilibrium. A canonical class of rules are those in which an agent's reward (apart from the initiator) is affected by her own success but not the success or failures of others. Our findings show that such simple reward rules invoking short-run individual incentives are sufficient to meet our long-run systemic objectives.

Keywords: Incentives, Sequential processes, Optimal reward rules, Nash equilibrium