Adverse Selection: A Primer

Information is the basis for our economic and financial decisions. As buyers, we collect information about products before entering into a transaction. As investors, the same goes for information about firms seeking our funds. This is information that sellers and fund-seeking firms typically have. But, when it is too difficult or too costly to collect information, markets function poorly or not at all.

This form of asymmetric information―where two parties to a potential transaction have unequal knowledge―is a particularly serious hindrance to the operation of financial markets. If, for some reason, conditions suddenly make the information asymmetry worse, the consequences can be catastrophic. In a recent post, we described how in August 2007, a sequence of events led financial intermediaries to suddenly question the quality of some securities that many of their counterparties already owned. Not being able to tell safe from unsafe, investors and institutions withdrew from lending. As credit evaporated, many potential transactions stopped taking place all at once.

Economists use the term adverse selection to describe the problem of distinguishing a good feature from a bad feature when one party to a transaction has more information than the other party. The degree of adverse selection depends on how costly it is for the uninformed actor to observe the hidden attributes of a product or counterparty. When key characteristics are sufficiently expensive to discern, adverse selection can make an otherwise healthy market disappear.

The term “adverse selection” comes from the fact that, when the hidden attributes are costly to observe, the quality of the products on offer or of the potential parties to a transaction deteriorates; that is, the pool becomes adverse relative to the full universe of goods (or counterparties) available.

In this primer, we examine three examples of adverse selection: (1) used cars; (2) health insurance; and (3) private finance. We use these examples to highlight mechanisms for addressing the problem.

Used cars. George A. Akerlof, A. Michael Spence, and Joseph E. Stiglitz received the 2001 Nobel Prize in Economics “for their analyses of markets with asymmetric information.” Akerlof’s contribution came first, in a 1970 paper titled “The Market for “Lemons”. He explained why the market for used cars—some of which may be mechanically deficient lemons—doesn’t function well.

Akerlof’s logic is surprisingly simple. Suppose the used-car market has only two cars for sale. One is immaculate (call it a peach); the other (a lemon) contains hidden damage because of the driver’s abusive handling. While the owners know the condition of their own cars, used-car shoppers do not.

Suppose that potential buyers are willing to pay $20,000 for a well-maintained car, but only $10,000 for a lemon. The owner of the peach won’t part with it for less than $18,000. The owner of the lemon would take $8,000 for it. If buyers can’t tell the condition of the two cars, without any additional information they might pay only the average price of $15,000. (A risk-averse buyer wouldn’t even pay that much.) That means the peach disappears from the market, leaving only the lemon for sale. The point is that buyers’ inability to observe the hidden attributes of the vehicles for sale undermines the entire market for used cars. Compared to the complete information environment, the exit of high-quality objects or sellers from the market results in a substantial welfare loss. This form of market failure is a classic result of adverse selection.

The solution to the adverse selection problem in the used-car market is to reduce the cost of detecting the car’s hidden attributes, helping buyers separate the peaches from the lemons. Because this is such an important market, people have developed a range of technologies and practices to improve its function. CARFAX, for example, provides potential buyers the detailed history, including reported accidents and airbag deployments, of a specific used vehicle. For a fee, a mechanic will check out a used car for a potential buyer. The sellers of peaches also have developed means to signal the quality of their product (Spence’s analysis of signaling earned him his Nobel.). Many car manufacturers offer warranties on the used cars they have certified. And, car dealers try to maintain their reputations by refusing to pass off a clunker as a well-maintained car.

These mechanisms for addressing the information asymmetry allow both good and bad used cars to sell at prices closer to their true value. So long as there exists a technology that lets buyers determine, at a reasonable cost, the hidden attributes of used cars for sale, the market works.

Health insurance. In its landmark 2015 decision upholding the Affordable Care Act (ACA, aka Obamacare), the U.S. Supreme Court explained why health insurance that is always available to everyone at a common price has to be mandatory: “Why buy insurance when you are healthy, if you can buy the same coverage at the same price when you are ill?”

The justices went on to describe how adverse selection (they used the term repeatedly!) can trigger a “death spiral” for health insurance programs. To see what they meant, suppose that an insurer initially bases the policy premium on the nation’s average cost of health care per person. If participation is voluntary, healthier people—typically young adults—will be less willing to enroll than those who anticipate greater-than-usual health care costs. Consequently, insurance outlays will exceed the norm, leading to losses for the insurance company. If the insurer then raises the premium, those relatively healthy individuals who were still in the pool will drop their insurance. The higher the premium, the smaller the enrollment, and the worse the health of those still insured. Eventually, the program becomes unviable.

Two mechanisms exist to address adverse selection in the health insurance market: allow price discrimination based on health status, or make insurance mandatory. The first option, allowing insurers to set premiums according to the health of the applicant—the way they do for life insurance—reduces the incentive for healthy individuals to drop out. But concerns about fairness often lead governments to forbid this kind of price discrimination. Since the point of insurance is to help us when we are unfortunate, discrimination based on how healthy we are seems unfair. Consequently, in most advanced economies, the solution to the adverse selection problem in health insurance is the second option—require participation. (Where behavior, rather than misfortune, influences health—like smoking or alcohol consumption—jurisdictions may still authorize discriminatory pricing.)

How was adverse selection managed in the United States before passage of the ACA? For the most part, employers who provided health insurance promised insurers broad participation, limiting the risk that healthy workers would drop out. However, the U.S. approach had major shortcomings. For example, before the ACA, an estimated 47 million U.S. residents—about 18 percent of the population—went uninsured. In addition, persons with preexisting conditions could not obtain new coverage if they changed employers, so many stayed in their jobs (even when that was inefficient) just to keep their health insurance.

Private finance. As our initial discussion about the financial crisis suggests, what is true for used cars and health insurance, is true for finance in general. For example, potential users of funds know more about the projects they wish to finance than prospective suppliers of those funds do. This information asymmetry can drive good equity and debt out of the financial market―another version of the lemons problem.

Consider the case of stocks. Suppose there are just two firms, one with good and one with bad prospects. Even in the absence of risk aversion, investors who are unable to tell the difference between the two, will only be willing to pay a price based on their average quality. However, knowing that it will be undervalued, the owners of the good company won’t issue stock in the first place. That leaves only the firm with bad prospects in the market.

The same thing happens with bonds. Risk requires com­pensation: the higher the risk, the greater the risk premium. Not able to observe whether a potential borrower is a good or a bad credit risk, a potential lender will demand a risk premium based on the average risk. Borrowers who know that they are low risk withdraw from the market to avoid paying the high cost, leaving only the bad credit risks applying for high interest rate loans.

As with used cars, we have found ways to reduce the impact of adverse selection in finance. These include disclosure, the use of collateral to secure loans, and requirements that funds-seekers have an equity stake (sometimes called skin in the game). The first―the CARFAX solution applied to finance―has its limits. As we saw in the financial crisis, if it suddenly becomes costly and time-consuming to uncover the truth about the value of a security or the creditworthiness of a counterparty, markets and credit can disappear. (By certifying the state of large banks’ balance sheets in May 2009, the Federal Reserve’s stress tests were a critical part of the remedy to some of the adverse selection problems at the root of the crisis.)

The second solution―the one that is common in real estate, business lending and municipal bonds markets―is to use collateral. Home and commercial mortgages are secured by buildings; while company loans (and even corporate bonds) can be secured by a firm’s physical assets, like factories or machines; and municipal bonds are often linked to specific revenue sources, like tolls from bridges or roads.

Next, there is the equity stake of the funds-seeker. Take the simple case of a home mortgage. For a lender, the risk is that the value of the house will fall. From their perspective, the homeowner’s equity serves as a cushion against this risk. The greater the owner’s stake, the less the lender needs to worry about the value of the collateral.

In practice, when people need funds, most turn to banks, because banks have developed technologies to overcome information asymmetries. Banks address the adverse selection problem by screening loan applicants. They expend what are often significant resources to collect enormous amounts of information about potential borrowers in order to estimate the likelihood that a loan will be repaid. This process allows banks to charge interest rates that differ across borrowers: the better someone’s personal credit score, for example, the lower the interest rate on a loan. That is, banks have systems in place that allow them to measure potential borrowers’ hidden attributes, and price their products accordingly.

Even with all of these mechanisms, adverse selection contributes significantly to the costs of obtaining finance for households and businesses.

Future Challenges. Since the problem arises from information asymmetries, it is natural to ask how advances in information technology will affect adverse selection. As information becomes cheaper to obtain and easier to share, will adverse selection disappear? Not everywhere. In fact, in the case of health insurance, the science of genomics is going to make matters worse, not better. Advances in DNA testing are making it more feasible to anticipate an individual’s future illnesses, giving new meaning to the term “preexisting conditions.” Governments could try to address this by forbidding DNA screening by health insurers. More likely, this advance of knowledge provides another reason why health insurance will be broadly available at a common price only if it is mandatory.

The case of finance is more complicated. We see two questions: Will better information technology make it cheaper and easier for firms and households to raise funding? Will the predominance of banks in consumer lending give way to a more market-based mechanism? On the former, the answer could very well be yes. The ability of individuals and firms to share large amounts of private information (possibly using blockchain technology) cheaply could lower the cost of screening and improve pricing. On the second, expressly because of adverse selection, we are skeptical about the prospects for platforms like peer-to-peer lenders. Instead, banks and finance companies are likely to remain the primary lender to consumers.

The bottom line: while changing technology will alter the challenges arising from adverse selection, they will not go away. Managing those challenges will be key to the success of any market (financial or otherwise) where it is costly to overcome information asymmetries.

Acknowledgement: The authors are grateful to their friend and colleague, Professor Maher Said, for thoughtful discussions regarding adverse selection.