Hidden Information (Adverse Selection)

In lesson 2 the danger showed up after you signed. You hired someone, the ink dried, and then — safely out of sight — they started slacking, cutting corners, or taking risks with your money. That’s moral hazard: hidden action, after the deal.

This lesson is the other brick, and it’s sneakier, because it strikes before you sign. The problem isn’t what the agent will do later. It’s what the agent already knows right now and you don’t. Call it hidden information: a fact about the agent’s type — how healthy, how risky, how good the car actually is — that is baked in before either of you says a word.

Here’s the twist that makes it a trap rather than a mere annoyance. When one side knows their own type and the other can’t tell them apart, the very act of offering a deal selects the wrong people into it. Offer cheap insurance and you attract the sick. Offer a loan at a high rate and you attract the desperate. The pool that says “yes, please” is skewed toward exactly the types you didn’t want. That self-selection is adverse selection, and it’s the villain of this lesson.

The market for lemons

The analogy. You’re at a used-car lot. Some cars are gems — call them peaches. Some are disasters that will die on the motorway — lemons. The seller has driven every one of them and knows exactly which is which. You, the buyer, see only shiny paint. You are betting blind, and the seller is not.

The definition. The market for lemons is George Akerlof’s 1970 model (it won him a Nobel Prize) showing that when buyers can’t observe quality but sellers can, the average quality traded collapses — and in the worst case the good goods disappear entirely and the market unravels. It is the canonical illustration of adverse selection.

Worked example — watch it unravel. Start with a market split evenly between peaches and lemons:

Step through it:

1. Buyers price the average. Can’t tell them apart, 50/50 mix, so a rational buyer offers the expected value: (10,000 + 4,000) / 2 = €7,000.
2. Peach owners refuse. A peach is worth €10,000 to its owner. Nobody sells a €10,000 car for €7,000, so peach owners withdraw from the market.
3. The pool re-skews. Now the only cars still for sale are lemons. The mix that self-selected into “willing to trade at €7,000” is adversely skewed — almost all lemons.
4. Buyers re-price down. Buyers aren’t fools; they notice only lemons trade at €7,000, so they cut their offer toward the lemon’s true value, €4,000.
5. The market unravels. At €4,000 you can only ever buy a lemon. The good cars have been driven out of a market they should have been the stars of. Bad quality has literally chased out good quality — “the bad drives out the good.”

The killer detail: nobody lied. Every seller was honest, every buyer was rational, and the good product still vanished — purely because quality was hidden. That’s what makes adverse selection so much scarier than simple fraud.

When to use it

Reach for the lemons model whenever one side can observe a quality the other can’t, and the seller of the good stuff can walk away. It explains why brand-new cars lose value the instant they’re “used” (buyers assume there must be a reason you’re reselling), why nobody trusts a suspiciously cheap listing, and why “no questions asked” markets breed junk. If you can’t verify quality and the sellers can self-select out, expect the average to rot.

The same trap, three markets

Lemons aren’t a car thing — they’re an information thing. The identical mechanism reappears anywhere the party who knows their own quality gets to choose whether to deal. In every case below, the good types get driven out and the pool curdles toward the bad.

Insurance — the death spiral. An insurer sets a premium for the average customer. But who’s keenest to buy generous health cover? The people who expect to use it — the already-sick, the high-risk. They flood in; claims run high; the insurer raises premiums to cover them. Now the premium is a bad deal for anyone healthy, so the healthy cancel. That worsens the risk pool, so premiums rise again, so the next-healthiest tier drops out… This runaway loop is the insurance death spiral, and it’s just the lemons unravelling wearing a different coat. Note the reversal: here the party with hidden info is the buyer, not the seller.

Credit — the risky borrower says yes. A lender who can’t tell safe borrowers from risky ones raises the interest rate to cover expected defaults. But a high rate is fine if you’re never going to repay — so the borrowers cheerfully accepting 30% interest are precisely the ones most likely to default. Push the rate up to compensate and you scare off the safe borrowers and keep the reckless ones. This is why lenders ration credit (refuse to lend at any price) instead of just charging more: past a point, a higher rate makes your pool worse, not your revenue better.

Hiring — the market for “used labour.” When you interview someone, they know their own true ability and their real reason for leaving the last job; you get a polished résumé. The strongest people are usually the ones firms fight to keep, so an unusually available candidate can carry a whiff of “why is this peach on the lot?” — the same lemons logic that dogs a car resold too soon.

When to use it

Use the three-markets lens to predict where a market will quietly rot: any deal where quality is hidden and the informed side chooses whether to participate. If you spot premiums rising while the healthy leave, or rates rising while the safe borrowers vanish, you’re watching adverse selection, not bad luck — and raising the price will make it worse.

Adverse selection vs moral hazard

This is the distinction the entire course pivots on, so let’s nail it side by side. Both are the principal–agent problem; they differ on when the hidden thing bites and what it is.

Notice the insurance row does double duty. Adverse selection: before you sign anyone, the sick disproportionately apply — you got a bad pool. Moral hazard: after they’re covered, people who’d otherwise be careful now take risks because the insurer eats the downside — good pool, bad behaviour. Same industry, two totally different failures, and the cures are different. Confusing them is the single most common mistake in the whole field.

Fighting back: screening and signalling

If hidden information is the disease, there are two cures — and the difference between them is simply who does the revealing.

• Screening — the uninformed party designs the deal so that the informed party’s choice reveals their type. You build a menu; which option they pick tells you who they are.
• Signalling — the informed party spends something costly to credibly prove their quality. They act; the act is the message.

The magic ingredient in both is a single-crossing condition: the revealing move must be cheaper (or more attractive) for the good type than for the bad type. If the two types would behave identically, nothing is revealed — the signal is just noise.

Screening — worked example (the deductible menu). An insurer can’t see who’s low-risk. So instead of guessing, it offers a menu:

A low-risk person rarely claims, so a big deductible barely costs them — they happily take Plan B and its low premium. A high-risk person expects to claim often, so a €1,000 deductible is painful — they pick Plan A. The insurer never asked anyone’s risk; the choice sorted them. That’s screening: the uninformed side engineered a self-selecting menu. (Warranties work the same way in reverse — a buyer offering “I’ll only pay for it with a warranty” screens sellers, since only a peach owner can afford to promise one.)

Signalling — worked example (Spence’s diploma). Michael Spence’s 1973 model (a Nobel alongside Akerlof): an employer can’t see who’s genuinely productive. A worker gets a degree — not necessarily because it taught them anything, but because finishing it is cheaper for a high-ability person (less effort, less agony, less risk of failing) than for a low-ability one. A degree can therefore be a credible signal of ability even if the coursework were useless, purely because of that cost gap. The good type can afford the signal; the bad type can’t fake it. Warranties offered by a seller are the used-car version (only a peach owner can afford to promise free repairs), and a founder putting their own money into the round signals belief the same way (worthless if the venture is a lemon).

When to use it

Screening is your tool when you’re the uninformed side and can design the menu (an insurer, a bank, a hiring manager writing a test). Signalling is the play when you’re the informed side and need the other party to believe you (a good seller, a strong candidate, a confident founder). Either way, sanity-check the single-crossing condition first: if the move isn’t genuinely harder for the bad type, you’ve built expensive theatre, not a filter.

Pitfall / When to use it

Adverse selection is a specific trap, not a synonym for “someone got fooled.” It needs three ingredients, and if any is missing the spiral never starts:

1. Hidden information — one side observes a quality the other can’t.
2. That info correlates with a hidden type — it isn’t random noise; it maps to “good” vs “bad” (peach vs lemon, healthy vs sick, safe vs risky).
3. Types can act on it — the informed side can choose whether to deal, so they can self-select.

Kill any leg and the trap collapses. If information is symmetric (both sides can grade quality — think regulated, disclosed, inspected goods), there’s no lemons problem. If types can’t sort themselves (participation is mandatory, or nobody can opt out — like a universal, everyone’s-enrolled insurance pool), the adverse pool can’t form. That, incidentally, is why mandates and universal enrolment are the blunt-instrument cure for the death spiral: force everyone in and the healthy can’t select out.

And the cardinal error, one more time: don’t confuse it with moral hazard. If the problem is who signed up, it’s adverse selection (hidden type, before the deal). If the problem is how they behave once signed up, it’s moral hazard (hidden action, after the deal). Different disease, different cure.

Recap

• Adverse selection = hidden information, before the deal. The other side already knows their own type; offering a deal makes the wrong types self-select in.
• The market for lemons (Akerlof): buyers price the average, so peach owners withdraw, so the average falls, so the market can unravel — bad drives out good, and nobody even lied.
• The same trap hits insurance (the sick rush in → death spiral), credit (risky borrowers accept high rates → credit rationing), and hiring — sometimes the seller holds the hidden info, sometimes the buyer.
• Adverse selection vs moral hazard: type vs action, before vs after, wrong-people-sign-up vs signed-people-misbehave. This split is the spine of the course.
• Two cures: screening (uninformed side builds a self-selecting menu) and signalling (informed side sends a costly-but-credible proof). Both need the move to be cheaper for the good type than the bad type — otherwise it’s theatre.

Next up — lesson 4, “Agency Costs & the Alignment Toolkit” — where screening and signalling join monitoring and incentives in the full kit for closing the gap between principals and agents.