How to reward honesty?

Abstract

An agent and a principal are engaged in an ongoing relationship, where the agent receives project opportunities of random value and is biased towards implementation. Each period, the principal elicits a (non-contractible) recommendation from the agent, compares it to a random outside option and decides whether to implement the project or not. The key distortions in the stationary decision rule used by the principal that will sustain honest reporting by the agent are (i) discrimination against best projects, so that sometimes very good projects are not implemented even if they would benefit the principal, (ii) general favoritism, where proposals other than the best are sometimes implemented even when they have negative value to the principal, and (iii) focused leniency towards average projects, where the bias towards implementation is larger for average projects than for bad or good projects. The first limits the agent’s incentives to exaggerate by decreasing the effectiveness of exaggeration, the second increases the overall value of the relationship to the agent while the last targets the projects for which the incentive to exaggerate is the largest to sustain truth-telling.