A standard test for adverse selection in health insurance examines whether people with characteristics predicting high health care utilization are more likely to buy insurance (or buy more generous nsurance). George Akerlof’s theory of adverse selection suggests a test based on prices: those who purchase insurance at the regular price will have higher expected utilization than those buying insurance when offered a deeply discounted price. Both tests provide (different) lower bounds on self-selection. We use a randomly allocated coupon for deeply discounted health insurance in rural Cambodia coupled with a longitudinal survey to test for adverse selection. While the standard test can show only a small amount of self-selection, the Prices test shows vastly more self-selection – providing a much more informative lower bound.