This study uses a natural experiment approach to evaluate the effect of health insurance on infant and child mortality. In the 1970s Costa Rica adopted national health insurance, which expanded children's insurance coverage from 42 percent in 1973 to 73 percent by 1984. Aggregate infant and child mortality rates dropped rapidly during this period, but this trend had begun prior to the insurance expansion, and may be related to other changes during this period. We use county-level vital statistics and census data to isolate the causal insurance effect on mortality using county fixed effects models. We find that insurance increases are strongly related to mortality decreases at the county level before controlling for other time-varying factors. However, after controlling for changes in other correlated maternal, household, and community characteristics, fixed effects models indicate that the insurance expansion could have explained only a small portion of the mortality change. These results question the proposition that health insurance can lead to large improvements in infant and child mortality, and that expanding insurance to the poor can substantially narrow socioeconomic differentials in mortality.