A large literature now exists on the cross-national correlation between income inequality and population health, but existing studies suffer from sparse data, poor operationalization of income inequality, and the use of low-power statistical models. This paper sets out to estimate the ecological correlation between income inequality and indicators of population health in a very broad panel of countries, to demonstrate that this relationship is largely non-artifactual, and to test whether this relationship might be causal. Gini coefficients of national income inequality in 1970 and 1995 are correlated with life expectancy, infant mortality rates, and murder rates, controlling for national income per capita. In cross-sectional analyses, inequality is significantly correlated with life expectancy, infant mortality, and (inconsistently) the murder rate. The health correlations are shown to be not primarily due to the "convexity effect" of the non-linear relationship between individual income and individual health, which seems to account for no more than one-third of the relationship between inequality and health, and likely much less. Change in inequality 1970-1995 is significantly related to change in life expectancy and infant mortality, suggesting a causal relationship, but these correlations are not robust with respect to sample or controls. It can be concluded that there is a strong, consistent, statistically significant, non-artifactual correlation between national income inequality and population health, but though there is some evidence that this relationship is causal, the relative stability of income inequality over time in most countries makes causality difficult to test.