Game Theory · GT-24

Adverse Selection

Game Theory

When one side of a deal knows more than the other before the contract is signed, the market can fill up with exactly the people you least wanted.

A market failure arising from hidden information that exists before a transaction: if a seller (or buyer) can't distinguish quality types, pricing to the average draws in disproportionately more of the low-quality (or high-risk) types, driving up the average risk of whoever remains willing to transact — potentially unraveling the market entirely.

Formalized in George Akerlof's landmark 1970 paper 'The Market for Lemons,' which won him a share of the 2001 Nobel Memorial Prize alongside Michael Spence and Joseph Stiglitz for work on information asymmetry.

The Mechanism

The used-car market unraveling, one round at a time

Rounds of market adjustment Average price / quality remaining in market Round 1: price reflects average car quality Good-quality sellers exit — average quality of remaining pool drops Price falls again to match new (lower) average — more good sellers exit

Akerlof's 'lemons' result: because buyers can't tell good cars from bad ones, they'll only pay the average expected price — which is a bad deal for owners of genuinely good cars, who exit the market, dragging the average (and the price buyers are willing to pay) down further, in a cycle that can unravel the market almost entirely.

01 · THE INFORMATION ASYMMETRY EXISTS BEFORE THE CONTRACT

This is what separates it from moral hazard

Adverse selection is about hidden information the informed party already has before any deal is struck (a seller already knows their car is a lemon, an applicant already knows their health history) — this timing distinction is the key thing that separates it from moral hazard, which concerns hidden actions taken after a contract begins.

02 · UNRAVELING CAN DESTROY MARKETS THAT WOULD OTHERWISE BE MUTUALLY BENEFICIAL

Akerlof's central, startling result

In the extreme case, adverse selection can drive a market's average quality down round after round until only the worst types remain willing to transact at any price buyers will pay — a market that, absent the information asymmetry, would have supported plenty of mutually beneficial trades at fair prices for good-quality goods.

03 · REAL MARKETS FIX IT WITH SIGNALING, SCREENING, AND WARRANTIES

The theoretical failure mode is common, but so are its remedies

Vehicle history reports, third-party certifications, warranties (a seller offering a warranty signals confidence their product won't need it), and screening menus (see Screening) are all real institutional responses that emerged specifically to counteract adverse selection, restoring markets that would otherwise degrade toward the Akerlof 'lemons' outcome.

Where It Fails / Inversion

Where it fails / inversion

Not every information asymmetry produces meaningful adverse selection — if the cost of misjudging quality is low, or reputation and repeat interaction already discipline sellers effectively, the theoretical unraveling may never materialize in practice. Assuming every information gap is an adverse-selection crisis can lead to over-engineering costly certification and verification systems where lighter-touch reputation mechanisms would suffice.

How To Use It

Worked example · health insurance risk pools

Health insurers pricing a single premium to a broad, unscreened population predictably attract a disproportionate share of higher-risk applicants (who know their own health status better than the insurer does) relative to healthier applicants, who may find the pooled price unattractive and opt out — pushing the pool's average risk, and therefore the premium, upward over time. This is precisely why real insurance markets rely heavily on medical underwriting, mandated enrollment (reducing the ability of healthy people to opt out), and risk-adjustment mechanisms, all designed specifically to counteract this dynamic.

How to use it

Before pricing any deal to an unscreened population — a hiring pool, a lending portfolio, an insurance product — ask who self-selects into transacting at that price. If the answer is systematically the types you least want, you're facing adverse selection, and the fix is either better screening, mandatory participation, or a warranty/signaling mechanism — not simply adjusting the average price, which often makes the unraveling worse.

See Also

Screening → Signaling Theory → Moral Hazard → Winner's Curse →