Concept/Economics/No. 0056

Asymmetric Information

You test-drive a used car for twenty minutes; the seller has driven it for five years.

Also called Information Asymmetry

a concept: name it

Evidence: Well established
Read: 7 min
Links: 11 connections
Useful when: Metrics and incentives · Money and investing · Negotiating · Organizations and bureaucracy · Complex systems and policy

01You've seen this when…

in life
An apartment looks quiet during your afternoon viewing. The landlord knows the restaurant downstairs unloads deliveries at 5 a.m.
at work
A software vendor promises an easy migration. Its engineers know your oldest records will need manual cleanup; your team has only seen the demo.
out in the world
A city pays a contractor to maintain streetlights. The contractor knows which repairs actually happen; officials see invoices and scattered complaints.

02The idea

The apartment viewing gives you a snapshot. The landlord has a history. Both of you are negotiating over the same place with different knowledge.

Asymmetric information means one participant knows relevant facts that another lacks. The difference matters when those facts affect what someone should pay, agree to, or expect. Unequal knowledge can persist even when everyone tells the truth. A seller can answer every question truthfully while a buyer never thinks to ask the important one.

This is different from uncertainty everyone shares. Neither a farmer nor a grocer may know next month’s weather. But the farmer may know that today’s shipment has already spent several hours unrefrigerated. The first problem is shared uncertainty; the second is an information gap between participants.

Two common consequences deserve separate names. Adverse selection concerns hidden characteristics: before agreeing, you cannot reliably distinguish the better risks, products, or partners from the worse ones. Moral hazard concerns incentives and actions: after agreeing, someone can change their behavior while another party bears some of the consequences.

Neither consequence is inevitable. The central question is what the less-informed party can learn, verify, or protect against—and at what cost.

03Why it matters

Prices respond to both quality and uncertainty. A buyer who cannot verify a car’s condition may offer less than a good car is worth. The seller knows the discount is unfair to this particular car; the buyer knows it is sensible given the information available.
Worthwhile deals can disappear. If good sellers cannot demonstrate their quality, they may refuse the price buyers offer. Their departure makes the remaining market worse, giving buyers another reason to lower their offers. An information problem can shrink trade itself, with effects beyond the distribution of its proceeds.
Agreements change incentives. Paying someone by the hour, by the completed job, or by the result creates different temptations when effort and quality are hard to observe. This connects information gaps to the principal-agent problem: the person doing the work may know much more than the person paying for it.
Verification becomes part of the service. Inspections, audits, references, warranties, and reputation systems help people trade despite unequal knowledge. They also create transaction costs. A bargain requiring weeks of checking may no longer be a bargain.

These responses show up in market data. In a 2011 study of eBay Motors, economist Gregory Lewis examined sellers’ voluntary disclosure through listing text and photographs and how that information related to auction prices. The information presented in the listing was part of the economics of the sale.

04A worked example

Consider a deliberately simplified used-car market, adapted from Akerlof’s classic thought experiment. These numbers are illustrative, not observations from a study.

Half the cars are sound; half are worn. Buyers would pay $12,000 for a sound car and $6,000 for a worn one. Owners of sound cars will not sell below $10,000. Owners of worn cars will accept $4,000. Sellers know their own car’s condition, but buyers cannot distinguish the two types.

What it looks like A reasonable compromise is $9,000, the average of the two values to buyers. Assuming buyers are risk-neutral and believe each type is equally likely, that offer reflects the information they have. A buyer can make it in good faith and with sound reasoning.

What’s actually going on The $9,000 guess is unstable. Owners of sound cars refuse it, leaving worn cars for sale. Once buyers recognize that change, their valuation falls to $6,000, the value of the remaining cars to them. The sound-car trades disappear even though buyers value those cars more than their owners do. There was room for a mutually beneficial deal; identifying it required reliable information that buyers lacked.

What would have helped A credible inspection could distinguish sound cars from worn ones. A financially dependable seller might offer a warranty that is more expensive to honor on a worn car. These mechanisms need to cost less than the value of the trade they rescue. A seller’s unsupported assurance would leave buyers facing the same uncertainty: both types could give it equally cheaply.

The lesson is that the mix of things offered for sale can change in response to buyers’ uncertainty. A used-car market can experience that shift and still keep operating.

05Where people trip up

Greater knowledge is compatible with honesty. Expertise, experience, and specialization naturally produce unequal knowledge. Start by naming the missing fact and why it matters. Claims that the informed party is hiding something need evidence beyond the information gap. The party with the advantage can also be the buyer: an insurance applicant may know more about their own risk than the insurer does.
Cheap claims are weak tests of quality. A polished presentation is easy for both strong and weak suppliers to produce. Ask what evidence would be harder for the weaker supplier to provide. Signaling is the informed party’s attempt to convey what they know; costly signaling explains why the difficulty or expense of a signal can sometimes make it credible. Expense alone proves nothing.
The less-informed side can design the test. Take the initiative by requesting an independent inspection, a paid trial, or a choice between contracts. This is screening: structuring a decision so relevant differences become easier to see. Choose a test that targets the uncertainty. A test that merely generates more paperwork leaves the information gap in place.
A promise needs someone able to honor it. A warranty from a seller about to disappear is poor protection. Check the provider’s resources, exclusions, and enforcement arrangements. Guarantees may align incentives while quality remains partly hidden, but only if the promised consequences take effect.
Longer contracts still leave gaps. With incomplete contracts, some important actions or circumstances remain unspecified or difficult to prove. Use observable milestones, sampling, and clear remedies where possible. Adding clauses can leave the same facts beyond verification.

06When it isn’t an information problem

Two people can know the same facts and still disagree about price. Their valuations may differ, and either unequal bargaining power or a lack of good alternatives can shape the price. An information gap is only one possible explanation for an unfavorable deal.

Risk can exist even when knowledge is shared. If both parties have the same evidence about an uncertain outcome, the issue is risk versus uncertainty; unequal knowledge would need to be established separately.

Finally, trade can survive unequal knowledge. Repeat business and enforceable guarantees can support transactions, while professional standards and good market design help them work even when participants remain unequally informed. The practical target is a dependable decision that can be made even with some information gaps.

07Roots

At Berkeley, George Akerlof turned the familiar problem of buying a used car into a challenge to standard economics. His 1970 paper divided cars into good ones and lemons. Buyers might overpay. More strikingly, their inability to tell the cars apart could drive good cars out of the market altogether.

Economists already knew that people possessed different information. Akerlof showed how putting that difference at the center of a model could explain missing markets and failed trades. The price both reflected the available cars and changed which cars their owners would offer.

Michael Spence’s 1973 work followed the better-informed participant’s response. A job applicant could use education to signal ability, provided obtaining that education imposed different costs on different types of applicant. Joseph Stiglitz and collaborators explored the other side: how the less-informed participant could construct choices that reveal information. Michael Rothschild and Stiglitz’s 1976 insurance paper examined contracts that people with different risks would choose differently.

Together, these approaches made information part of the machinery of markets: what a participant knows and can demonstrate shapes what others infer from a choice. In 2001, Akerlof shared the Nobel Memorial Prize in Economic Sciences with Spence and Stiglitz for this work.

08How solid is this?

Unequal information and its effects on market behavior are supported by extensive theoretical and empirical research. The theory predicts complete market unraveling only under specific conditions; verification, reputation, and contracts often limit the damage.

09Connections

Often confused with Risk vs. Uncertainty
Countered byScreening, Signaling, Costly Signaling, Market Design, BATNA
Can lead to Moral Hazard
Includes Adverse Selection, Principal-Agent Problem
See alsoTransaction Costs, Incomplete Contracts

+ 1 more in the list

10Origin and sources

George Akerlof’s 1970 used-car model established a foundational account. Michael Spence developed signaling in 1973; Joseph Stiglitz and collaborators developed screening models during the 1970s.

Suggest an edit· Updated 2026-10-02