The Used-Car Model

In 1970 a young economist named George Akerlof wrote a short paper that three journals rejected — one editor sniffed that if it were true, economics would be different. It was true, economics did change, and Akerlof collected a Nobel Prize for it. The paper was about used cars. Not because Akerlof cared about carburettors, but because a second-hand car lot is the cleanest possible laboratory for one unsettling fact: when the seller knows the quality and the buyer can’t, a market can destroy trades that would have made everyone better off — with nobody in it behaving badly.

This lesson builds that model by hand. No hand-waving, no “trust me, it collapses.” We’ll put real dollar figures on peaches and lemons, compute the price a rational buyer offers, watch the good cars walk off the lot one tier at a time, and end at the eerie knife-edge where the whole market can implode to nothing. By the end you’ll be able to reconstruct Akerlof’s result on the back of a napkin.

The setup: gains from trade, and one thing the buyer can’t see

Start with a market that should work beautifully. There are used cars, and each one is either a peach (reliable, will run for years) or a lemon (a money-pit that will die on the motorway). Two facts make this a market at all:

• Every car has a buyer who values it more than its owner does. The owner has driven it, wrung out the joy, and would happily take cash instead; the buyer needs a car and values it a little higher. Economists call that gap the gains from trade — the surplus a sale creates. When gains from trade exist for every car, every car should sell. A world where some of them don’t is leaving free money on the table.
• Only the seller knows the quality. The person who owned the car knows every rattle and warning light. The buyer, kicking the tyres for ten minutes on a Saturday, genuinely cannot tell a well-loved peach from a lemon dressed up with a car wash.

That second fact has a name, and it is the engine of this entire course.

Asymmetric information is when the two sides of a potential deal do not have the same information about the thing being traded — specifically, one side knows something material about its quality (or type) that the other side cannot verify before agreeing. In the car market, the seller knows the car’s type; the buyer doesn’t; and no cheap kick of the tyres closes the gap.

Two things it is not. It is not merely “the seller knows more trivia about the car” — the asymmetry has to be about something the buyer would pay differently for if only they could see it. And it is not about the seller lying. Every seller in this lesson tells the truth or says nothing at all. The damage comes entirely from what the buyer can’t check, not from anything the seller does. Hold onto that; it’s the most counter-intuitive part of the whole model, and we’ll hammer it again at the end.

Reservation prices and the pooled offer

Here’s the move that starts the rot, and it’s completely reasonable.

A reservation price is the worst deal you’ll accept before walking away. For a seller it’s the lowest price they’ll take (below it, they keep the car). For a buyer it’s the most they’ll pay. Reservation prices are where “how much I value this” turns into “what I’ll actually do.”

Now put yourself in the buyer’s shoes. You can’t tell a peach from a lemon, so you can’t offer the peach price (you might be buying a lemon) or the lemon price (you might be handing back a peach for scrap money). The only rational thing is to offer what the car is worth on average — the expected value of a random car from the pool in front of you. If the lot is half peaches and half lemons, you pay a blend of the two. That blended number is the pooled offer: one price for a pool of cars you can’t sort.

Let’s make it concrete. Take our two types, with a buyer valuation and a seller reservation for each:

Notice every car should trade: the buyer values each one $1,000 above its owner’s floor. In a world with no hidden information — where the buyer could read the quality off a windshield sticker — a peach sells for something between $9,000 and $10,000, a lemon between $3,000 and $4,000, and everybody goes home happy.

Now hide the quality and suppose the buyer believes the lot is 50/50. The pooled offer is the expected value:

And $7,000 is a disaster dressed up as fairness. Compare it to each seller’s reservation:

The “average” price is above a lemon’s floor and below a peach’s floor. So the peach owners take their good cars home, the lemon owners rush to sell, and the buyer — who offered a perfectly fair average — ends up buying a lemon for $7,000. The pooled offer didn’t split the difference; it selected which sellers would take it. That’s why it’s called adverse selection: the pricing quietly picks out the bad types and repels the good ones.

The unravelling, round by round

The two-type story stops after one step: peaches leave, only lemons remain, done. But real quality comes in grades, and that’s where the model shows its teeth — because each tier that leaves drags the average down onto the next tier’s head.

Let’s build a lot with six quality grades, A (best) down to F (worst), in equal numbers. To keep it clean, say a seller values each car at 80% of what a buyer would — so gains from trade exist for every single car (the buyer always values it 25% more than the owner), yet the owner’s floor rises with quality:

Each round, the buyer offers the average buyer-value of the cars still for sale. Any seller whose floor is above that offer withdraws. Then the buyer recomputes the average over whoever’s left, and offers again. Watch it cascade:

The lot started with six healthy grades, every one of which should have traded, and it ends with only grade F changing hands, at $2,000. Look at the feedback loop that did it: every time a good tier leaves, the pool that remains is worse, so the honest average drops, so the offer drops, so the next tier up finds its floor is now above the price and leaves too. Each exit manufactures the next exit. Nobody coordinated this, nobody cheated — arithmetic did it.

This is the moment to stop reading and watch it. Below is the same market as a simulator: 24 cars, each coloured by a true quality only the seller can see. Keep the hidden-information slider at 100% (buyers see nothing but the average) and drag the round scrubber left to right. The top tiles — the good cars — go dim one band at a time as their owners refuse the pooled price, and the average-quality bar sinks after every exit. That dimming cascade is the table above, animated.

The knife-edge: when the market collapses all the way to zero

Six grades bottomed out at grade F — the very worst still traded. But make quality continuous and the collapse can be total: in the limit, no car trades at all, even though every possible sale would create surplus. This is Akerlof’s sharpest result, and it’s worth seeing the argument in full because it feels impossible until you follow the algebra.

Set it up as cleanly as possible. A car’s quality is the seller’s own valuation, spread uniformly across the range $0 to $10,000 — some cars are near-worthless, some nearly new, and everything in between, in equal density. And say the buyer values every car 50% more than its owner does. That’s a huge, universal gain from trade: for any car, buyer value is 1.5× seller value, so a sale always creates surplus. Surely this market can’t fail?

Follow the buyer’s logic at any offered price $p$:

1. Who sells at price $p$? Only owners whose valuation is at or below $p$ — those with a car worth $$p$ or less to them. So the cars actually on offer are the ones with quality spread uniformly from $0 up to $$p$.
2. What’s the average quality of that pool? Uniform from 0 to $p$ means the average seller-value is the midpoint, $p/2$.
3. What will the buyer pay for it? They value a car 50% above the owner, so their willingness to pay for the average car on offer is $1.5 \times (p/2) = 0.75p$.

Now stare at that. The buyer offered $p$, but once they reason about which cars that price attracts, the pool is only worth $0.75p$ to them. The offer always exceeds what the attracted cars are worth — by exactly a quarter. So the rational buyer lowers the offer. But at the new, lower price the same logic repeats: whatever you offer, the cars it attracts are worth only 75% of it. There is no positive price where the buyer breaks even. The only price that survives is $p = 0$, and at $0 nothing sells.

A market where the buyer values every car 50% more than its owner — gains from trade dripping off every trade — clears zero cars. That’s the knife-edge: hidden information didn’t just skim the top tier off, it unravelled the entire thing to a single point at the bottom. The good cars, the mediocre cars, all the cars stay home, and a mountain of surplus that everyone agreed existed simply never gets realised.

Whether a market lands at “only lemons trade” or “nothing trades” turns on how fast quality and seller-floors rise relative to the buyer’s markup. When the buyer’s premium can’t keep pace with the average dropping as good cars exit, the floor of the market falls out from under it completely.

When the used-car model applies

The car lot is a metaphor with a precise anatomy, and it pays to know exactly when a real situation shares that anatomy — because when it does, you can predict the unravelling, and when it doesn’t, reaching for this model will mislead you. Look for all of these:

• Hidden quality that varies. The items genuinely differ in a way that matters to the buyer, and that difference is known to the seller but not cheaply checkable by the buyer. Identical items, or quality the buyer can verify in five minutes, don’t unravel.
• A pooled price. The uninformed side is forced to price the good and bad together because it can’t sort them. One price, many hidden types.
• Gains from trade worth protecting. There are good deals being destroyed — the market isn’t collapsing because the goods are simply bad, but because pricing is repelling the good ones.
• The informed side can walk. Good types must have a viable outside option (keep the car, stay uninsured, don’t take the loan). If they can’t withdraw, there’s nothing to select.

That last point hides the real subtlety, and it’s the pitfall to carry forward. The model does not predict that “markets with bad average quality fail.” It predicts that markets fail when a pooled price selects the good types out — which needs the buyer to be genuinely unable to verify quality cheaply. The instant verification gets cheap — a trusted inspection, a warranty, a rating, a credential — the pooling breaks, the good types stop fleeing, and the market heals. That’s not a footnote; it’s the entire strategy of every cure you’ll meet next. The disease is the information gap, so the treatment is always to shrink it.

Where this leaves us

You can now derive the whole thing from scratch: gains from trade exist for every car, hidden quality forces a pooled offer, the pooled offer is a bad deal for the good types so they withdraw, their exit worsens the pool, the offer drops, and the next tier leaves — until only lemons trade, or in the knife-edge case nothing trades at all. Real numbers, no magic.

The next lesson, the general shape, lifts this off the car lot: the deep point isn’t about cars, it’s that hidden information about type selects who is willing to trade — so the mere fact that a deal is on offer to you is itself information about how good it is (the willing counterparty problem). After that we tour the arenas where it bites — insurance death spirals, lending, labour, marketplaces — and then, finally, the cures: signals, screening, and institutions that shrink the information gap and coax the peaches back onto the lot.