The Principal-Agent Problem: Why Your Agents Don't Always Do What You Want

Seen in: Medici Bank , Decline of the Ottoman Empire

What this model means

The principal-agent problem is what happens when you hire someone (the agent) to act on your behalf, but they have their own goals and better information than you do. If incentives and monitoring aren’t aligned, they can act in ways that are good for them and bad for you.

You’re the principal. You want something done. The agent does the work. But the agent knows more about what they’re actually doing, and their interests may not match yours. That gap creates room for slacking, corner-cutting, or outright self-dealing.

Why it matters

This model is everywhere once you look for it. Shareholders (principals) hire CEOs (agents) who might prefer empire-building to returns. Patients (principals) rely on doctors (agents) who might over-test for liability protection. Voters (principals) elect politicians (agents) who might prioritize donors over constituents.

The principal-agent problem explains why “just hire good people” is never enough. Even good people respond to incentives. And when the incentives are misaligned, predictable problems follow.

Examples

1. The Medici Bank (15th century)

The Medici family in Florence ran a network of bank branches across Europe. Early on, they made branch managers partners with their own capital at stake—so the agent shared the downside. It worked for decades. But under later, less attentive leadership, managers like Tommaso Portinari in Bruges started making risky loans to impress local nobility, knowing Florence would eat the losses. Classic principal-agent breakdown: the agent got prestige, the principal got the bill. Read more in Medici Bank.

2. Ottoman tax farming (18th century)

The Ottoman Empire auctioned off the right to collect taxes to private contractors called tax farmers. The state (principal) wanted sustainable revenue; the tax farmer (agent) wanted to squeeze as much as possible before the contract ended. With weak monitoring and short-term contracts, tax farmers over-extracted from peasants, enriched themselves, and left the treasury with a depleted tax base. Read more in Decline of the Ottoman Empire.

3. Real estate agents

When you sell a house, your agent earns a percentage commission. But their incentive is to close fast, not to squeeze out the last $10,000 for you. Research shows agents leave their own homes on the market longer and sell them for more—because when it’s their money, the incentives are perfectly aligned.

4. Mutual fund managers

Fund managers charge fees regardless of performance. Their incentive is often to gather assets and avoid looking too different from their peers, not necessarily to maximize your returns. Heads they win (fees), tails you lose (underperformance).

5. Corporate expense accounts

Employees on expense accounts tend to spend more than if it were their own money. The company (principal) can’t perfectly monitor every meal or flight choice. The employee (agent) knows what they actually need vs. what they want.

How to use it / common failure mode

Any time you scale through managers, franchises, or remote teams, assume principal-agent issues by default. Then ask:

  • Skin in the game: Does the agent share downside, not just upside?
  • Information flow: Can you actually see what’s happening, or are you relying on their reports?
  • Credible exit: Can you cut ties if things go wrong, or are you locked in?

If your lieutenants can get status and money now while leaving you with long-dated risks, eventually they will.

Failure mode: Overcomplicating monitoring until you spend more watching agents than getting work done. At some point, you have to trust. The goal is to design structures where trust is warranted, not to eliminate the need for it entirely.

In one line: The principal-agent problem warns that when you delegate, your agents may pursue their interests unless you design incentives and oversight to align them with yours.


This article was produced with AI assistance and human editing. Last updated Dec 14, 2025.