article in this Special Theme Issue of Medical Care by Naessens et al1 is of particular interest to a health economist, because it provides a confirmation, under almost laboratory conditions, of important theoretical results from more than 30 years ago. Choice of insurance coverage (all else equal) leads people to self-select into more and less high-risk groups. Greater choice thus has the effect of placing a greater financial burden on those at higher risk the elderly and the chronically ill. The theory works, and has immediate relevance to present-day health policy. In 1970, Akerlof s classic Lemons paper2 explored the dynamics of markets with asymmetric information. In the extreme case, transactors on one side of an exchange might have complete information about the relevant characteristics of a given product (sellers of used cars, buyers of individual health insurance contracts) whereas those on the other side might know only the average characteristics of used cars offered for sale, or the average risk status of would-be insurance purchasers. Sellers of higher quality cars would then be offered too low a price for their cars; low risk buyers would be overcharged for their coverage. If the highest quality used cars and the lowest risk buyers should respond by leaving the market, this would lower the average quality of cars for sale and raise the average risk of those with health insurance. Depending upon the relative numbers and risks, this could set off a chain reaction such that more higher quality cars and lower risk people leave. The market could go into a death spiral and completely disappear. Rothschild and Stiglitz,3 however, demonstrated that even when sellers of health insurance contracts do not know the risk status of individual purchasers, they could induce (rational, fully-informed, self-interested, and risk-averse) buyers to reveal their status by offering multiple contracts with differing premiums and levels of coverage. Their demonstration model includes high-risk and low-risk buyers. If 2 different contracts are offered, low-risk buyers will choose the lower premiums and less comprehensive coverage more potential user charges while high-risk people will pay more for more comprehensive coverage. As they did at Mayo. In the real world of Naessens et al, the separation is of course not complete. But the higher-risk employees are clearly self-selecting into the higherbenefit, higher-premium pool, and conversely the theory works. (Akerlof and Stiglitz shared the 2001 Nobel Prize in Economics with Michael Spence for their work on the economics of imperfect information.) It should be noted, indeed emphasized, however, that this separation process has significant distributional effects relative to the single premium contract. Income is transferred from the less to the more healthy members of the insured population, in 2 distinct ways. (There is also a transfer from employees to the employer, insofar as the increases in premiums and user charges permit lowering the employer's contribution.) Low-risk buyers gain because their premiums no longer have to cover part of the costs of high-risk people. And the latter correspondingly pay more for whatever coverage they buy. But in addition, because self-selection is motivated by increased user charges on