Hart & Holmström (2016): The Science of Writing Better Contracts

The Invisible Architecture of the Economy

Contracts are everywhere. Your employment agreement, your mortgage, your insurance policy, the terms of service you never read — all are contracts that shape behavior, allocate risk, and create incentives. The modern economy runs on contracts.

But writing good contracts is hard. Two fundamental problems make it so: you can’t always observe what the other party does (moral hazard), and you can’t anticipate everything that might happen (incompleteness). The 2016 Nobel Prize honored two economists who built the theory for dealing with both problems.

Bengt Holmström showed how to design optimal contracts when you can’t observe effort — solving the principal-agent problem that underlies everything from CEO compensation to insurance design. Oliver Hart showed what happens when contracts are inevitably incomplete — and proved that the allocation of ownership and control rights is the key to making incomplete contracts work.

Holmström: Optimal Incentive Contracts

The principal-agent problem: A company (the principal) hires a CEO (the agent) to run the business. The company wants the CEO to work hard and make good decisions. But it can’t directly observe the CEO’s effort — only the results, which depend on both effort and luck.

How should the company design the CEO’s compensation?

Holmström’s informativeness principle (1979) provided the foundational answer: the optimal contract should tie pay to any measure that provides information about the agent’s effort — and only to such measures.

• If the company’s stock price reflects the CEO’s effort (plus noise), tie pay to stock price
• If the industry’s overall performance is just noise (luck), filter it out — pay the CEO based on performance relative to industry peers, not absolute performance
• If a colleague’s output provides information about shared conditions (luck vs. effort), use it as a benchmark

This principle has profound implications:

• Stock options for executives: Tying CEO pay to stock price makes sense — but only if you also adjust for market-wide movements. A CEO who rides a bull market shouldn’t be rewarded for luck
• Relative performance evaluation: Comparing a manager’s results to peers in the same industry filters out common shocks and provides a cleaner signal of effort
• Team incentives: When individual contributions can’t be separated, group bonuses may be optimal — even though they create free-rider problems

Holmström: Beyond Simple Incentives

Holmström extended the basic model in several crucial directions:

The multitask problem (1991): Most jobs involve multiple tasks. A teacher must both teach well and do paperwork. A doctor must both treat patients and keep records. If you reward only the measurable task (test scores, paperwork), the agent neglects the unmeasurable one (inspiring students, bedside manner).

• This explains why many jobs use flat salaries rather than performance pay — when important dimensions of the job are hard to measure, strong incentives on measurable dimensions distort effort away from unmeasurable ones
• It explains why government employees are often paid fixed salaries — their jobs involve too many dimensions to incentivize with simple metrics

Career concerns (1982/1999): Even without explicit incentive contracts, people work hard because the market is watching. A young lawyer works long hours not because of this year’s bonus, but because good performance today leads to better job offers tomorrow. Holmström formalized this insight and showed that career concerns can substitute for — or complement — explicit contracts.

Pay for luck: Holmström’s theory predicts that optimal contracts should filter out luck. But empirical evidence shows that CEOs are often rewarded for industry-wide booms they didn’t cause. This gap between theory and practice became a major topic in corporate governance research.

Hart: Incomplete Contracts and the Boundaries of the Firm

Oliver Hart tackled a different question: what happens when you can’t write a complete contract?

Real contracts are always incomplete. You can’t specify what should happen in every possible future state of the world. Your employment contract doesn’t say what happens if your company is hit by a pandemic, or if AI makes your job obsolete, or if your boss asks you to do something not in your job description.

When contracts are incomplete, who has the right to decide in situations not covered by the contract? Hart’s answer: ownership.

The property rights theory (with Sanford Grossman and John Moore):

• When a contract doesn’t specify what happens, the owner of the asset gets to decide — this is residual control rights
• Ownership matters because it determines bargaining power. If you own the factory, you can threaten to fire the workers and hire new ones. If workers own unique skills, they can threaten to leave
• The optimal allocation of ownership gives control rights to the party whose investment is most important and hardest to contract on

This framework answers fundamental questions:

• Why do firms exist? When two parties’ investments are complementary and hard to specify in a contract, merging into one firm (giving one party ownership) can be more efficient than contracting at arm’s length
• Make or buy? A company should own its supplier when the supplier’s investments are specific to the relationship and hard to protect contractually
• Public vs. private provision: Should prisons be public or private? Hart showed that private prisons have strong incentives to cut costs — but this includes cutting quality (guard training, inmate welfare) in ways that are hard to specify in a contract. When quality is hard to contract on, public provision may be better
• Financial contracting: Hart’s theory explains why debt and equity have different control rights — creditors get control when the firm can’t pay (bankruptcy), while shareholders control when things go well

The Complete Picture

Holmström and Hart address two sides of the same coin:

• Holmström: When you can write a contract but can’t observe effort, design incentives carefully — tie pay to informative signals, beware of multitasking distortions, and use career concerns
• Hart: When you can’t write a complete contract at all, allocate ownership and control rights to the party whose uncontractible investments matter most

Their 2016 Nobel Prize was awarded “for their contributions to contract theory.”

Explain It to a Child

Imagine you hire your friend to walk your dog while you’re on vacation. Problem one: you can’t watch — how do you make sure they actually walk the dog and don’t just sit on the couch? Holmström says: check clues. If the dog is tired and happy when you get back, that’s a good sign. Pay your friend based on how the dog seems, not just whether they say they walked it. Problem two: what if something unexpected happens — the dog gets sick, or there’s a snowstorm? You didn’t plan for that in your deal. Hart says: whoever “owns” the responsibility should get to make decisions when surprises happen. If you gave your friend the house key and full responsibility, they decide whether to take the dog to the vet. The key insight: you need both good incentives AND clear rules about who decides when the plan doesn’t cover everything.