The evidence for favorable selection into HMOs comes from a series of natural experiments that have the following framework: Suppose everyone in an employer group is enrolled in a conventional health plan that collects detailed information on employees’ use of covered health services. Then, at some open enrollment period, employees can choose to take a newly offered HMO or to remain in the existing plan. Once people have made their choices, the researcher goes back into the preceding year’s claims data and compares the health services use of those ultimately choosing the HMO with those ultimately choosing to stay in the conventional plan. If favorable selection into the HMO is present, we should see that, prior to having a choice, those who ultimately chose the HMO had lower claims experience. In contrast, if the conventional plan retained the low utilizers, the prior claims experience of its ultimate enrollees should be lower. If no favorable selection exists, then no difference in the reported levels of prior utilization should exist.