Adverse selection occurs when hidden pre-contract information changes who trades, borrows, or buys insurance, worsening the pool or terms offered.
Adverse selection occurs when one side of a potential transaction has private information about its risk, quality, or likely payoff, and that hidden information changes who accepts the offered contract. The participants most willing to transact may be systematically riskier or lower quality than the less-informed party assumed, worsening the pool of insureds, borrowers, securities, or trades.
Adverse selection is a pre-contract selection problem. It does not require fraud or deliberate exploitation. A high-risk applicant can answer every permitted question truthfully and still know more about future risk than an insurer or lender can observe.
The mechanism needs three ingredients:
The third condition is essential. Information can be asymmetric without creating adverse selection if the hidden characteristic does not affect participation, expected loss, or value.
flowchart TD
A["Hidden risk or quality"] --> B["Pooled price or terms"]
B --> C["Riskier types accept more"]
C --> D["Average pool risk rises"]
D --> E["Price or terms tighten"]
E --> F["Lower-risk types leave"]
F --> D
The feedback loop can weaken a market, but collapse is not inevitable. Better information, differentiated contracts, subsidies, mandates, competition, reputation, or cross-subsidization can alter the outcome.
Consider a deliberately simplified insurance pool with two applicant types. Expected loss means the probability-weighted claim amount, not a quoted premium.
| Applicant type | Number initially | Expected annual loss per person | Total expected loss |
|---|---|---|---|
| Lower risk | 100 | $200 | $20,000 |
| Higher risk | 100 | $1,400 | $140,000 |
| Initial pool | 200 | $160,000 |
If the insurer cannot distinguish the types, the initial pooled expected loss is:
An actual premium would also need expenses, capital, taxes, profit or contingency margins, contract terms, and applicable regulatory treatment. The $800 is only an expected-loss average.
Suppose the pooled offer is unattractive to many lower-risk applicants because they know their own risk is below the pool average. Eighty lower-risk applicants decline, while all 100 higher-risk applicants remain:
The average expected loss rises from $800 to $1,200, a 50% increase, even though neither type’s underlying expected loss changed. If the offer is repriced around the new pool, still more lower-risk applicants may leave.
This example demonstrates selection, not a forecast of insurance behavior. Real applicants have many risk levels, insurers observe some characteristics, coverage affects behavior, regulations differ, and willingness to buy depends on risk aversion, income, alternatives, and contract design.
An applicant generally knows more than the insurer about habits, symptoms, intended use, maintenance, or other risk factors that may not be observable or permitted in underwriting. If one price is offered to a broad pool, people expecting larger benefits may be more willing to buy more coverage.
Insurers may respond with underwriting, waiting periods where permitted, deductibles, limits, exclusions, contract menus, group pooling, or reinsurance. These responses can improve pricing but can also reduce access, add administration, or shift risk back to policyholders. The National Association of Insurance Commissioners defines adverse selection in insurance as greater demand for coverage among people with above-average loss probability.
Borrowers often know more than lenders about project quality, repayment priorities, future borrowing, or income stability. A lender that raises rates to compensate for uncertainty may unintentionally discourage safer borrowers whose lower-risk projects cannot support the higher payment, while applicants pursuing riskier, higher-upside projects remain willing to borrow.
This helps explain why price alone may not clear a credit market. Lenders may instead limit amounts, require collateral or covenants, verify information, monitor accounts, or apply Credit Rationing. Federal Reserve research on consumer loan markets illustrates why contract choice, interest rates, collateral, adverse selection, and moral hazard must be analyzed together rather than inferred from one outcome.
A dealer or market maker posts prices without knowing whether the next order comes from an uninformed liquidity trader or someone with better information about value or pending order flow. Informed traders are more likely to buy when the offer is stale and too low or sell when the bid is stale and too high. The liquidity provider is then selected into trades that are more likely to lose money after prices adjust.
Possible responses include wider bid-ask spreads, smaller quoted size, faster quote updates, inventory controls, or reduced participation during uncertainty. The spread contributes to Transaction Cost, but adverse-selection cost is only one component; order processing, inventory risk, tick size, competition, volatility, and market rules also matter.
An originator or current holder may know more than a buyer about loan quality, underwriting exceptions, servicing, documentation, or collateral. If buyers cannot separate good pools from weak pools, they may demand a discount based on average or stressed quality. Sellers of stronger assets may refuse that price, leaving weaker assets more likely to be offered.
Federal Reserve research on private-label securitization and asymmetric information models how information gaps can produce discounts or no trade for complex securities. The result is model-dependent; illiquidity, funding pressure, uncertainty, legal risk, and capital constraints can also reduce trading.
The seller of a used vehicle, machine, receivable, private company, or other hard-to-evaluate asset may know more about quality than a prospective buyer. If buyers offer only an average-quality price, owners of high-quality assets may withdraw. This is the Market for Lemons mechanism.
Due diligence, representations and warranties, escrow, holdbacks, seller financing, third-party inspection, audits, and reputation can help. None guarantees quality, and each changes cost or risk allocation.
| Concept | Timing | Core problem | Example |
|---|---|---|---|
| Asymmetric information | Before or after contracting | One party knows more than another | Borrower knows more about project risk than lender |
| Adverse selection | Before or at contract choice | Hidden types participate at different rates | Higher-risk applicants accept pooled insurance more often |
| Moral hazard | After contracting | Protection or financing changes action or effort | Insured party reduces precautions after coverage begins |
| Principal-agent problem | During delegated decision-making | Agent’s incentives or information differ from principal’s | Manager chooses a project shareholders would reject |
| Lemons problem | Before an asset sale | Hidden quality drives strong sellers away | Buyers discount all used assets toward average quality |
| Risk-based pricing | At underwriting or quotation | Observable differences receive different terms | Documented higher expected loss leads to a higher price |
The timing distinction is useful but not always sufficient for empirical work. A contract can contain both selection and incentive effects: a person may choose coverage because of private risk information, and the coverage may then change behavior.
The less-informed party screens. A lender verifies income and collateral; an insurer reviews permitted application information; a buyer inspects an asset; an investor performs due diligence. Screening can reduce uncertainty but is imperfect, costly, and subject to legal and data-quality constraints.
The better-informed party signals. A borrower contributes more equity, a seller offers a warranty, an issuer accepts an audit, or a firm builds a reputation through repeated performance. A useful signal must be sufficiently credible and harder for weak types to imitate; a cheap unsupported claim is not a reliable signal.
An uninformed party may offer choices designed to encourage self-selection. For example, insurance contracts may trade a lower premium for a higher deductible, or a lender may offer different combinations of rate, collateral, maturity, and covenants. The 2001 Nobel Prize overview of markets with asymmetric information explains how Akerlof’s adverse selection, Spence’s signaling, and Stiglitz’s screening address different parts of the information problem.
Repeated transactions can produce performance history and make reputation valuable. Banks, rating firms, auditors, exchanges, insurers, and other intermediaries may collect or certify information. They can reduce information gaps, but they introduce fees, model risk, conflicts of interest, and their own monitoring problems.
A useful review separates the theory from the evidence.
Suppose a lender observes that borrowers choosing low-collateral loans default more often. That pattern is consistent with adverse selection if borrowers privately know their risk and riskier borrowers prefer to pledge less collateral. It is not conclusive because:
A defensible analysis would preserve application-time data, compare otherwise similar offers, document underwriting variables, distinguish selection from post-origination behavior, and report uncertainty. It should not infer dishonesty from a risk correlation.
The Federal Reserve’s discussion of financial instability and information flows describes adverse selection and moral hazard as distinct ways information problems can disrupt credit and price discovery. The distinction is analytically useful, but real financial contracts often contain both.
These sources provide theory, regulatory context, or empirical research. Research findings depend on the sample, identification strategy, contract design, and institutional setting and should not be generalized automatically.
This article provides general economic and financial education. It is not an underwriting decision, insurance recommendation, credit decision, trading strategy, legal opinion, or assessment of any person or transaction.