Holmström & Milgrom (1991)

Citation: Holmström, B., & Milgrom, P. (1991). “Multitask Principal‑Agent Analyses: Incentive Contracts, Asset Ownership, and Job Design.” Journal of Law, Economics, & Organization.

Holmström and Milgrom’s 1991 paper extends classic agency theory to multitask settings. In many real jobs, agents perform multiple tasks, but performance can be measured only on some of them. This creates a new problem: strong incentives on measured tasks can distort effort away from unmeasured but valuable tasks.

The core idea is that incentive strength should depend on the measurability of tasks. If one task is easy to measure and another is hard, a strong performance contract on the measurable task can lead the agent to over‑focus on it, neglecting the unmeasured task. As a result, the optimal contract often uses lower‑powered incentives than would be ideal in a single‑task setting. Sometimes the best solution is to reduce incentives overall to prevent distortion.

The paper also connects incentive design to job design and asset ownership. If tasks are tightly linked, it may be optimal to assign them to the same agent with balanced incentives. If tasks are easily separable, it can be better to split them across agents or organizational units to avoid distorted effort. This insight helps explain why firms divide labor the way they do and why high‑powered incentives (like commissions) are common in some jobs but not others.

A related contribution is the analysis of asset ownership. Ownership provides residual control rights that can substitute for explicit incentives. If an agent owns an asset they use, their incentives may align naturally with performance across multiple tasks, reducing the need for strong explicit incentives. Conversely, if ownership is separated, stronger contracts may be required. This helps connect organizational structure with incentive design: ownership, control, and job boundaries are all tools for managing multitask agency problems.

The paper is theoretical but closely tied to real organizational puzzles. For example:

• Why do many professional service jobs rely on fixed salaries or low‑powered incentives despite measurable outputs?
• Why are commission structures common in sales but rare in R&D or management?
• Why do firms sometimes bundle tasks that seem unrelated, and other times split tasks across specialized units?

Holmström and Milgrom show that these patterns can be explained by the trade‑off between incentive strength and task distortion. When tasks are interdependent or when important dimensions are hard to measure, high‑powered incentives can reduce overall performance. In such cases, low‑powered incentives or broader monitoring may be superior.

The paper also has implications for performance measurement systems. If a firm can improve measurement on previously unobservable tasks, it can safely increase incentive intensity without causing distortion. This provides a mechanism linking information systems and organizational design: better measurement enables better incentives.

In short, the paper’s key contributions are:

• Incentive contracts must be designed with multitask distortions in mind.
• Strong incentives are not always optimal; low‑powered incentives can be rational.
• Job design and asset ownership interact with incentive design.
• Improved measurement systems can change the optimal contract structure.

Holmström and Milgrom (1991) thus broaden agency theory beyond single‑task models and explain why real‑world incentives often appear weaker than theory might predict. The presence of multiple tasks and imperfect measurement makes incentive design a balancing act between effort and distortion.