The Setup

Every time you hand a task to someone else, you build a tiny economy. There is a principal — the party who wants an outcome and can’t (or won’t) produce it alone — and an agent — the party who acts on the principal’s behalf. You hire a plumber, elect a senator, buy a mutual fund, tell a teenager to mow the lawn: in each case you have delegated, and someone else’s hands are now doing your work. Delegation is one of the great engines of civilisation. It is also where a specific, recurring kind of trouble lives.

The intro course gave you the two ingredients — a principal, an agent, and the friction between them. This lesson gives you the anatomy of that friction. Here is the thesis, and it is worth memorising because the entire course hangs off it: the wall between you and your agent is built from exactly two bricks. One brick is that your agent’s interests are not your interests. The other is that you cannot fully see what your agent knows or does. Pull out either brick and the wall falls — the delegation becomes safe. It is only when both bricks are stacked that you have a genuine principal–agent problem: a situation where a self-interested agent, whom you cannot fully monitor, can profit at your expense.

You already know from the incentives course that people respond to the rewards they actually face, and from mechanism-design that the fix is to arrange the rules so the behaviour you want is the agent’s own best move. Keep both in your pocket. This lesson is about diagnosing the disease precisely, so that later lessons can prescribe the cure.

Brick one: interests diverge

The analogy. Imagine hiring a taxi in a city you don’t know, where the fare is metered by distance. You want the short route. The driver’s paycheck grows with the long route. Neither of you is a monster — the meter simply points your wallets in opposite directions. That opposition is the first brick.

The precise definition. Interests diverge when the agent’s payoff is not a faithful function of the principal’s outcome — when the agent can do better for themselves by doing worse for you. The technical name for this gap between what your agent maximises and what you’d want them to maximise is incentive misalignment. It is not the feeling of conflict; it is the structure of the rewards.

A worked example. Suppose you invest €500,000 with a fund manager who is paid a flat 1.5% of assets under management per year, with no share of your gains or losses. Line up the two payoffs:

The manager earns €7,500 whether you make a fortune or lose a fifth of your capital. What actually moves their pay is the size of the pot, not its growth — so their real job, rationally pursued, is to gather more assets and retain yours, not to beat the market for you. Their best move might be to hug a benchmark, avoid anything that could scare you into withdrawing, and spend their energy on marketing. None of that is fraud. It is just the meter running in the wrong direction.

When to use it

Reach for the “interests diverge” lens whenever you notice a payoff mismatch: the person acting for you gets paid, promoted, or praised on a metric that isn’t your true goal. Salespeople paid per deal (not per good deal), surgeons reimbursed per procedure (not per cure), a salaried employee who bears none of the downside of a slow afternoon — all show the first brick. But spotting divergence alone doesn’t prove a problem exists. Hold that thought; you need the second brick too.

Brick two: information asymmetry

The analogy. You take your car to a mechanic who tells you the transmission is failing. Maybe it is. Maybe it’s a €40 sensor and he fancies a €2,000 job. You cannot tell, because he knows the inside of your gearbox and you know the inside of your glovebox. The knowledge sits on his side of the counter, not yours. That gap is the second brick.

The precise definition. Information asymmetry is any situation in which one party to a transaction knows something relevant that the other party does not — and cannot cheaply verify. In an agency relationship the asymmetry runs in the agent’s favour: the agent typically knows more about their own effort, their own competence, and the true state of the task than the principal can observe.

That asymmetry comes in two distinct flavours, sorted by when the hidden thing arises relative to the moment you hire:

• Hidden action — you can’t observe what the agent does after you hire them. The salesperson who golfs on a “client visit”; the manager who takes the safe project because failure would embarrass them. The action is concealed. This is the engine of moral hazard, the subject of lesson 2.
• Hidden information — you can’t observe what the agent knew before you hired them, or a quality they carry that you can’t inspect. The used-car seller who knows the car is a lemon; the job applicant who knows they’re a poor fit. The type is concealed. This is the engine of adverse selection, the subject of lesson 3.

Keep the timing straight — hidden action is about behaviour after the deal, hidden information is about type before the deal — and half of this course is already organised in your head.

When to use it

Deploy the information-asymmetry lens when you catch yourself thinking “I just have to trust them” — because trust is the word we use for the gap where verification should be. Doctors, lawyers, plumbers, financial advisers and software contractors are classic expert agents: we hire them precisely because they know things we don’t, which is the same fact that makes us unable to check them. The more expert the agent, the deeper the asymmetry, the sharper the second brick.

You need BOTH bricks

Here is the crux of the whole lesson, and the part most people get wrong. Neither brick, on its own, is a problem. You need the two stacked together. Let’s prove it with two clean thought-experiments.

Thought-experiment A: misaligned interests, but perfect observation. Recall the metered taxi, but now imagine you have a live GPS map of the whole city and you know the shortest route. The driver still wants the long way — interests diverge, brick one is firmly in place. But you can see every turn. The instant he heads the wrong way, you object, and you dock the fare. Knowing this, the rational driver takes the short route from the start. No problem. When you can watch the action, you can simply pay for the action you want, and misalignment has nowhere to hide.

Thought-experiment B: hidden action, but perfectly aligned interests. Now suppose your agent is your own mother managing your money, and she genuinely wants exactly what you want — your interests are identical, brick two of behaviour is hidden but brick one is gone. You cannot see a single thing she does with the portfolio. Does it matter? Not in the slightest. Unobserved, she does the right thing anyway, because right for you is right for her. No problem. When goals coincide, monitoring is pointless — there’s no gap to exploit.

The lesson is stark: misalignment is harmless if you can watch, and hidden action is harmless if goals coincide. It is only the product of the two that bites. Put them in a grid:

Three of the four cells are safe. Exactly one cell — misaligned and hidden — is where the principal–agent problem lives. Every remedy in this course is, at bottom, a way of dragging a delegation out of the bottom-right cell: either by aligning interests (so you slide left) or by buying observation (so you slide up).

Agency chains: it’s principals and agents all the way down

The analogy. Think of a relay race where the baton is your intent. You hand it to a runner, who hands it to the next, who hands it to the next. At every exchange the baton can be fumbled, and worse, each runner has their own reasons to run their own way. Delegation is rarely a single handoff. It’s a chain — and every link is a fresh principal–agent gap.

The precise definition. An agency chain is a nested sequence of delegations in which each agent is, in turn, the principal for the next agent down. The gap at one link does not cancel the gap at the next; the gaps compound, and each intermediary can shave a little of your intent off for themselves.

A worked example — the state. Follow the baton down a government:

By the time “the citizens want a safe, affordable bridge” reaches the welder, it has passed through three interest gaps and three information gaps. This is why grand intentions arrive at the ground floor bent out of shape — not because anyone betrayed anyone, but because each link quietly optimised for itself.

The corporate version runs the same way: shareholders → board → CEO → division managers → front-line workers. Shareholders want long-run value; the board wants to keep its seats; the CEO wants a bigger empire and a fat package; managers want their unit to look good; workers want an easy shift. Five links, five gaps.

When to use it

Whenever a decision feels mangled by the time it reaches you — a policy hollowed out, a strategy garbled on the shop floor — resist the urge to find the one villain. Walk the chain link by link and ask, at each handoff, whose interests, and what can’t the link above see? The distortion is usually the sum of many small, individually-rational shavings, not one act of sabotage. Fixing it means fixing the worst link, not firing the last person to touch the baton.

Common agency: one agent, many principals

The analogy. Picture a real-estate agent who represents both the buyer and the seller of the same house — “dual agency.” Whose corner are they in? Both, officially. Neither, effectively. The agent sits at the centre of a tug-of-war holding both ropes, and the rational move is to let go of whichever rope pulls weakest.

The precise definition. Common agency is the mirror image of the chain: instead of one principal delegating down a line, multiple principals delegate to a single agent at once. The agent now serves several masters whose goals conflict, and the agent gets to choose — often invisibly — which master to actually serve.

A worked example. A corporate manager answers, simultaneously, to at least three principals:

When these pull apart, the manager can play the principals against each other — telling shareholders the regulator forced a costly change, telling the regulator that shareholders won’t allow more, telling customers that both are to blame. Each principal hears a story in which someone else is the obstacle, and the agent keeps the discretion.

When it’s NOT a problem

The most useful thing a good lens does is tell you where not to spend your worry. The two-bricks model does exactly that: a delegation stops being a problem the moment a brick is gone. Two clean escape routes exist.

Route one — interests were already aligned. Sometimes you don’t have to engineer alignment because it comes for free. A parent managing a child’s welfare, a founder-owner running the company they’ll inherit the upside of, a genuine mission-fit hire who’d do the work for love — here the agent’s payoff already tracks yours, so brick one never gets laid. The classic case is the owner-operator: when the person doing the work is the residual claimant who keeps the profit and eats the loss, principal and agent collapse into the same body. There is no gap because there are no longer two parties.

Route two — observation is cheap or output is perfectly measurable. Sometimes you can simply see everything that matters, so brick two never gets laid. A pieceworker paid per flawless widget, where a machine counts and grades every widget; a call-centre agent whose every call is recorded and scored; a data-entry clerk whose keystrokes are logged and checked. When the output is a hard, cheap-to-verify number, hidden action has nowhere to hide, and even a badly misaligned agent is leashed by measurement.

The trade-off

Beware of forcing an escape route that isn’t really there. “Just measure the output” sounds like a universal cure — but many jobs have output that is genuinely hard to measure (research, teaching, long-horizon strategy), and slapping a crude metric on them doesn’t defeat the agency problem, it relocates it: the agent games the metric instead. Likewise, “just hire aligned people” works only where alignment is real and durable, not merely claimed in an interview. The escape routes are real, but only when the brick is actually missing — not when you’d like it to be.

Recap

• Every delegation is a principal (wants the outcome) and an agent (does the work).
• The wall between them is built from exactly two bricks: misaligned interests (the agent’s payoff isn’t your outcome) and information asymmetry (you can’t fully see what the agent knows or does).
• Misalignment isn’t villainy — it’s the structure of the rewards. You fix structure by changing structure, not by hiring nicer people.
• Information asymmetry splits into hidden action (behaviour after you hire → moral hazard, lesson 2) and hidden information (type before you hire → adverse selection, lesson 3).
• You need both bricks. Misalignment with perfect observation is safe; hidden action with aligned interests is safe. Only the combination — misaligned and hidden — is the agency problem.
• Agency chains stack the gap link by link (citizens → politicians → bureaucrats → contractors), and monitoring the monitors is itself an agency problem.
• Common agency — one agent, many principals — lets the agent play principals against each other and serve the loudest.
• A delegation is no problem when a brick is missing: interests already aligned, or observation cheap and output perfectly measurable. The lens tells you where to spend worry.

Next up — Hidden Action (Moral Hazard): what happens in that dangerous bottom-right cell when you can’t see what your agent does after the deal is signed.