Hidden Action (Moral Hazard)

In lesson 1 you learned the wall between you and your agent is built from exactly two bricks: misaligned interests and something hidden. And the hidden thing comes in two kinds, split by when it hides. Some things are hidden before you sign — what the agent already knows, their private type, the true state of the used car. Others are hidden after you sign — what the agent chooses to do once the ink is dry, out of your sight.

This lesson takes the second kind: hidden action. It’s the failure that arises after the contract, when the agent picks how much effort to spend or how much risk to run, and you can’t watch which. It has a much older, much more famous name, borrowed from the insurance industry that discovered it: moral hazard.

Moral hazard, defined

Long before economists got hold of it, fire and marine insurers in the 1800s used moral hazard to describe a maddening pattern: the very act of insuring a warehouse seemed to make it more likely to burn. Not because policyholders were arsonists, but because a fully-covered owner stops sweeping up oily rags, stops fixing the dodgy wiring, stops caring — the insurer, not the owner, now eats the loss. The insurance made the careful person careless.

Strip out the smoke and the general model is this:

Notice the two conditions have to hold together, exactly like the two bricks. Shielding alone isn’t enough: if I don’t bear my losses but you can watch my every move, you’ll catch me the instant I slack. Hiddenness alone isn’t enough either: if I can’t be observed but I bear my own losses in full, I’ll still be careful, because carelessness costs me. It’s the combination — I don’t pay for the downside and you can’t see what I did — that turns a rational, ordinary person into someone who takes chances they otherwise wouldn’t.

A worked example — the insured driver. Meet Dana. Uninsured, Dana drives like someone who owns the car she’s driving, because she does: a €4,000 dent comes straight out of her pocket. Suppose careful driving costs Dana €500 a year in inconvenience — she leaves early, skips the risky overtake, parks far from trolleys — and cuts her expected crash cost from €2,000 to €400. Uninsured, that’s an easy yes: she spends €500 to save €1,600. Now she buys a policy with zero deductible — the insurer pays every euro of any crash. Dana’s private maths flips. The €500 of care still costs her €500, but the €1,600 it saves now lands entirely on the insurer. Her rational move is to stop trying: pocket the €500, and let someone else eat the crashes. She isn’t a worse person than she was yesterday; her incentives changed, and the insurer can’t sit in the passenger seat to check.

When to use it

Reach for the moral-hazard lens the moment someone is partly or fully insulated from the downside of a hidden choice — literal insurance, sure, but also bailouts, limited liability, “too big to fail,” bonuses with no clawback, salaried effort, a rental car, a company credit card. If you catch yourself saying “they’d never be that careless with their own money/car/reputation,” you’ve found moral hazard: the tell is that the exposure has been quietly moved off the person making the call.

Flavour one: too little effort (shirking)

The first way hidden action bites is the quiet one: the agent simply does less. Economists call it shirking — supplying less effort than the principal is paying for, because effort is unobservable and its rewards mostly flow to someone else.

The analogy. Picture a rowing crew where everyone is paid the same flat sum regardless of the boat’s speed, and no one can see who’s actually pulling. Each rower feels 100% of the ache in their own arms but shares the boat’s speed with seven others — so each captures maybe an eighth of the benefit of pulling hard. Rationally, everyone eases off a touch. The boat crawls, and no single rower is “to blame.”

The precise mechanism is that mismatch, and it’s the heart of the whole flavour: the agent bears essentially 100% of the cost of effort but captures only a fraction of its benefit. Effort is privately costly (it’s tiring, it’s time, it’s forgone leisure); its payoff — a better outcome for the principal — is largely captured by the principal. So the agent’s private calculation, “does this next unit of effort pay me more than it costs me?”, tips toward no long before the principal’s calculation would. The result: rational effort falls short of what the principal wants, and because effort is hidden, the principal can’t simply point and say “you slacked there.”

A worked example — cost-plus vs. fixed-price. A city hires a contractor to build a bridge under cost-plus terms: the city reimburses every euro of cost and pays the contractor a fixed 10% fee on top. Ask what effort the contractor will spend hunting for a cheaper steel supplier, negotiating harder, or trimming waste. Each euro saved by that effort is a euro the contractor no longer gets reimbursed for — and since the fee is fixed, saving money doesn’t raise the fee. Effort to control costs is all cost, no benefit to the contractor; predictably, costs balloon. Now flip to a fixed-price contract: the contractor is paid €10,000,000 flat and keeps every euro they don’t spend. Suddenly cost-cutting effort pays the contractor directly — €1 saved is €1 earned — and the shirking on cost control evaporates. (It reappears elsewhere, as we’ll see; you rarely delete moral hazard, you relocate it.) Same task, same people; the arrangement decided how hard they’d try.

Flavour two: too much risk (risk-shifting)

The second flavour is the loud one, and in some ways the more dangerous: not too little effort, but too much risk. Here the agent captures the upside of a gamble while offloading the downside onto the principal — so the agent’s rational move is to swing for the fences with someone else’s money. The name is risk-shifting (or asset substitution), and its folk motto is perfect: “heads I win, tails you lose.”

The precise mechanism. Whenever a payoff is chopped so that the agent keeps gains above some line but is protected from losses below it, the agent’s payoff becomes convex — it rises with the good outcomes and flattens out over the bad ones. And anyone facing a convex payoff loves volatility, because more variance means fatter tails on the side they keep and no extra pain on the side they don’t. So they reach for the riskiest bet available, even one with a negative expected value for the pair as a whole, because the split makes it positive for them.

Worked example one — the trader. A trader is paid a 20% bonus on annual gains and nothing clawed back on losses; the worst personal outcome is losing the job. Consider a wild bet: 50% chance of a €10,000,000 gain, 50% chance of a €10,000,000 loss — expected value zero for the bank, before you even count the risk. What’s it worth to the trader? On the win, personal bonus ≈ €2,000,000. On the loss, personal cost ≈ zero (fired, maybe, but the €10M comes out of the bank’s capital, not the trader’s savings). So the trader’s expected payoff from a coin-flip that nets the bank nothing is roughly 0.5 × €2,000,000 = €1,000,000. The trader will happily bet the bank into the ground; the maths that’s insane for the principal is a screaming yes for the agent.

Worked example two — limited liability and the bailed-out bank. The same convexity is baked into the legal furniture of the modern economy. Limited liability means a bank’s shareholders can lose at most what they put in — the downside is capped at zero, the upside is unbounded. That already tilts a bank toward leverage and risk. Now add a bailout expectation: if a bank is “too big to fail,” its worst-case losses get socked to taxpayers instead of to the bank. The downside is capped twice. So the rational move is to lever up, chase the fat tail, and let the public backstop catch the fall — which is a fair one-line summary of the run-up to 2008: moral hazard at industrial scale, where the gains were private and the catastrophic downside was, quite literally, shifted onto everyone else.

Worked example three — renter vs. owner. Milder, everyday version: hand someone the keys to a rental car and watch how it gets driven versus their own. The renter captures the upside (getting there faster, the fun of flooring it) but offloads the downside — wear, a scraped bumper, the strained gearbox — onto the owner. Not malice; just a payoff where the joyride is yours and the damage is someone else’s.

Now sort the wild from the lazy — same root cause, two very different symptoms:

When to use it

Look for risk-shifting wherever a payoff has been sliced so the agent keeps the top and someone else eats the bottom: bonuses without clawbacks, options grants, limited liability, deposit insurance, bailouts, “other people’s money” of any kind. The diagnostic question is: whose money is at risk if this bet goes wrong, and is it the same person deciding to make it? When the answer is “no,” expect more volatility than anyone signed up for — and don’t mistake it for recklessness of character. It’s the payoff shape talking.

Why you can’t just say “try harder”

Faced with all this, the instinct is to exhort — to tell the driver to be careful, the employee to hustle, the trader to be prudent, the bank to behave. It almost never works, and moral hazard tells you exactly why: the two ingredients that create the problem are the same two that make exhortation useless.

First, the consequences are diffuse. The insured driver’s carelessness costs the insurer; the trader’s blowup costs the bank; the shirker’s coasting costs the principal. The person you’re lecturing doesn’t feel the sting your words are trying to invoke, because the whole setup routed the sting elsewhere. Second, the action is hidden. Even if the agent nodded along, you can’t verify whether they actually took more care or ran less risk — so your exhortation isn’t backed by anything. You’re asking someone to override their standing incentives, unwatched, out of goodwill, forever. That’s the “hope, not a system” trap from lesson 1, wearing a new hat.

So the only thing that reliably moves the needle is to change what the agent bears — to route some of the consequence back onto the person taking the action, or to make the action (or its outcome) partly visible so it can be tied to consequences. That’s the whole toolkit of lesson 4, but here’s the preview:

The middle one leans directly on incentives: you don’t get prudent behaviour by asking for it, you get it by arranging the payoff so the agent’s own best move is the prudent one — putting them back in the passenger seat of their own decision. A deductible makes the careful driver careful again not by moral appeal but by handing back a share of the crash. Skin in the game tames the trader not by preaching but by making the blowup theirs too. That’s the through-line: stop trying to change the agent’s heart; change what the agent carries.

Recap

Hidden action is the first of the two hidden things, and its old name — moral hazard — carries the two-century-old warning that it looks like a character flaw and is actually a structure. Someone shielded from the consequences of a choice you can’t see will, quite rationally, take less care or more risk than they would if fully exposed.

The one line to keep: moral hazard is hidden action after the deal, and you beat it by changing what the agent carries, never by asking them to be better. The full toolkit — and the price tag on the problem — is lesson 4.

Next up: lesson 3, Hidden Information (Adverse Selection) — the other hidden thing, the one the agent knows before you sign, and why the bad risks crowd out the good ones.