Chapter 16 — Information Asymmetry, Adverse Selection, and the Markets That Almost Don't Work

Part III has walked through four kinds of market failure: externalities (Chapter 11), public goods and commons (Chapter 12), inequality and distributional failures (Chapter 13), and healthcare's triple failure of information, moral hazard, and adverse selection (Chapter 14). Climate change (Chapter 15) tied the externality framework to the defining challenge of the century.

This chapter addresses a fifth and final kind of market failure: information asymmetry. When one side of a transaction knows something the other doesn't — about the quality of a product, the riskiness of a borrower, the health of an insurance applicant, the true condition of a used car — the market can fail in ways that are distinct from externalities or public goods. The failures have their own names (adverse selection, moral hazard), their own framework (the lemons model), their own solutions (signaling, screening, reputation systems, contract design), and their own most dramatic real-world manifestation (the 2008 financial crisis).

16.1 Adverse selection: the lemons problem

In 1970, George Akerlof — then a young economist at Berkeley, later a Nobel laureate (2001) — published a paper called "The Market for 'Lemons': Quality Uncertainty and the Market Mechanism." The paper was rejected three times by major journals before being published in the Quarterly Journal of Economics, and it became one of the most influential papers in 20th-century economics.

Akerlof's insight was deceptively simple. Consider the used-car market. Some used cars are good ("peaches") and some are bad ("lemons"). The seller of a used car knows whether their car is a peach or a lemon. The buyer does not — all they see is the car's appearance and a test drive, which don't fully reveal the car's true condition.

What happens?

Step 1 — Buyers assume the worst. Knowing that some cars are lemons, rational buyers offer a price that reflects the average quality of cars on the market — not the quality of any particular car.

Step 2 — Good-car sellers withdraw. If the market price reflects average quality, sellers of good cars (whose cars are worth more than average) find the price too low. They withdraw their cars from the market or sell them privately (where they can credibly communicate the car's quality). Only sellers of lemons — whose cars are worth less than the average price — are happy to sell at the market price.

Step 3 — Quality spirals down. As good-car sellers withdraw, the average quality of cars remaining in the market falls. Buyers, observing this, lower their offers. More good-car sellers withdraw. Quality falls further. In the extreme case, the market collapses entirely — only lemons are traded, and many potential trades that would benefit both parties never happen.

This is adverse selection: the "selection" of participants in the market is "adverse" (skewed toward low-quality or high-risk types) because information is asymmetric.

Adverse selection: a situation in which asymmetric information causes the composition of market participants to be skewed toward undesirable types, because the party with better information acts on it in ways that disadvantage the less-informed party.

Real-world examples of adverse selection

Used cars (Akerlof's original example). Used-car prices are lower than you'd expect given the average quality of used cars, because buyers can't tell good from bad and assume the worst. Sellers of good cars lose. Buyers of lemons lose. The market is less efficient than it would be with full information.

Health insurance (Chapter 14). People who expect to be sick are more likely to buy insurance. The pool skews sick. Premiums rise. Healthy people drop out. The spiral can destroy the market (the "death spiral").

Credit markets. Borrowers who are most likely to default are the most eager to borrow at any given interest rate. Banks, unable to distinguish good from bad borrowers, either charge high rates (which drives away good borrowers and worsens the pool) or ration credit (refuse to lend even to some creditworthy borrowers).

Labor markets. Employers can't perfectly observe a job applicant's ability. High-ability applicants can get jobs elsewhere; low-ability applicants are more likely to accept a below-market wage. An employer who offers below-market wages gets a pool of applicants skewed toward lower ability.

16.2 Market solutions to adverse selection

Markets have developed several mechanisms to overcome the lemons problem:

Signaling

Signaling: an action taken by the informed party to credibly communicate their type to the uninformed party.

Michael Spence (Nobel 2001) showed that education can function as a signal in the labor market. A college degree doesn't just teach you things (human capital); it also signals to employers that you are the kind of person who can complete a four-year degree (organized, persistent, intelligent). The signal is credible because it's costly — getting a degree requires years of time, effort, and money. A low-ability person would find it too costly to get the degree just for the signal, so the degree separates high-ability from low-ability applicants.

Other signals: - Warranties in the used-car market. A seller who offers a warranty is signaling that the car is good (a lemon seller wouldn't offer a warranty because they'd have to pay for the repairs). The warranty is credible because it's costly to the lemon seller. - Brand names. A company that invests heavily in building a brand reputation is signaling quality. Destroying the brand by selling a bad product would cost the company dearly, so the brand investment is a credible signal. - Professional certifications. An accountant with a CPA license, a doctor with board certification, a lawyer who passed the bar — each is signaling competence through a costly credential.

Screening

Screening: an action taken by the uninformed party to induce the informed party to reveal their type.

Insurance companies screen applicants by offering menus of policies: a high-deductible, low-premium plan and a low-deductible, high-premium plan. Healthy people (who expect low medical bills) rationally choose the high-deductible plan. Sick people (who expect high bills) choose the low-deductible plan. The insurer doesn't need to know who is healthy and who is sick; the menu design induces self-selection.

Other screening mechanisms: - Job interviews and probationary periods. The employer can't perfectly assess ability from a résumé. The probationary period lets the employer observe the worker's actual performance before making a permanent commitment. - Credit scoring. Banks use credit scores to screen borrowers. The score doesn't perfectly predict default risk, but it's much better than no information. - Trial periods for subscriptions. "Try free for 30 days" is a screening device: only people who genuinely value the product are likely to continue after the trial.

Reputation systems

Digital platforms have created powerful new mechanisms for reducing information asymmetry. eBay's seller ratings, Amazon's product reviews, Uber's driver and rider ratings, Yelp's restaurant reviews, Airbnb's host and guest reviews — all of these are reputation systems that aggregate past transaction information and make it available to future participants.

Reputation systems work because they reduce the asymmetry: a buyer on eBay doesn't know whether a particular seller is trustworthy, but they can see the seller's feedback score from hundreds of past transactions. The score is costly to fake (it requires actual transactions) and costly to lose (a bad rating reduces future business). So the reputation system functions as a credible signal of quality.

Limitations of reputation systems: they can be gamed (fake reviews), they can create "reputation lock-in" (new sellers can't compete because they have no reviews), and they don't work for one-shot transactions (where no reputation is built). But for repeat transactions on platforms, they have dramatically reduced the adverse-selection problem.

16.3 Moral hazard, revisited

We encountered moral hazard in Chapter 14 (healthcare). The concept is broader than healthcare:

Moral hazard: when one party to a contract changes their behavior after the contract is signed, in ways that are costly to the other party and that the other party cannot fully observe or prevent.

Healthcare moral hazard: once insured, patients use more healthcare than they would if they paid full price. The insurer can't fully observe whether each use is necessary.

Auto insurance moral hazard: once insured, drivers may be slightly less careful (because they know the insurance will cover an accident). The insurer can't observe how carefully each driver drives.

Employer-employee moral hazard: once hired, employees may exert less effort than they promised (shirking). The employer can't perfectly observe effort.

Bailout moral hazard: if banks believe the government will rescue them in a crisis ("too big to fail"), they take on more risk than they otherwise would. The government can't fully monitor the risk.

The distinction from adverse selection

Adverse selection is about who enters the market (the hidden-information problem before the transaction). Moral hazard is about how they behave once they're in the market (the hidden-action problem after the transaction).

• Adverse selection in insurance: the sickest people are most likely to buy insurance (before the contract).
• Moral hazard in insurance: once insured, people use more healthcare (after the contract).

Both are information problems. Both arise from asymmetry. But they have different solutions. Adverse selection is addressed by screening and signaling (before the transaction). Moral hazard is addressed by contract design — deductibles, co-pays, monitoring, performance incentives — that align the informed party's behavior with the uninformed party's interests (after the transaction).

16.4 The 2008 financial crisis as an information failure

Chapter 2's case study introduced the 2008 crisis as a story about model assumptions. Now we can tell a deeper version: the crisis was, at its core, an information failure — the most dramatic and consequential adverse selection and moral hazard event of the 21st century.

The information chain

Step 1 — Mortgage origination. Banks and mortgage companies originated home loans. In the 2003–2006 period, lending standards deteriorated sharply. Borrowers with low credit scores, no documentation of income, and little or no down payment received mortgages. The originator knew (or should have known) the risk of these loans, but it didn't matter because the originator planned to sell the loan to someone else rather than hold it.

This is a moral hazard at the origination level: the originator bears no long-term risk, so it has no incentive to lend carefully.

Step 2 — Securitization. The mortgages were bundled into mortgage-backed securities (MBS). The bundling was supposed to reduce risk through diversification (many mortgages from many regions — the regional-independence assumption from Chapter 2). The MBS were sold to investors: pension funds, hedge funds, insurance companies, foreign banks.

This created an information asymmetry: the originators and the securitizers knew more about the quality of the underlying mortgages than the investors who bought the MBS. The investors relied on credit ratings (from Moody's, S&P, Fitch) instead of doing their own due diligence. The credit rating agencies had their own information problems and conflicts of interest (they were paid by the issuers, not by the investors).

Step 3 — Credit default swaps (CDS). AIG and other institutions sold insurance (credit default swaps) on the mortgage-backed securities. A CDS promised to pay the MBS holder if the underlying mortgages defaulted. AIG sold hundreds of billions of dollars of CDS protection without setting aside adequate reserves to pay claims.

This was moral hazard at the insurance level: AIG collected premiums for insuring risk it didn't fully understand and didn't adequately reserve for. The moral hazard was compounded by the fact that AIG was "too big to fail" — and ultimately was rescued by the government.

Step 4 — The collapse. When house prices fell nationally (violating the regional-independence assumption), defaults rose across the country simultaneously. MBS values plummeted. CDS claims exploded. AIG couldn't pay. Lehman Brothers, which held massive positions in MBS, went bankrupt. The entire financial system seized up because no one knew how much risk anyone else was holding. The information asymmetry that had been papered over by credit ratings, diversification assumptions, and CDS insurance was suddenly, violently revealed.

What the information framework tells us

The 2008 crisis was: - Adverse selection at the MBS level: the worst mortgages were the ones most likely to be securitized and sold to uninformed investors - Moral hazard at the origination level: originators who didn't bear the risk didn't lend carefully - Moral hazard at the insurance level: AIG insured risk it didn't understand because it collected premiums without bearing the consequences - Information failure at the rating-agency level: the agencies that were supposed to provide information had conflicts of interest that rendered their ratings unreliable

The crisis was not primarily about greed (though greed was present), or about deregulation (though deregulation contributed), or about housing-market speculation (though speculation was a factor). It was primarily about information: the system generated, processed, and distributed information about risk in ways that were systematically wrong, and when the wrongness was revealed, the system collapsed.

Post-crisis reforms (Dodd-Frank, stress testing, capital requirements) are largely designed to address these information failures: requiring banks to hold more capital (reducing moral hazard), improving disclosure (reducing information asymmetry), and reducing the interconnectedness that made one institution's failure contagious.

16.5 What this chapter — and Part III — taught

Part III began with the assertion that markets sometimes fail. Six chapters later, you have seen the five main types of market failure:

1. Externalities (Chapter 11) — costs or benefits imposed on third parties
2. Public goods and commons (Chapter 12) — goods with non-rivalry or non-excludability
3. Distributional failures (Chapter 13) — markets that are efficient but inequitable
4. Healthcare's triple failure (Chapter 14) — information asymmetry + moral hazard + adverse selection in the most personal market
5. Climate change (Chapter 15) — the largest externality in history
6. Information failures (Chapter 16) — adverse selection, moral hazard, and the lemons problem

Each of these is a case where the supply-and-demand model from Chapter 5 does not produce the welfare-maximizing outcome. Each requires some intervention — government regulation, taxation, information provision, contract design, community governance — to move the market closer to efficiency.

The next part — Part IV — goes inside the firm. We leave market failures behind (temporarily) and ask: how do firms make decisions? What do their cost structures look like? How does competition (or the absence of it) shape prices and output? The tools from Part III will come back when we look at monopoly power and labor markets.

Key terms recap: asymmetric information — one party knows more than the other adverse selection — skewed participation because of hidden information (the lemons problem) moral hazard — changed behavior because of hidden action (after the contract) lemons problem (Akerlof) — quality spirals down when buyers can't observe quality signaling (Spence) — costly credible action by the informed party (warranties, degrees, brands) screening — menu design by the uninformed party to induce self-selection reputation system — aggregated past-transaction information (eBay ratings, Uber stars, Yelp) principal-agent problem — the agent acts on behalf of the principal but may have misaligned incentives CDS (credit default swap) — insurance on financial instruments

Themes touched: Markets power+imperfect (information failures are the fifth and final type of market failure in Part III), Disagreement (about the relative importance of information vs. other factors in the 2008 crisis), Behavioral (reputation systems exploit behavioral shortcuts), Affects daily life (used cars, insurance, credit, online platforms, the 2008 crisis).

PART III COMPLETE. You now understand the five main types of market failure: externalities, public goods/commons, distributional failures, healthcare, climate, and information asymmetry. Part IV — Firm Behavior and Market Structure — begins with Chapter 17.