We adopt a new representation of the relationship between emissions and income using long-run growth rates. Our approach allows us to test multiple hypotheses about the drivers of per capita emissions in a single framework and avoid several of the econometric issues that have plagued previous studies. We find that for carbon dioxide emissions, scale, convergence, and resource endowment effects are statistically significant. For sulfur emissions, the scale and convergence effects are significant, there is a strong negative time effect, and non-English legal origin and higher population density are associated with more rapidly declining emissions. The environmental Kuznets effect is not statistically significant in our full sample for either carbon or sulfur.