OBJECTIVE: To investigate how baseline health insurance coverage affects subsequent out-of-pocket costs and utilization of health services over a two-year period. DATA SOURCE: The first two waves of the Health and Retirement Study, a nationally representative survey of the noninstitutionalized population, ages 51 to 61 at baseline. Interviews were conducted in 1992 and 1994. Our sample consisted of 7,018 respondents who did not report public insurance as their sole source of coverage at baseline. STUDY DESIGN: We compared self-reports of physician visits, hospitalizations, and out-of-pocket health care costs, measured as payments to physicians, hospitals, and nursing homes, by type of insurance coverage at the beginning of the period. We estimated multivariate models of costs and service use to control for individual health, demographic, and economic characteristics and employed instrumental variable techniques to account for the endogeneity of insurance coverage. PRINCIPAL FINDINGS: Controlling for personal characteristics and accounting for the endogeneity of insurance coverage, persons at midlife with job-related health benefits went on to spend only about $50 per year less in out-of-pocket payments for health services than persons who lacked health insurance at the beginning of the period. However, they spent about $650 more per year in insurance premiums than the uninsured. The uninsured used relatively few health services, except when they were seriously ill, in which case they were likely to acquire public insurance. CONCLUSIONS: The medically uninsured appear to avoid substantial out-of-pocket health care costs by using relatively few health services when they are not seriously ill, and then relying upon health care safety nets when they experience medical problems. These results suggest that the main impact of non-insurance at midlife is not to place the locus of responsibility for costly health care upon individuals. Instead, it discourages routine care and transfers the costs of care for severe health events to other payers. Our findings on the high cost of employment-based coverage are consistent with evidence that the proportion of workers accepting health benefits from employers has been declining in recent years.