Essay two of four

The market for lemons

A used car market where the seller knows what the car is and the buyer does not. Set how far quality varies and how much of it a buyer can see, then run the market and watch the good cars leave one round at a time.

A hundred and twenty people each want to sell a car. Each knows exactly what their own car is worth. The buyers know only that the cars in front of them are somewhere in a range. Every car is worth half again as much to a buyer as to its current owner, so there is a gain to be had on every single one of them, and under full information every one of them would sell.

What follows is not a story about dishonest dealers. Nobody in the model lies. Every seller makes the one decision available to them, which is to accept the offered price or to keep the car. Every buyer offers what the cars in front of them are worth on average. From those two entirely reasonable rules, the market takes itself apart.

01The lot

Each mark below is one car, ordered from the worst on the left to the best on the right, with its quality on the vertical scale. Set how wide the spread of quality is and how much of that quality a buyer can actually see before handing over money, then run the market. The line that appears is the quality above which an owner would rather keep the car than take the price on offer.

Figure one the cars, round by round

02Why it unravels

The loop has four steps and it closes on itself. Buyers offer a price built on what they believe the average car on the lot is worth. At that price the owners of the best cars decline, because their car is worth more to them than the offer. The cars actually on offer are therefore worse than the cars in principle available. Buyers see that, revise the average down, and offer less. Which drives out the next layer of good cars.

Step through the rounds and watch the line come down. The two series below are the average quality of the cars that changed hands in each round and the average price paid for them. They fall together, because the price is only ever an estimate of the quality, and the estimate is chasing a quantity that the estimate itself is pushing down.

Figure two quality and price over the rounds

Two things are worth noticing about where it stops. First, it does not always go to zero. If the worst car on the lot is still a decent car, the spiral runs out of room and the market settles smaller and worse than it should be, but alive. Second, if the worst car is worthless, there is no floor to stop it, and the market falls all the way through. Push the spread control to its maximum and watch that happen.

03The number that decides it

The whole spiral is one line of arithmetic. A buyer who sees a fraction of the quality and guesses the rest offers a price built from both. An owner sells when that offer is at least what the car is worth to them. Working the inequality through gives the highest quality that will still be offered for sale.

highest quality still offered = 1.5 x (1 minus visibility) x believed average / (1 minus 1.5 x visibility) With no visibility at all the expression is just one and a half times the believed average. As visibility rises towards two thirds the denominator goes to zero and every car sells whatever the belief.

That gives a clean threshold, and it is the reason the visibility control has a point where the behaviour changes rather than a smooth slope. When the worst car on the lot is worthless, the market survives only if buyers can see more than a third of what they are buying. Below a third the belief falls faster than the floor can hold it and the whole thing goes to nothing. Above a third it holds. Set the spread to its maximum and move visibility across a third to see the switch.

04Two ways to put it back together

Nothing above required anybody to behave badly, so nothing above is fixed by asking people to behave better. What fixes it is changing what a buyer's money is exposed to. The remedy control offers two ways of doing that, and they work differently enough to be worth trying one after the other.

A certificate attacks the ignorance directly. An independent inspection reveals most of what the buyer could not see, for a fee. Switch it on and the line jumps above the top of the lot: with quality visible, there is no gap between what a car is and what it fetches, so nobody has a reason to withdraw. The fee is a real cost and it is subtracted from what the seller receives, so the market that comes back is slightly smaller than the ideal one. Information is not free.

A warranty attacks the exposure instead. The seller promises to make good any shortfall against a stated standard, so the buyer is paying for the standard rather than for a guess. Switch it on and watch what happens at the two ends of the lot. The bad cars now sell, because their owners can carry the repair bill and still come out ahead. The very best cars withdraw, because a warranty pays the same for an excellent car as for an adequate one and the owner of an excellent car will not accept that. The remedy rescues the market and truncates it at the top.

05What Akerlof actually wrote

The paper is George Akerlof's, it is thirteen pages long, and it was published in the Quarterly Journal of Economics in August 1970. The used car market is the opening illustration rather than the subject. Akerlof was after something more general: that where one side of a trade knows the quality and the other does not, the bad drives out the good, and the loss falls on the honest seller who can no longer prove that their car is good.

The model above is a numerical version of that argument, not a reproduction of his. Akerlof's arithmetic is starker. In his example buyers value cars at three halves of what owners do and quality is spread evenly from zero upward, and he shows that the only price consistent with itself is zero: the market does not shrink, it vanishes. The spread control on this page reaches that case at its maximum, which is why the maximum is worth visiting once.

It is worth knowing how the paper was received. Akerlof has written that the American Economic Review and the Review of Economic Studies both turned it down as too trivial to publish, and that a referee at the Journal of Political Economy objected that if the argument were correct then no goods could be traded at all. It went into the Quarterly Journal of Economics. Thirty one years later Akerlof shared the 2001 economics prize with Michael Spence and Joseph Stiglitz for the analysis of markets with asymmetric information.

06Where the model stops

Assumptions and limits

One shot buyersEvery buyer in the model arrives once, buys or does not, and leaves. There are no repeat customers, no reputation and no word of mouth, which in a real used car market are among the strongest forces holding quality up. Their absence is what makes the unravelling as clean as it is here.
Beliefs update crudelyBuyers set this round's belief to last round's realised average. That is adaptive rather than rational: a fully rational buyer would jump straight to the fixed point in one step, and there would be no spiral to watch. The rounds are a teaching device. The place the model ends up is the same either way.
Quality is one numberReal cars are good in some ways and bad in others, and buyers can usually see some dimensions and not others. Compressing that into a single hidden scalar is what makes the picture drawable and is also the largest thing the picture leaves out.
The remedies are sketchesThe certificate is modelled as a jump in visibility with a flat fee attached. The warranty is modelled as a promise to make up any shortfall against a fixed standard, again with a flat administrative fee. Real warranties are priced, limited, disputed and sometimes not honoured, and real certification schemes can be captured by the sellers who pay for them. Neither is the clean instrument drawn here.
The warranty is a transferThe repair a warranty pays for is treated as money moving from seller to buyer rather than as work that consumes real resources, so the only cost the warranty adds to the total is its administrative fee. A treatment in which repairs consume resources would leave the market smaller than it appears here.
Nothing is estimatedThe multiplier of one and a half, the hundred and twenty cars, the fees and the fourteen rounds are chosen so that the mechanism is visible. None of them is measured from any real market, and no number on this page describes the price of any real car.
Adverse selection is not the only readingUsed cars sell below new cars for reasons that have nothing to do with hidden information, wear and depreciation among them. That a market shows the pattern is not proof that this mechanism produced it.

07Sources

The paper itself, and the numerical example in which buyers value cars at three halves of the owner's valuation and quality is spread evenly, so that no price above zero is consistent with itself. George A. Akerlof, The Market for “Lemons”: Quality Uncertainty and the Market Mechanism, The Quarterly Journal of Economics, volume 84, number 3, August 1970, pages 488 to 500. DOI 10.2307/1879431. doi.org/10.2307/1879431 Bibliographic details confirmed at en.wikipedia.org/wiki/The_Market_for_Lemons. The setup, a buyer willing to pay three halves of the seller's valuation, quality distributed uniformly, and a unique equilibrium price of zero at which no car is sold, is set out in George Georgiadis, Information Economics, Module 14: Adverse Selection, Kellogg School of Management, kellogg.northwestern.edu/faculty/georgiadis/Teaching/Ec515_Module14.pdf. Checked 2026 08 21.
Akerlof's own account of the rejections: the American Economic Review and the Review of Economic Studies declined the paper on the ground that they did not publish work on subjects of such triviality, and a referee at the Journal of Political Economy held that if the paper were correct then no goods could be traded. George A. Akerlof, Writing the “The Market for ‘Lemons’”: A Personal and Interpretive Essay, published by the Nobel Foundation, 2001. nobelprize.org/prizes/economic-sciences/2001/akerlof/article Read through the summary at equitablegrowth.org. Checked 2026 08 21.
The 2001 Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel was awarded jointly to George A. Akerlof, A. Michael Spence and Joseph E. Stiglitz for their analyses of markets with asymmetric information. The Nobel Foundation, prize summary for 2001. nobelprize.org/prizes/economic-sciences/2001/summary Checked 2026 08 21.
Warranties, vehicle history reports, brand reputation and licensing are the institutions usually named as market answers to the problem, rather than statutory ones. David R. Henderson, editor, The Concise Encyclopedia of Economics, biography of George Akerlof, Library of Economics and Liberty. econlib.org/library/Enc/bios/Akerlof.html Checked 2026 08 21.