Gaining Control by Losing Control

Abstract

We study an agency model with two vertical tasks an upstream (e.g. capital-raising) and a downstream task (e.g. implementation). The effort for the upstream task cannot be verified, and the principal can do it herself (self-performance), or delegate it to the agent (delegated performance). Only the agent can do the downstream task whose environment is his private information. We show the following. When the cost of effort for the upstream task is small, self-performance is optimal - under delegated performance, the upstream task is more likely to fail due to the agent's shirking incentive. When the cost of effort is large, by contrast, delegated performance is optimal - under self-performance, the principal makes an excessive effort for the upstream task, and as a result, the distortion in the downstream task's output schedule due to the agent's private information becomes more of an issue. Our result also has an implication for the principal's bias in favor of self- or delegated performance.