Holmström (1979)

Citation: Holmström, B. (1979). “Moral Hazard and Observability.” The Bell Journal of Economics.

Holmström’s 1979 paper is a classic in contract theory and agency economics. It addresses a common problem: a principal hires an agent to perform a task, but the agent’s effort is not directly observable. The principal can only observe outcomes, which are influenced by both the agent’s effort and random noise. This creates moral hazard—the agent may shirk because the principal cannot perfectly monitor effort.

The central question is how to design a contract that motivates effort while sharing risk efficiently. Holmström shows that the optimal contract must balance two competing objectives:

1) Incentive provision: Pay must respond to performance in order to motivate effort. 2) Risk sharing: If the agent is risk‑averse, excessive performance sensitivity exposes the agent to random shocks and can be inefficient.

The paper formalizes the trade‑off between these two goals. The more output reflects true effort (low noise), the stronger the incentive component can be. The more output is driven by luck (high noise), the weaker the incentive component should be. This insight explains why purely performance‑based pay is rare in highly uncertain environments and why fixed wages remain common.

A major contribution is the informativeness principle: any performance measure that provides information about the agent’s effort should be included in the optimal contract, even if it is noisy. In other words, if a signal helps distinguish effort from luck, it has value in motivating the agent. This principle underlies modern compensation designs that combine multiple performance metrics rather than relying on a single outcome.

Holmström also introduces the idea that contract design is constrained by what is observable. If effort were directly observable, the principal could pay for effort and fully insure the agent against outcome risk. But when effort is hidden, outcome‑based pay is the only lever. The noisier the environment, the more expensive it becomes to provide incentives, because stronger incentives impose higher risk costs on the agent.

The paper is theoretical, using formal principal‑agent models rather than empirical data. But its conclusions have broad implications for how firms pay employees, how performance contracts are structured, and why real‑world contracts tend to blend fixed and variable components. It also helps explain why organizations invest in monitoring and measurement systems: improved observability reduces the cost of incentive provision.

In practical terms, Holmström’s framework suggests that:

• Performance pay is more effective in roles where output closely tracks effort (e.g., sales).
• Fixed pay is more common in roles where outcomes are heavily affected by noise or external factors.
• Hybrid contracts (base salary + bonus) are often optimal because they balance risk sharing with incentives.

Holmström’s work is foundational because it provides a rigorous, general explanation for why compensation structures look the way they do. It shifts the discussion away from “good” or “bad” contracts and toward the underlying informational constraints that make certain contracts optimal. The core lesson is that incentives are costly when effort is hard to observe—a principle that underpins later work on multitask incentives, career concerns, and organizational design.

When effort is hidden and outcomes are noisy, the optimal contract must trade off incentive strength against risk exposure—using every informative signal available. That trade-off is the central constraint of incentive design.