Fundamentals

Adverse Selection: What It Is and How It Arises

By Daniel Sardá · Published on October 14, 2026

6 min read1,220 words

Adverse selection arises when one party knows more about quality or risk before an agreement is made. That difference can change prices, participants, and outcomes.

What happens when a buyer cannot tell an excellent product from a defective one? Or when an insurer knows less than its customer about the likelihood of a loss? In both cases, setting a price is difficult. Charging as though everyone were alike can drive away precisely those with lower risk or higher quality.

This mechanism is known as adverse selection. It arises before an agreement is made, when one party has relevant private information about its quality or risk and the other cannot fully observe it. The information gap changes the terms of exchange and, sometimes, who chooses to take part.

It does not necessarily involve fraud, bad faith, or irresponsible conduct. It can occur even when no one lies. What matters is that an important characteristic cannot be observed and that this lack of information affects the price or the contract.

Key idea: Adverse selection is not conduct that occurs after a contract is made. It is an information problem that affects who enters into an agreement and on what terms.

How adverse selection works

The mechanism can be understood as a chain of four steps.

1. There are different types of goods or people: for example, high- and low-quality cars, or policyholders with different levels of risk.
2. One party knows its own type better than the other.
3. Because it cannot distinguish those types precisely, the less informed party offers a price or terms based on average quality or risk.
4. That offer is less attractive to some favorable types, which reduce their participation or leave the exchange. The composition of the remaining group worsens, and the initial calculation is no longer sustainable.

An average price conveys an estimate based on incomplete information and changes incentives. The process can reduce exchanges or lower average quality, as George A. Akerlof showed in The Market for “Lemons”.

The outcome, however, is not predetermined. The presence of asymmetric information does not mean that good products must always disappear or that every market will collapse. The result depends on the distribution of quality or risk, the cost of obtaining information, and the institutions available to make differences credible.

The used-car example

Imagine that a car owner knows how the vehicle has been treated, which breakdowns it has had, and whether it needs expensive repairs. A prospective buyer can inspect it only to a limited extent. Even if some cars are excellent, the buyer will not want to pay the price of an excellent one if there is a risk of receiving a defective vehicle.

If the buyer offers a price based on average quality, owners of the best cars may find the offer too low and decide not to sell. As those cars leave the market, the average quality of the cars still available declines. Buyers then revise their expectations and offer even less.

The selection is “adverse” because uncertainty tends to leave a less favorable mix for the party without the information. It does not mean that every seller deceives or that every used car is bad.

What happens in insurance

In insurance, a person may know more than the company about certain aspects of their risk. Someone who believes they are highly likely to suffer a loss will usually value coverage more than someone who considers themselves low risk. If the insurer cannot distinguish between them and charges everyone the same premium, the policy may be expensive for lower-risk customers and attractive to higher-risk ones.

If some lower-risk customers leave, the expected cost of the group rises. The insurer may respond by raising the premium, which can in turn prompt further departures. The Insurance Information Institute describes adverse selection in this context as the greater tendency of high risks to seek or maintain coverage.

Here, too, automatic conclusions should be avoided. It is not enough for people to have different risks: there must be a relevant difference between the risk they know and the risk the insurer can observe. In addition, deductibles, questionnaires, records, and different coverage options can change the result.

Key idea: Different levels of risk alone do not create adverse selection. The problem arises when unequal information prevents those risks from being distinguished and changes the composition of those who accept an agreement.

Adverse selection and moral hazard are not the same

The simplest distinction is temporal:

• Adverse selection: private information exists before the contract and affects who enters it.
• Moral hazard: the contract can change behavior after it is made because some of its costs fall on someone else.

In insurance, knowing more about a pre-existing condition belongs to adverse selection. Taking fewer precautions after obtaining coverage would be a possible case of moral hazard. The first question is, “Whom am I contracting with?” The second is, “What will they do afterward?”

Neither concept is, by definition, fraud. They are problems of information and incentives that help analyze contracts, though in some cases they may coexist with deliberate deception.

Signaling and screening: two different responses

People do not usually remain passive in the face of uncertainty. They develop mechanisms to communicate or uncover information.

Signaling begins with the party that knows more. A broad warranty, for example, can communicate confidence in quality because it would be more costly for a seller of defective goods. In his analysis of signaling, Michael Spence showed that a signal distinguishes types only when its costs or incentives make it credible.

Screening begins with the party that knows less. It consists of designing tests, requirements, or menus of options that help separate types. An insurer may offer different combinations of premiums and deductibles so that customers with different profiles reveal information through their choices. Joseph Stiglitz studied this process as a way of identifying qualities.

Key idea: The party with the information sends a signal; the party without it designs a screening mechanism to obtain or infer it.

Institutions that make exchange possible

Reputation, certification, independent inspections, warranties, and specialized intermediaries reduce the information gap. So does the design of voluntary contracts: provisions on returns, maintenance, deductibles, or responsibilities allocate risks and make some commitments observable.

These decentralized solutions reflect a process of learning. Prices convey signals, competition rewards certain arrangements, and parties adapt their contracts. From a classical liberal perspective, adverse selection illustrates a problem of dispersed knowledge; it is not, by itself, proof against voluntary exchange or an automatic justification for a single regulatory response.

Private solutions should not be idealized, either. Inspections are costly, a warranty must be enforceable, and reputation takes time to build. Rules that protect property and allow contractual commitments to be enforced are part of the setting that makes such mechanisms useful.

The relevant question is not whether perfect information can be achieved, but which arrangements reduce uncertainty at a reasonable cost. Where parties can experiment with warranties, reputation, verification, and contractual design, adverse selection can be contained without disappearing altogether.

Understanding adverse selection requires tracing three elements: who knows what, how that difference changes incentives, and which institutions make information credible. With this guide, the concept stops being an abstract label and becomes a tool for explaining why some exchanges become difficult, how they adapt, and why their failure is not inevitable.