Adverse Selection

Adverse selection is the tendency for the people who expect to use coverage most to be the ones who buy it, at the highest limits, in a market where the seller cannot fully observe how risky each buyer is. The National Association of Insurance Commissioners, the standard-setting body for state insurance regulators, defines it in its glossary as "the social phenomenon whereby persons with a higher than average probability of loss seek greater insurance coverage than those with less risk." Note the framing: NAIC calls it a social phenomenon rather than a behavior, because nobody involved has to be acting in bad faith for it to happen. A person who knows their family history and buys more life insurance because of it is behaving sensibly.

The problem is not confined to insurance. In economics it names any market in which one side knows more about what is being traded than the other, and the better-informed side's choices systematically shift the average quality of what gets sold. The general treatment economists cite is George Akerlof's 1970 paper in the Quarterly Journal of Economics on quality uncertainty in the used car market, which gave the field the shorthand it still uses.

Why a single price cannot hold. Suppose an insurer must charge everyone in a group the same premium and cannot ask any questions. The premium has to cover the group's average expected cost. To a buyer who privately expects to cost far more than average, that price is a bargain and they buy generously. To a buyer who privately expects to cost far less, the same price looks like poor value and they buy little or nothing. Their departure raises the average cost of whoever remains, which raises the price, which pushes out the next tier of relatively low-risk buyers. Economists call the extreme version of this a death spiral. It is worth being precise about what causes it: not that some people are riskier than others, which is always true, but that they know it and the insurer does not.

The mechanisms built against it are the boring parts of an insurance contract, and each targets a different moment. Asking health, driving or property questions before issuing a policy is the direct answer, and published material on underwriting covers how the answers turn into a rate class. Waiting and elimination periods delay the start of benefits so that someone cannot buy coverage in the week they expect to need it. Enrollment windows restrict when anyone may join at all; that application is developed at length on the open-enrollment page and is not repeated here. Contract provisions letting an insurer investigate and rescind for a misstatement during an early window address the buyer who answered the questions untruthfully. And where the law or the product forbids all of this, the price carries the burden instead: a policy issued with few or no health questions costs more per dollar of coverage precisely because the insurer cannot separate the applicants.

The trade-off runs in both directions, which is why this is a policy question and not just an actuarial one. Every mechanism above works by excluding or surcharging people according to characteristics they mostly did not choose. A market with unrestricted underwriting prices risk accurately and leaves the sickest people unable to buy at any price; a market with none of it prices everyone the same and risks unraveling. Real systems sit in between and use substitutes: guaranteed issue paired with a limited enrollment window, or with a subsidy that keeps low-risk buyers in the pool even at a blended price. Which combination a country or a state chooses is a distributional decision wearing technical clothes.

In investment markets the same structure appears without the word. A seller who knows more about an asset than the buyer sells when the asset is worth less than the price and holds when it is worth more, so the buyer's average purchase is worse than the average asset. Market makers widen the spread they quote to cover the losses they expect to informed counterparties, and issuers of complex or illiquid securities face higher costs of capital for the same reason. The responses look different from waiting periods but do the same job: disclosure rules, audited financials, warranties and reputational intermediaries all exist to shrink the gap in information rather than to price around it.

Start with a pool the insurer cannot sort. Suppose 1,000 people want coverage. Five hundred of them are low-risk and will cost $500 a year on average; five hundred are high-risk and will cost $3,000 a year on average. Each person knows which group they are in and the insurer does not, and the law requires one price for everybody.

The insurer must charge the average. Total expected cost is 500 times $500, which is $250,000, plus 500 times $3,000, which is $1,500,000, for $1,750,000 across 1,000 people. That is a premium of $1,750 each. To the high-risk half, paying $1,750 for $3,000 of expected cost is an obvious purchase. To the low-risk half, paying $1,750 for $500 of expected cost is a poor one, and some of them decline.

Now suppose 300 of the 500 low-risk people drop out. The remaining pool is 200 low-risk and 500 high-risk, 700 people in all, with expected costs of 200 times $500 plus 500 times $3,000, which is $100,000 plus $1,500,000, or $1,600,000. Divided across 700 people, the premium has to rise to about $2,286. That new, higher price is worse value still for the 200 low-risk people who stayed, so more of them leave, and the arithmetic repeats. The figures are invented for the illustration, and the movement in one direction is the whole point: nothing in the mechanism pushes the price back down.

Every countermeasure interrupts one of those steps. Letting the insurer ask questions splits the single price into two, so neither group is subsidizing the other. Restricting when people may join removes the option to wait until you need it. Subsidizing the premium keeps the low-risk buyers in the pool at a price that would otherwise drive them out.

Why underwriting against it helps buyers

• Splitting one blended price into risk-based prices means a low-risk buyer can purchase coverage at something close to their own expected cost.
• Without it, a voluntary market can shrink to the people who need it most, at a price only they will pay, which serves nobody well.
• Waiting periods and enrollment windows are cheaper countermeasures than outright exclusion, and they keep coverage available to people whose risk is already known.

The costs, which are real

• Every mechanism sorts people by characteristics they largely did not choose, and the people sorted out are usually the ones with the most to lose.
• A buyer whose risk is known and high may face a price that is accurate and unaffordable at the same time.
• Anti-selection rules make coverage harder to buy at exactly the moment a household realizes it needs it, which is when most people go looking.
• Where the law removes underwriting without adding an offsetting mechanism, the price rises for everyone rather than the problem disappearing.