Asymmetric information — when one side knows more

Two cars that look the same

Amara wants to buy a used car. Two cars of the same model and the same year stand side by side. Both are clean. Both start first time. Both sellers ask the same price.

One of the cars has been looked after carefully. The other has a gearbox that will fail within a few months. The seller of that car knows it. Amara does not, and she cannot find out by looking.

So what should Amara pay? If she pays a good car's price, she may be buying the bad one. If she offers less, the owner of the good car will refuse. The theory of markets you have built so far assumes that buyers and sellers both know what is being traded. Here that assumption breaks. One side knows much more than the other.

The lecture on quasi-public goods ended with exactly this problem: markets also fail when one side knows much more than the other. This lecture shows how such a gap can stop good trades from happening, and in the worst case can shrink a market until almost nothing is left.

Information is part of the product

When you buy a car, you buy a machine and also a belief about it: how long it will last and how often it will break. If you cannot check that belief, the market cannot price it properly.

The important word is "one side". Many deals involve uncertainty for everyone. That is not asymmetric information. The problem appears only when the uncertainty is lopsided, because then the side that knows more can act on what it knows.

Gaps like this are everywhere. Sellers know their products, insurance buyers know their own health, job applicants know their own ability, and borrowers know how risky their plans are.

Before the deal and after it

Information gaps cause two different problems. The way to tell them apart is to ask when the hidden thing matters: before the deal is made, or after it.

Adverse selection comes from hidden information before the deal. One side cannot see what kind of car, customer or worker it is dealing with. As a result, the people who choose to take part are mostly the ones the other side would least like to deal with. "Adverse" means unfavourable, and "selection" means who ends up in the deal.

Moral hazard comes from hidden action after the deal. Once the deal is signed, one side changes how it behaves, because someone else now carries part of the cost of its choices. The other side cannot watch closely enough to stop it.

The market for lemons

In 1970 the American economist George Akerlof published a paper called "The Market for 'Lemons'". In American slang, a "lemon" is a car that turns out to be bad. Akerlof showed that hidden quality can do more than make a market a little less efficient. It can make the good part of a market disappear.

Start with the simplest version. Half the used cars on sale are good, and worth $10 000. Half are lemons, worth $5 000. Sellers know which kind they own. Buyers cannot tell. A buyer who will not risk more than the average value offers this many dollars:

Now think like a seller. The owner of a lemon is delighted: $7 500 for a car worth $5 000. The owner of a good car refuses: $7 500 for a car worth $10 000 is a bad deal. So the good cars are taken off the market. Buyers soon notice that only lemons are left, and they offer $5 000. The good cars have been driven out, even though buyers would happily pay $10 000 for a car they knew was good.

That is it. That is the whole idea. The low offer drives out the best cars, and their leaving makes the next offer lower still.

With more than two kinds of car, this happens one step at a time. Try it in the model below. It is invented, but it follows Akerlof's logic exactly.

What did you see? At 20%, the best four cars left in the first round. Each round, the average car on sale got worse, so the offer fell, and the best remaining owners left. After six rounds, only the $1 000 car was sold. Yet every one of the ten cars was worth more to a buyer than to its owner. Every sale would have made both sides better off. Almost none of them happened.

At 50%, the market stopped with three cars sold. At 100%, nothing unravelled at all, because the offer for an average car, $11 000, was above every owner's value. So the size of the damage depends on two things: how much buyers gain from a car, and how widely quality varies. The model leaves out everything sellers and buyers might do about the problem, such as inspections, warranties and reputation. Those are the subject of the next lecture.

This is a market failure, as with externalities: too few trades happen, and the lost gains are a deadweight loss. But nothing spills over onto a third person. The cause is the gap in what the two sides know.

Adverse selection beyond cars

The same logic appears wherever price is set for an average that the buyer cannot see behind.

Hiring works the same way. An employer who cannot judge ability offers a wage based on the average applicant. The most able applicants can earn more elsewhere, so they are the most likely to refuse.

Lending works this way too. Two invented borrowers each want $1 000 for one year. Kwame's safe project will certainly return $1 150. Lucía's risky project returns $1 500 half the time and nothing half the time, and if it fails she repays nothing. At 10% interest, both borrow: Kwame keeps $50, and Lucía expects $0.5\times (1500-1100)=\$200$. At 20% interest, Kwame would make a loss of $50, so he stops borrowing. Lucía still expects $150, so she stays. Raising the interest rate has driven away the safe borrower. The bank's expected repayment per loan actually falls, from $825 to $600. So banks may refuse some loans rather than charge more, an argument made by Joseph Stiglitz and Andrew Weiss in 1981.

A real case: Harvard's health insurance, 1995 to 1997

Moral hazard: what happens after the deal

Adverse selection is about who signs. Moral hazard is about what people do once they have signed.

Diego buys insurance for his phone. Before, a broken screen would have cost him $150. Now it costs him almost nothing, so he stops using a protective case. What is hidden is his behaviour after signing, because the insurer cannot watch him. The insurance itself has made the insured event more likely.

The same pattern appears in health care. When a visit to the doctor is free to the patient, people use more care, because the cost is paid by the insurer. The most famous test of this was the RAND Health Insurance Experiment in the United States, which ran from 1971 to 1982. About 2 750 families, more than 7 700 people in six places, were placed at random in different plans. Some received free care. Others paid 25%, 50% or 95% of their bills, up to a limit. People who paid 95% spent about 30% less on health care than people whose care was free. For the average participant, the lower spending did not make health worse. The exception was the poorest and sickest participants, who did better with free care for some conditions. (RAND Corporation, research brief on the Health Insurance Experiment.)

Read that result carefully. Free care led to more use, which is moral hazard. But some of the extra care mattered for the poorest and sickest. Moral hazard is real, but it does not mean every extra visit is wasted.

Moral hazard also appears at work and in finance. A worker on a fixed salary who is rarely supervised may put in less effort than the employer pays for. A bank that expects a government rescue keeps the gains when its risky bets pay off, while taxpayers may bear the losses. This is called "too big to fail". Many economists see it as one factor behind the financial crisis of 2008, though how large a factor is still debated.

The structure is always the same. One person, the agent, takes actions. Another, the principal, bears part of their cost and cannot fully observe them. This is the principal–agent problem.

Now sort some cases yourself. For each one, ask the timing question: is something hidden before the deal, or after it? And is anything hidden at all?

Notice that insurance appeared in two boxes. The smoker buying life insurance was adverse selection. The student and his phone case was moral hazard. The market does not decide the box. The timing does.

Three classic mistakes