Using prospective cohort data from the 1979 National Longitudinal Survey of Youth, this study examines the extent to which health insurance coverage and the source of that coverage affect adult health. While previous research has shown that privately insured nonelderly individuals enjoy better health outcomes than their uninsured counterparts, the same relationship does not hold for those publicly insured through programs such as Medicaid. Because it is unclear whether this finding reflects a true causal relationship or is in fact due to selection bias on socioeconomic status and health, previous estimates of the contribution of health insurance to inequities in health may have been biased. This study attempts to disentangle these competing hypotheses of causation or selection bias by using fixed effects models with sibling clusters to corroborate--or contradict--the results of a conventional OLS regression. By controlling for unobserved factors shared by siblings, such as parental genetic influences, sibling models estimate health insurance effects that are less affected by selection bias. Findings suggest that, among the U.S. birth cohorts of 1957 to 1961, the negative relationship between public health insurance and health is not causal, but rather due to prior health and socioeconomic status. Conversely, the lack of health insurance coverage has a strong cumulative negative impact on adult health. (Contains 2 tables, 2 figures, and 7 notes.)