It's difficult to interpret this because the paper defines income throughout the text in terms of income quantile (not absolute, COL-adjusted levels). And oddl

It's difficult to interpret this because the paper defines income throughout the text in terms of income quantile (not absolute, COL-adjusted levels). And oddly, in the FT, they show a significant negative correlation b/t Gini index and life expectancy in the upper quartile, and not the lower quartile (you'd expect the opposite). In the lower quartile, it is p=0.11, r=0.2; i.e., a nonsignificant positive association between income inequality and income quartile. This suggests the effects of income inequality might actually affect LE in different directions for high and low earners. This would explain why they say in the abstract they don't see a LE-Gini correlation for the distribution as a whole. The story I would spin is that, within one geographic region, a high Gini actually means that rich and poor are living in proximity and sharing social structures. If you have low Gini in a compact region it likely means everyone there is (uniformly) poor. IOW, they very likely have a correlation b/t Gini and income levels they aren't controlling for. This is why we use multivariate regression with interactions, not just run a Pearson between every pair of variables. The statistics in this article are really bad. JAMA strikes again.