The Clinton proposal recognizes the need for successful risk adjustment and calls for the National Health Board to promulgate a risk adjustment formula by 1 April 1995. Unfortunately, risk adjustment technology is primitive; using observable characteristics such as age only slightly ameliorates the flawed incentives of not adjusting at all. Without major improvements in risk adjustment technology we face a trade-off between giving plans an incentive to select good risks and an incentive to produce at lowest cost. Pure capitation maximizes both incentives; pure fee-for-service minimizes both. I suggest experimentation with paying plans partly on the basis of risk-adjusted capitation and partly on the basis of a fee schedule reflecting actual use (partial capitation). In the draft Clinton plan, the option given to alliances not to offer plans priced above 120 percent of the weighted average premium appears to assume better risk adjustment ability than is now possible. This option should be relaxed or abandoned.