Adverse selection

When buyers can't see quality, they pay only for average quality. Sellers of good units then withdraw, leaving "an adverse selection of low-quality goods, the lemons" (as summarized in the Nobel committee's 2001 scientific background).

One blended price or bid draws in the worst slice of applicants, vendors and leads.

Where vocabulary such as moral hazard begins. Arrow warns that rivals offering "lower premiums to preferred risks" leave a plan that doesn't discriminate "with only an adverse selection of them."

The cream-skimming problem, named in 1963. If you don't sort, a competitor will sort for you.

Insurers screen with a menu: full cover at a high price that attracts high risks, and a cheaper policy with a deductible that attracts low risks. A single pooled price can't survive, because a rival can "cream-skim" the low risks.

Tiers and commitments sort customers. One flat offer invites competitors to take your best ones.

Raising the interest rate attracts riskier borrowers and changes how borrowers behave, so "it may be optimal for banks to ration the volume of loans instead of raising the lending rate" (as summarized in the Nobel committee's 2001 scientific background).

The terms you set decide who applies. Loosening terms to clear more volume can lower the quality of the book.

When Harvard switched to contributing the same amount to every employee health plan, competition "reduced Harvard's premiums by 5 to 8 percent," but adverse selection caused "a welfare loss of 2 to 4 percent of baseline spending." Einav and Finkelstein call it "an empirical case study of a death spiral"; a companion chapter (Cutler and Zeckhauser) reports the generous plan disappeared "within three years."

A real death spiral. Flat pricing plus free choice sorts the costliest customers into your richest offer.

Adverse selection "exists in some insurance markets but not in others." Some markets show advantageous selection, where the most eager buyers are the most careful and cheapest to serve.

The counterpoint: adverse selection isn't everywhere, and sometimes the most eager buyers are the best. Measure it in your own book. It's also the clearest one-chart picture of the mechanism.

From randomized mail solicitations, "respondents to solicitations are substantially worse credit risks than nonrespondents," and people who accept inferior offers are "significantly more likely to default," even after controlling for everything the issuer knew.

Response rate isn't quality. The worse the offer, the worse the responders.

People who respond to "inferior solicitation offers have poorer credit quality attributes" and "are significantly more likely to default ex post."

A peer-reviewed study with the same finding: offer design selects the customer.

58,000 randomized offers from a South African lender separated who signs up from how they behave afterwards. The study found "strong evidence of moral hazard and weaker evidence of hidden information"; "perhaps 13% to 21% of default is due to moral hazard."

Terms change behavior after the sale, not just who buys. Don't blame everything on selection.

At a subprime auto lender, demand is "highly sensitive to down-payment requirements," "default rates rise significantly with loan size," and "higher-risk borrowers demand larger loans."

Two signals in one paper. The size of the ask carries information, and a down payment works as a screen.

Applicants insurers reject hold enough private information "to explain a complete absence of trade" in long-term care, disability and life insurance.

Some segments can't be served profitably at any price, and turning them away is rational. Not every "no" is a lost sale.