To combat adverse selection, governments increasingly base payments to health plans and providers on enrollees’ scores from risk-adjustment formulae. In 2004, Medicare began to risk-adjust capitation payments to private Medicare Advantage (MA) plans to reduce selection-driven overpayments. But because the variance of medical costs increases with the predicted mean, incentivizing enrollment of individuals with higher scores can increase the scope for enrolling "overpriced" individuals with costs significantly below the formula's prediction. Indeed, after risk adjustment, MA plans enrolled individuals with higher scores but lower costs conditional on their score. We find no evidence that overpayments were on net reduced.