The Paper Nobody Wanted
In 1966, a young economist named George Akerlof wrote a thirteen-page paper with an irresistible title: "The Market for 'Lemons.'" He sent it to the American Economic Review. They rejected it. Too trivial, they said. He sent it to the Review of Economic Studies. Rejected. He sent it to the Journal of Political Economy. Rejected again — this time because, as one reviewer put it, "if this were correct, economics would be different."1
The reviewer was right. Economics would be different. The paper was finally published in the Quarterly Journal of Economics in 1970, and thirty-one years later, it won Akerlof the Nobel Prize. The idea was simple enough to explain to a child, profound enough to reshape how we think about markets, and mathematical enough to make the whole thing precise. Which, if you think about it, is the hat trick of great economics.
Here is the question Akerlof asked: What happens to a market when one side knows something the other side doesn't?
The answer, it turns out, is that the market can destroy itself. Not through fraud, not through malice, not through anyone doing anything wrong — but through the perfectly rational behavior of perfectly reasonable people. This is the tragedy of adverse selection, and once you see it, you'll find it everywhere: in used car lots, in insurance markets, in job interviews, on dating apps, and buried deep inside the reason your health insurance costs so much.
The Used Car Problem
Let's say you're buying a used car. You've found a 2019 Honda Civic on Craigslist, and the seller wants $15,000. Is it worth that? You kick the tires, check the paint, take it for a spin. But you can't really know. Is the transmission about to fail? Does it burn oil when you're not looking? Has it been in a flood? The seller knows these things. You don't.
This asymmetry — the seller knowing the car's true quality while the buyer is guessing — is the seed from which everything grows.
Suppose cars come in a range of qualities, from terrible (worth $0) to excellent (worth $20,000). Let's say quality is uniformly distributed: every value between 0 and 20,000 is equally likely. Sellers will sell their car if the offered price meets or exceeds their car's true value. Buyers, who can't observe quality directly, will pay the expected value of whatever cars are actually on the market.
Here's where the spiral begins. If buyers offer $10,000 (the average of 0 to 20,000), which sellers will actually sell? Only those whose cars are worth less than $10,000. An owner of a car worth $15,000 won't accept $10,000 — they'll keep driving it. So the cars actually on the market are worth between $0 and $10,000.
But buyers are smart. They know this. If only cars worth $0–$10,000 are for sale, the expected quality is $5,000. So they revise their offer down to $5,000. Now only cars worth $0–$5,000 remain. Expected quality: $2,500. Offer drops again. And again. And again.2
The adverse selection death spiral. Each round, buyers revise down, good cars leave, and the market collapses toward zero.
In the mathematical limit, the market collapses entirely. The only car you can buy is the worst one. Economists call this market unraveling, and it's one of those results that's simultaneously obvious once you see it and deeply counterintuitive before you do. Nobody cheated. Nobody lied. The market ate itself.
Watch the Market Unravel
Theory is nice, but nothing beats watching it happen. Below is a market of 40 used cars, each with a hidden quality level. Hit "Start Market" and watch as buyers offer the average value, good cars withdraw (their owners aren't fools), the average drops, and the spiral takes hold. Drag the asymmetry slider to see what happens when buyers can partially observe quality.
🍋 Lemon Market Simulator
40 cars with random quality (green = good, red = lemon). Watch the market unravel round by round.
Notice what happens at 100% asymmetry: the market collapses almost entirely. But slide it toward 0% — giving buyers perfect information — and every car trades at its true value. The spiral is a creature of the information gap. Close the gap, and it vanishes.
When Bodies Replace Cars
Now let's replace used cars with something more consequential: human bodies. The health insurance market has exactly the same structure as Akerlof's lemon market, except instead of sellers hiding car defects, buyers are hiding their own health risks from insurers.3
Think about it. Who most wants health insurance? People who expect to need it — the chronically ill, the injury-prone, those with family histories of expensive diseases. Who least wants it? The young, the healthy, the invincible twenty-five-year-olds who think they'll live forever.
An insurance company sets a premium based on the average expected cost of its enrollees. But at that average price, the healthiest people — the ones whose expected costs are well below the premium — rationally drop out. Why pay $500 a month when you spend $100 a year on healthcare?
When healthy people leave, the remaining pool is sicker. Costs rise. The insurer raises premiums. Now slightly less healthy people drop out too. Costs rise again. Premiums rise again. This is the insurance death spiral, and it's not a metaphor — it's the same fixed-point equation as the lemon market, just wearing a hospital gown.4
This is precisely why the Affordable Care Act included an individual mandate — a requirement that everyone buy insurance, not just the sick. The mandate was the government's way of saying: we know the math, and the only way to stop the spiral is to keep healthy people in the pool. Force everyone in, and the average cost stays low enough for the market to function.5
The insurance death spiral: healthy people leave, costs rise, premiums increase, more healthy people leave. Rinse and repeat.
💀 Insurance Death Spiral Simulator
A population of 60 people with varying health costs ($100–$10,000/yr). Set a starting premium, then watch who stays and who drops. Toggle the mandate to see the difference.
With the mandate off, watch how quickly the pool shrinks and premiums skyrocket. Turn the mandate on, and suddenly the market stabilizes. Same people, same health costs — the only difference is that everyone has to play.
Signals and Screens
If adverse selection is the disease, information is the cure. But when you can't directly share information — when saying "trust me, I'm healthy" or "trust me, this car is great" is cheap talk — you need something more clever. Economists have found two main treatments: signaling and screening.
Michael Spence (Nobel 2001) asked a deceptively simple question: why do people go to college?6 Not "why do they say they go" — the standard answers about learning and personal growth. But why does the job market reward a degree so handsomely?
Spence's answer: because getting a degree is hard, and it's harder for low-ability types than high-ability types. A smart, disciplined person can grind through four years of problem sets and all-nighters. A less capable person might fail out. So a degree doesn't necessarily teach you anything the employer needs — it proves you're the kind of person who can survive the gauntlet.
This is a costly signal. It works precisely because it's expensive (in time, effort, money). If degrees were free and easy, everyone would get one and it would signal nothing. The cost is the point.
Think of a peacock's tail. It's metabolically expensive, it slows the bird down, it makes it easier for predators to catch. That's what makes it a reliable signal of genetic fitness — only a genuinely healthy peacock can afford such an absurd luxury.7 Your college degree is your peacock's tail. (Sorry.)
Rothschild and Stiglitz (1976) flipped the script. Instead of the informed party sending a signal, the uninformed party designs a menu of options that makes people reveal their type through their choices.8
Insurance companies do this constantly. They offer you Plan A: low premium, high deductible. Plan B: high premium, low deductible. Healthy people choose Plan A (they don't expect to use it much). Sick people choose Plan B (they know they will). By offering the menu, the insurer gets people to self-select into categories, partially solving the information problem without ever asking "how sick are you?"
The genius of screening is that nobody has to reveal anything explicitly. The menu does the sorting. It's like putting out two bowls at a party — one with celery sticks, one with chocolate cake — and learning something real about your guests from who reaches for what.
Two cures for information asymmetry: signaling (the informed party proves themselves) and screening (the uninformed party designs revealing choices).
Lemons Are Everywhere
Once you have the adverse selection lens, the world starts looking different.
Dating apps are lemon markets. The people most aggressively on dating apps are — on average and with many exceptions — the ones who've been unable to find partners through other channels. The best "catches" get snapped up quickly or never need the app at all. This doesn't mean everyone on Hinge is a lemon. But it does mean the pool is selected, and sophisticated daters know it. Profile verification, mutual friends, linked Instagram accounts — these are all signals designed to fight the same information asymmetry Akerlof identified in 1966.
Job markets are lemon markets. Why do employers obsess over pedigree — which school you attended, which companies you worked at? Because they can't directly observe your ability. A resume from Google signals competence, the way a college degree signals discipline. It's Spence all the way down.
Online reviews are a partial solution. Before Yelp and Amazon reviews, buying anything online was a lemon market — you couldn't inspect the goods, and the seller knew more than you. Reviews create a reputational mechanism that partially closes the information gap. Of course, now we have fake reviews, which is just the lemon problem one level up: you can't tell the real reviews from the fake ones, and the fake ones are more likely to be effusively positive...
eBay's feedback system, Carfax reports for used cars, home inspections before purchases, warranties and guarantees — every one of these is a market institution that evolved specifically to combat information asymmetry. They're all answers to Akerlof's question.
Adverse selection isn't about dishonesty. It's about structure. Even in a world of perfectly honest people, if one side knows more than the other, the market's logic can make it devour itself. The math doesn't care about your intentions. This is why "just let the free market work" is sometimes terrible advice — not because markets are bad, but because some markets need institutional scaffolding to survive their own incentive structure.
Akerlof's paper was rejected for being "trivial." But there's nothing trivial about the discovery that rational actors, free markets, and perfect honesty can combine to produce catastrophic failure. The three journals that rejected the paper missed something important: sometimes the simplest ideas are the ones that change everything. You just have to be willing to look at the lemons.