Is inequality deadly and for whom? A Bayesian Model Averaging analysis

The persistent correlations between income and mortality, and inequality and mortality in the US are well established. It has been demonstrated with various empirical models at different levels of analysis. However, there is no consensus on the relevant confounding factors or even the functional specification of the income-inequality-mortality relationship across the literature. We interpret this as significant model uncertainty when it comes to the correct specification of the implied econometric model and provide a systematic approach to address it. Using BMA, we conduct a large-scale analysis considering millions of models to determine a model-averaged inequality effect explaining cross-sectional variations in mortality at the county level. The results also suggest a best set of confounding factors and emphasize the importance of controlling for unobserved State-specific factors. Furthermore, we uncover a robust nonlinear income-inequality-mortality relationship that challenges typical assumptions in the literature.