What Is Adverse Selection? The Lemons Problem Explained

Search Fintrev

Search our guides, reviews and glossary.

Adverse selection is a market problem that arises before a deal is made, when one side knows something relevant about the quality or risk of what is being traded and the other side does not. Because the uninformed party can only price for the average, the people most eager to trade at that price tend to be the ones the average flatters: owners of the worst products, or buyers carrying the highest risk. The pool of participants shifts toward bad outcomes for the uninformed side. Prices, premiums or interest rates then move against everyone, and in extreme cases the market shrinks or disappears.

What Adverse Selection Means

The term describes a selection effect: who chooses to enter a transaction. It is adverse because that self-selection works against the party with less information. An insurer quoting one price to everyone does not choose its customers. They choose the insurer, and the ones who find the price most attractive are those who privately know they are likely to claim.

Economists class adverse selection as a market failure caused by an information failure. The market can settle at a price where mutually beneficial trades fail to happen, even though a fully informed market would have completed them.

Is adverse selection the same as asymmetric information?

No. Asymmetric information is the underlying condition: one party knows more than the other. Adverse selection is one problem it can create, when the hidden information concerns a characteristic that already exists at the time of the deal (a car’s true condition, an applicant’s health, a borrower’s real repayment ability). Asymmetric information need not produce adverse selection, for instance if the uninformed side can cheaply verify the facts or the information has no bearing on price.

How Adverse Selection Works: The Lemons Problem

Akerlof set out the mechanism in “The Market for ‘Lemons’: Quality Uncertainty and the Market Mechanism,” published in The Quarterly Journal of Economics in 1970 (vol. 84, no. 3). Lemons is American slang for defective used cars. Akerlof shared the 2001 Nobel Memorial Prize in Economic Sciences with Michael Spence and Joseph Stiglitz for work on markets with asymmetric information.

Hypothetical example: why adverse selection lowers prices

Assume a used-car market where:

• Half the cars are good and half are lemons. Sellers know which they own; buyers cannot tell them apart.
• Buyers would pay $10,000 for a car they knew was good and $4,000 for a lemon.
• Owners of good cars will not sell for less than $8,000. Owners of lemons will sell for anything above $3,000.

With full information, every car trades: good cars somewhere between $8,000 and $10,000, lemons between $3,000 and $4,000.

Without information, a buyer facing a 50/50 chance prices at the expected value:

Expected value = (0.5 × $10,000) + (0.5 × $4,000) = $7,000

At $7,000, owners of good cars refuse to sell, because $7,000 is below their $8,000 minimum. Lemon owners happily accept. Buyers who reason this through realize that any car offered at $7,000 is almost certainly a lemon, so they lower their offer toward $4,000. The good cars leave the market entirely, and only lemons trade.

The average price fell not because cars got worse but because the mix of cars offered got worse, and trades that would have benefited both sides were lost. Each drop in price pushed out more high-quality supply, which justified a further drop. This unraveling is the core of the adverse selection problem.

The same structure appears whenever quality is hidden and the seller knows it: business equipment, private company stakes, loans sold between banks, or any product whose flaws only show up after purchase.

Adverse Selection in Insurance

Insurance reverses the roles. Here the buyer holds the hidden information, because applicants know more about their own health, habits and circumstances than the insurer does.

Hypothetical example: the insurance death spiral

An insurer sells one policy to two equal-sized groups. It cannot distinguish them.

• Low-risk customers expect $200 a year in claims.
• High-risk customers expect $800 a year in claims.

Ignoring costs and profit, a break-even pooled premium is the average, $500. A low-risk customer who values the cover at, say, $350 finds $500 a poor deal and drops out. The pool is now mostly high-risk, average claims rise toward $800, and the insurer must raise the premium. Any remaining moderate-risk customers now face an even higher price and leave too. If nothing interrupts the cycle, the product ends up covering only the highest risks at a premium few others will pay.

Concealment sharpens the problem. A smoker who declares themselves a nonsmoker gets cover priced for a lower risk. An auto applicant who gives a low-crime address, or claims the car is garaged at night when it sits on a busy street, shifts theft and damage costs onto the insurer. If enough applicants do this, the insurer’s pricing no longer matches its pool.

How insurers respond

Underwriting is the main defense. Life insurers typically assess health, medical and family history, occupation, hobbies, driving record and lifestyle risks such as smoking, and may require a paramedical exam or doctors’ records. They then decide whether to offer cover and at what premium, so a race car driver or a smoker pays more.

Other tools change who joins rather than what the insurer knows:

• Group coverage through an employer pools people who joined for reasons unrelated to their health.
• Waiting periods and exclusions for preexisting conditions stop people buying cover only after they learn they need it.
• Deductibles and policy menus let customers sort themselves: low-risk buyers often choose a cheaper high-deductible plan, revealing information through their choice.
• Mandates require broad participation. The U.S. Affordable Care Act paired a ban on pricing for preexisting conditions with an individual mandate penalty to keep healthy people in the pool. Congress cut that federal penalty to zero from 2019. Limiting what insurers can price on increases selection pressure unless participation is supported another way.

Adverse Selection in Lending, Investing and Trading

Credit markets

A lender knows less about a borrower’s true creditworthiness than the borrower does. Raising the interest rate to cover average default risk can backfire: cautious borrowers with safe projects find the loan too expensive, while borrowers willing to pay a high rate are disproportionately those with riskier plans or less intention of repaying. Stiglitz and Andrew Weiss showed in 1981 that this can lead lenders to ration credit, refusing some applicants outright rather than raising rates further. Credit scores, collateral requirements and income verification are all ways of narrowing the information gap.

Issuing shares

Company managers know more about the firm’s prospects than outside investors. If managers tend to issue new stock when they believe it is overvalued, investors learn to treat an equity offering as a warning sign and discount the price. Stewart Myers and Nicolas Majluf formalized this in 1984, and it underpins the “pecking order” view of corporate finance: firms prefer internal funds, then debt, and turn to new equity last because it carries the strongest adverse-selection discount.

Market making and the adverse selection cost

Market makers quote prices to anyone who wants to trade, so they cannot avoid trading with people who know more than they do. Informed traders buy when the quoted price is too low and sell when it is too high, so the market maker systematically loses on those trades.

The bid-ask spread compensates for this. Market microstructure often splits the spread into order-processing costs, inventory costs and an adverse selection cost, the part covering expected losses to informed traders. A reported adverse selection cost percentage is usually that share of the spread (or of price impact), and its definition depends on the model used.

Hypothetical example (in the spirit of the Glosten-Milgrom model)

• A stock is worth either $99 or $101, with equal probability.
• 20% of traders know the true value; the other 80% buy or sell at random, 50/50.
• The market maker sets the ask at the stock’s expected value given that someone wants to buy.

If the value is $101, buy orders come from all informed traders (0.20) plus half the uninformed (0.40): a 0.60 probability. If the value is $99, only uninformed buyers arrive: 0.40. A buy order therefore raises the probability of the high value to 60%.

Ask = (0.6 × $101) + (0.4 × $99) = $100.40 By symmetry, Bid = $99.60

The $0.80 spread is pure adverse selection: there are no processing or inventory costs in this setup. Raise the share of informed traders and the spread widens, a cost borne by uninformed traders.

How Adverse Selection Is Reduced

The fixes fall into two groups.

Screening is done by the uninformed party: underwriting, credit checks, vehicle inspections, due diligence on acquisitions, and contract menus designed so that different types choose different options.

Signaling is done by the informed party, who takes a costly action that low-quality types would find too expensive to copy. Spence’s 1973 work used education as a signal in the labor market. In product markets, a warranty works the same way: a seller who knows a car is sound can offer a guarantee cheaply, while a seller of a lemon cannot.

Third parties and rules help too:

• Reputation and reviews: dealer brands, user reviews and specialist assessments make quality visible before purchase.
• Certification and inspection: certified pre-owned programs, vehicle history reports and audited financial statements.
• Regulation: U.S. state lemon laws give buyers of defective vehicles remedies, and securities disclosure rules require issuers to publish material information.

Adverse Selection vs. Moral Hazard

Both come from asymmetric information, but they differ in timing and in what is hidden.

A banking example shows moral hazard: if employees expect regulators to rescue a failing bank, they may take excessive risks to earn bonuses. Nothing about them was hidden at hiring; the problem is their behavior once the protection exists. The two often coexist: an insurer can face both riskier applicants and policyholders who become less careful after buying cover.

Limitations of the Concept

• Markets rarely collapse fully. The lemons model assumes no remedies. Real used-car and insurance markets function because warranties, inspections, reputations and underwriting narrow the information gap.
• The size of the effect varies. Theory predicts that people who buy more insurance should be riskier, but that correlation is not always found.
• Selection can run the other way. If cautious people are both likelier to buy insurance and less likely to claim, the insured pool can be safer than average, sometimes called advantageous selection.
• Remedies have costs. Underwriting can price high-risk individuals out of coverage, while mandates force some low-risk people to buy cover they would not choose. Choosing between those outcomes is a policy judgment.
• It is easy to mislabel. Fraud, moral hazard and bad luck can produce the same losses. Calling every such loss “adverse selection” blurs which remedy would help.

FAQs

The Bottom Line

Insurance buyers, lenders, investors and anyone buying used goods face adverse selection whenever the other side knows more about quality or risk. The practical response is to narrow the information gap through underwriting, inspections, credit checks, warranties and disclosure, or to change who participates through group coverage and mandates. Each remedy carries costs, and losses from fraud, moral hazard or bad luck call for different fixes.