What the Seller Knows: The Problem Prices Cannot Solve

The Lot on the Edge of Town

You are standing on a used car lot. The man across the desk knows everything about the car in front of you. You know what he decides to tell you.

He knows about the noise the transmission makes when the engine is cold. He knows the previous owner traded the car in after three months. He knows whether the number on the odometer is real.

You know the price, the mileage on the sticker, and how the seats feel.

That gap - one side of a deal knowing more than the other - is one of the most useful ideas in economics, and one of the least talked about. It explains why the lot has a service bay behind it, why “certified pre-owned” exists, why your car insurance has a deductible, and why the argument over health insurance never seems to resolve.

It also produces a prediction that sounds absurd the first time you hear it: that a market with this kind of gap can destroy itself.

The Thought Experiment That Predicted Collapse

In 1970, George Akerlof published a paper in the Quarterly Journal of Economics called “The Market for Lemons: Quality Uncertainty and the Market Mechanism.” It won him a share of the 2001 Nobel Memorial Prize in Economic Sciences, thirty-one years later.

The setup is simple. We will use round numbers.

Two kinds of used cars sit on the lot. A good one is worth $12,000 to a buyer who knows what he is getting. A bad one - a lemon - is worth $6,000. Sellers know which car is which. Buyers do not.

A buyer who cannot tell the difference has to price the risk. If half the cars on the lot are good, he offers something near the average: $9,000.

Now watch what happens at $9,000.

The owner of a good car looks at that offer and refuses. His car is worth $12,000 to someone who can verify it, and he knows it. The owner of a lemon accepts immediately, because $9,000 for a $6,000 car is a fine day.

So the cars that sell are the bad ones. Buyers notice. They revise their estimate of the mix, decide that more than half the lot must be lemons, and offer less. The good cars leave. The offer falls again.

Akerlof’s model shows the market thinning from the top until, in the limiting case, no market for good used cars exists at all.

If you have read about prices as signals, this is the dark version of that idea. A price can carry knowledge into a market. It can also carry ignorance, and when it does, the market can eat itself.

That is a different problem from the one Hayek described in The Information Problem. There, knowledge is scattered across millions of people and no planner can gather it. Here the knowledge is not scattered at all. It is sitting in one person’s head, and he has a reason not to share it.

The Section Nobody Remembers

Here is what the paper is famous for: the collapse.

The paper has a fourth section called “Counteracting institutions.” Akerlof did not stop at the collapse. He asked what grows in the space where trust used to be, and he listed the answers - guarantees, brand names, chain stores, licensing.

That list is the real story, because the used car market did not collapse. It is enormous. You can buy a car this afternoon.

So what is standing between you and the lemon?

• The warranty. A dealer who offers twelve months of coverage is making a claim that costs him money if it is false. That is what makes it credible. A warranty is expensive for the seller of a bad car and cheap for the seller of a good one, so offering one separates the two types without anyone inspecting anything.
• The brand and the franchise. A manufacturer who sells you a car and then disappears does not get to sell you the next one. Reputation is a hostage. It is slow to build and fast to lose, which is precisely what makes it useful.
• The inspection. A pre-purchase check by a mechanic you hired moves the knowledge from his side of the table to yours. It costs about $150 and it is the cheapest insurance in the industry.
• The disclosure. On eBay Motors, sellers post photographs, service records and descriptions. Gregory Lewis tested what that does to prices in a 2011 paper in the American Economic Review. Voluntary disclosure, he found, works like a contract to deliver the car shown in the listing. Disclosures are important determinants of price, and when disclosure becomes more costly, sellers disclose less and cars sell for less.
• The price itself. A car priced $4,000 below book value is not a bargain. It is a message.

Notice the pattern. None of these devices requires the buyer to become an expert. Each one changes the cost of lying. That is the whole trick, and it is the same trick that makes a degree worth something as a signal, which is the subject of The Credential Con, or that makes a restaurant with an open kitchen worth more than one without.

The market’s answer to the information problem is not better information. It is institutions that make bad information expensive.

The Same Problem, Reversed

Everything above is about the seller knowing more than the buyer. Insurance runs the other way, and it is harder to see, because the person with the private information is you.

Think about what an insurer is buying. It is not buying your car or your house. It is buying a promise about your future, and you are the only person with a rough idea what that promise is worth.

Let us use round numbers again.

A hundred drivers want cover. Fifty of them will cost the insurer about $1,000 a year in claims: careful drivers, garaged cars, no teenagers on the policy. Fifty will cost about $3,000: the ones who drive at night, park on the street, and have a seventeen-year-old at home.

The insurer cannot tell which is which. It can only price the pool. Break-even is $2,000 a head.

The careful drivers look at that and do the arithmetic you would do. They are subsidizing the other fifty by $1,000 a year. Some of them leave. Some buy the thinnest policy the law allows. Either way the pool now holds a larger share of the expensive drivers, so the break-even price rises, so more of the careful ones leave.

The spiral runs the other way here, but it is the same spiral.

Michael Rothschild and Joseph Stiglitz worked out what a competitive market does about this in a 1976 paper in the same journal, “Equilibrium in Competitive Insurance Markets.” The answer is that it stops offering one price.

It offers a menu.

The Deductible Is a Question

A policy with a $1,000 deductible and a policy with a $250 deductible are not the same product at different prices. They are two questions with a price attached, and the answers sort the customers.

The careful driver, who does not expect to claim, takes the high deductible and the lower premium. He is buying the right to keep his own money. The driver who expects to claim takes the low deductible, because he knows something about his coming year that the insurer does not.

That is screening. The insurer did not learn anything about you by asking. It learned by watching which contract you chose, because choosing costs you something.

Once you see the deductible this way, the angry version of the argument falls apart. The deductible is not a trick to make you pay twice. It is the device that keeps the careful driver in the pool, which is what keeps the premium from rising for everyone. It is also the thing that makes the low premium possible in the first place.

Moral Hazard Is Not a Character Flaw

There is a second problem, and its name does it no favors.

Kenneth Arrow’s 1963 paper on the economics of medical care identified it: insurance changes the price a person faces at the moment he uses a service, and people respond to prices. If the visit costs you nothing at the counter, you go more often than you would if it cost $150. Some of those visits are unnecessary. Some are the visit where something was caught early.

“More often” is not fraud. It is not a moral failing either. It is the most reliable pattern in economics: lower the price, and people take more of the thing.

The professional name for it - moral hazard - came out of the insurance industry, where it once meant something close to arson. That history is why the conversation about it turns hostile within a minute. The mechanism is duller than the name. When the marginal cost of a service is zero, the quantity demanded rises.

Every insurance contract is therefore a compromise between two things we want that pull in opposite directions: spreading risk as widely as possible, and keeping a price at the point of use so the service is not overused.

The co-pay is the second thing. It is not a punishment. It is the price, put back.

Where the Theory Meets Data

Here is what separates this from ideology.

The theory is clean, and clean theories are dangerous. Akerlof’s model can be made to predict trouble in any market with a quality gap. If it were right everywhere, most markets would not exist. But they do exist. So the interesting question is not whether asymmetric information is possible. It is whether it is large enough, in a given market, to matter.

Economists have spent thirty years measuring it, and the answers do not line up.

The first result is the one to focus on. Pierre-André Chiappori and Bernard Salanié took data on French car insurance contracts and accidents and looked for the signature that adverse selection leaves: the customers with more cover should be the ones who claim more. They did not find it.

That does not falsify the theory. It means the theory is not a license. You cannot point at a market, note that one side knows more, and conclude that it is failing.

In the UK annuity market, where turning a pension pot into an annuity was, for most people, not optional, Amy Finkelstein and James Poterba found the pattern clearly: how long annuitants lived tracked the features of the contracts they picked. The same theory, two markets, opposite results. What differs between them is the design of the contracts and the rules around them, which is a fact about institutions rather than about human nature.

The Case Where the Sorting Has a Victim

Now the uncomfortable part.

Screening works by separating people into groups and charging each group its own price. For cars, that is a good outcome. Nobody mourns the lemon.

For people, it means the person with a condition that began before he bought the policy is, from the insurer’s side of the table, the expensive driver. The price he is offered reflects a real expected cost. The insurer is not being cruel; it is pricing a risk it can see. And the result is an outcome most people, economists included, find unacceptable.

This is where the free-market argument usually overreaches. It is true that the price is doing its job. It is also true that the job produces something we do not want, and saying “the price is correct” does not answer the person who cannot buy cover.

The reverse overreach is just as common. “The market produced a bad outcome, therefore the government should set the price” skips the step where you ask what the problem actually is.

It also skips the step where you identify which market failure you are looking at. The other classic one, the externality, is a different shape entirely: the price is wrong because a cost has been left out, and the fix is to put the cost in, through a price, a right or a liability. That case is set out in Externalities - When the Price Is Wrong. Here the price is not wrong. It is doing exactly what a price does when the information underneath it is uneven. Confusing the two is how a policy ends up aimed at the wrong target.

If the problem is that one side cannot see what the other knows, the remedy is information: disclosure, standardized contracts, mandatory participation so the pool stays whole, a backstop for the genuinely uninsurable. If the problem is defined as “the price is too high,” the remedy is a cap - and a cap on a risk pool does what a cap does everywhere else. It suppresses the signal and brings back the shortage, which is the lesson of What Price Controls Make Invisible.

The distinction matters because the two remedies fail in different ways. Disclosure can be gamed, but it does not empty the market. A cap cannot be gamed, and it does.

And there is one piece of arithmetic that no amount of good intention changes. If you require insurers to charge the healthy and the sick the same price, and you do not require the healthy to buy, the healthy leave, the pool becomes the sick, and the price rises until the market empties. That is not a prediction about greed. It is the same spiral from the start of this piece, running on the same logic.

The Two Questions

None of this resolves into a slogan, which is why it is worth knowing.

What it gives you is a pair of questions to ask the next time a price looks wrong.

The first: what does the other side know that I do not? On a car lot, at an insurance broker, in a hospital, at a job interview. The answer is usually specific, and it usually explains the shape of the deal better than the size of the number.

The second: what is this price doing besides being high?

A deductible is doing something. A warranty is doing something. A suspiciously cheap car is doing something. A premium that a sick person cannot pay is doing something too, and the thing it is doing is the reason the politics of insurance is so hard. Someone has to carry the cost of the unknown risk. The only real argument is about who, and who decides.

That is not a question economics can answer for you. It can only tell you what you are choosing between.