Where Adverse Selection Strikes

You’ve watched the used-car lot eat itself, and you’ve extracted the general shape: hidden information about quality selects who is willing to trade, so the very fact that a deal is available to you is itself a clue about how good it is. That’s the engine. This lesson is the field guide — the same engine, bolted into five very different machines.

Because once you know the signature, you start seeing it everywhere: the health plan that keeps getting pricier, the loan you were offered at a punishing rate, the colleague who’s suspiciously eager to leave, the five-star freelancer with no track record, the startup raising a round the founders quietly think is overpriced. None of these look like a used-car lot. All of them are one. Let’s walk the lot in each.

Insurance and the death spiral

This is adverse selection’s natural home — the arena Akerlof himself pointed at, and the one where the model does the most economic damage in the real world. The twist that makes insurance special: the informed side is the buyer, not the seller. On the car lot the seller knows the hidden quality. In insurance, you — the applicant — know your own health, your family history, that cough you haven’t mentioned. The insurer is the one guessing.

The mechanism is the same filter, run in reverse. The people keenest to buy generous cover are, on average, precisely the people who privately expect to use it. Offer everyone a single price, and you’ve built a magnet for exactly the customers who cost the most.

Let’s put numbers on it. Suppose an insurer faces a pool of 1,000 people who quietly fall into three risk tiers — but the insurer, unable to see inside anyone, must charge a single community-rated premium equal to the average expected claim.

The fair average premium is $3,300. Now watch the selection bite. A low-risk person expects only $1,000 of claims but is asked for $3,300 — a lousy trade, so they drop the plan (or never buy it). The high-risk person expects $6,000 of claims for a $3,300 premium — a gift, so they cling to it. The low-risk tier walks.

Recompute with the 300 low-risk gone. The pool is now 700 people costing $3,000,000, so the new average — and the new premium — is $3,000,000 / 700 ≈ $4,286. But now the medium-risk person, expecting $3,000 of claims, is being asked for $4,286. Another bad deal; they leave too. Recompute again: 300 high-risk people costing $1,800,000, premium $6,000. The plan now sells only to the people who cost exactly what they pay, at a price most people fled long ago. That upward ratchet — each exit making the next premium worse — is the death spiral, and it is the used-car unravelling with the labels swapped.

Here it is as the same simulator you drove in the intro, only now each tile is an applicant instead of a car: green tiles are healthy, cheap-to-insure people; red tiles are chronically ill, expensive ones. “Good cars driven out” becomes “healthy people leaving the risk pool.” Scrub the rounds and watch the healthy go dim tier by tier as the average cost — the premium — climbs.

Notice the two levers that stop the spiral, because they foreshadow the whole next lesson. Drag the hidden risk toward zero and the insurer can price each person accurately — that’s risk-based underwriting, and it cures the market but leaves the sickest facing unaffordable prices. The other fix is to forbid exit: make coverage mandatory (or heavily subsidise it) so the healthy can’t flee the pool. That’s the real reason health systems pair community rating (one price for all) with an individual mandate — the mandate exists precisely to stop the death spiral that community rating alone would trigger. Hold that thought; the cures lesson unpacks it.

Credit and lending

Now flip to money, and to the arena where adverse selection quietly rewrote how economists think about interest rates. A bank lends at some rate. The intuitive move, if defaults look likely, is to raise the rate to compensate. But here’s the trap: raising the price selects the borrowers you least want.

Think about who is thrilled to borrow at 25%. Not the cautious entrepreneur with a steady project returning 8% — at 25% their loan is a guaranteed loss, so they simply don’t apply. The borrower who happily signs at 25% is the one betting on a long-shot with a huge payoff if it hits, who reasons: “If my gamble pays off, I can afford any rate; if it fails, I default and the bank eats it either way.” The high rate screens out the safe borrowers and screens in the reckless ones.

Put a number on it. A safe borrower has a project returning a reliable 8%; they’ll borrow only while the rate stays below 8%. A risky borrower has a project that returns 40% with probability ½ and 0% otherwise (expected return 20%) — and crucially, they only repay in the good state. To that borrower, a rate of 25% costs nothing in the bad state (they’ve defaulted anyway) and is easily covered in the good state. So as the bank pushes its rate from 8% up past the safe borrower’s ceiling, the safe types vanish and only the gamblers remain. The bank’s expected repayment can actually fall as it raises the rate, because the rate hike has poisoned the pool of who borrows.

This is why banks don’t simply price risk away with ever-higher rates. Past a point, they cap the rate and ration credit instead — lending to fewer people than want loans at the going rate, turning applicants away rather than raising the price. That’s the famous Stiglitz–Weiss result: credit rationing is a rational response to adverse selection, not a market failure to be competed away.

Labour markets

Adverse selection doesn’t stop at things you buy — it operates on people you hire. The insight, sharpened by Bruce Greenwald, is that the workers most willing to leave a job at a given wage, or to accept a low one, may be exactly the ones their current employer is quietly relieved to lose.

Here’s the asymmetry. Your current employer has watched you for years — they know your true quality far better than any outside firm reading a CV. So when a worker comes onto the market, the hiring firm faces a lemons problem: why is this person available? A genuinely excellent worker is usually retained — counter-offered, promoted, kept — so the pool of job-changers is tilted toward the people other firms chose not to fight for. The outside firm, knowing this, discounts what it offers, which makes switching even less attractive to the good workers, who therefore stay put. The market for used workers thins from the top down, just like the car lot.

The sharpest version is the layoff stigma. When a firm cuts staff, outsiders can’t cleanly tell who was let go for blameless reasons (the division closed) versus who was quietly a lemon the firm was glad to shed. That uncertainty taxes everyone laid off — the “used-worker” discount — and it helps explain job lock: good employees clinging to a current role partly because stepping onto the open market is itself read as a bad signal. Notice the cure hiding in plain sight — glowing references, verifiable track records, portfolios, credentials — all ways to shrink the information gap, which is the next lesson’s whole business.

Online marketplaces and dating

Wherever strangers trade and quality is hidden, the lemons filter switches on. eBay in its early days was a pure Akerlof laboratory: a photo and a description, a seller who knows the item’s real condition, a buyer who doesn’t, and a shipping label between them. Left alone, the best sellers — whose goods are genuinely as described — get pooled with the worst and priced like everyone else, so they have every reason to leave. The eager counterparty is, once again, a mild warning: why is this listed so cheap, by someone with nothing to lose?

The same logic runs through freelancer platforms (is the five-star newcomer a hidden gem or a gamble?) and dating apps (the strikingly available, strikingly eager match invites the question what does everyone who’s already met them know that I don’t?). It would be a bleak picture — except these markets mostly work, and they work because of the patch you can already name: reputation systems. Ratings, reviews, verified histories, buyer protection and repeat-play all convert hidden quality into something a stranger can partly verify, which is why eBay didn’t collapse into a bazaar of pure lemons. Reputation is a cure, and we’ll give it its due next lesson — for now, just register that the eager, unrated counterparty is the one to scrutinise.

IPOs and securities

Finally, the arena where the informed side is a company selling a slice of itself. A firm’s managers know its true prospects far better than the investors they’re selling shares to. So ask the lemons question: which firms are most eager to sell equity at today’s price? Disproportionately, the ones whose managers privately suspect the stock is overvalued — selling something for more than they think it’s worth is a good trade for them. Firms that believe they’re undervalued are reluctant to issue cheap shares and dilute existing owners.

Rational investors see this coming, so they discount newly issued equity — which is a big part of why a company’s share price often drops on the announcement of a new stock issue, and why firms, per the pecking-order logic (Myers–Majluf), prefer to fund themselves first from internal cash, then debt, and only reluctantly from new equity. The whole edifice of underwriting, mandatory disclosure, audited accounts, prospectuses and lockup periods exists to shrink exactly this information gap — to let the good firms credibly separate themselves from the ones cashing out at a rich price. Same filter, same cure family, pinstripe suit.

The common signature

Step back and the five arenas rhyme perfectly. In every one, the party who is most willing to trade at the going terms is, on average, the party you’d least want — because their private information is exactly what makes them willing. And in every one, the uninformed side, sensing this, rationally retreats: it lowers the price it’ll pay, or rations quantity, or exits the market entirely. That’s the universal shape.

Sort the arena: who holds the hidden information?

Here’s a distinction that trips up even people who “get” adverse selection — and it’s worth getting sharp. In the classic lemons story, the informed side is the seller of a good or a risk: they know the quality and are trying to offload it. But in insurance and lending, the informed side is the buyer — the one seeking the cover or the loan — who privately knows the risk they’re bringing into the deal. Same filter, opposite side of the table. Sort each situation by which side holds the private information.