In this paper we empirically test the Kuznets curve notion for mortality. In particular, we assess if the relationship between income and overall as well as cause-specific mortality rates changes for different levels of income. Moreover, we find the income levels for which this relationship changes.
We follow a similar empirical approach as in the EKC literature. We use a panel data set of overall and broad categories of mortality rates and GDP per capita for the 32 states in Mexico from 1993 to 2008. We use Mexico because there is a large range and heterogeneity in GDP per capita at the state level, where the per capita income of the poorest states is similar to that of Ecuador and the richest states are closer to the Czech Republic. Moreover, the Mexican mortality data are considered to be high quality, data collection procedures are homogenous and all the states are under the same public health system.
We adapt the latest econometric specification of the EKC by performing state-level regressions of the mortality rates against a linear, a quadratic, and a cubic term of the GDP per capita. In all our regressions we use fixed effects at the year and state-level, cluster errors by state and weight mortality by the square root of the population. We separately regress overall mortality and three broad mortality categories: communicable, non-communicable and injuries. Moreover, we use quadratic and cubic specifications separately to find the best fit for the relationship.
We do not find a statistically significant relationship for overall mortality. However, for communicable diseases the relationship statistically significant and consistent with the cubic specification. In this case, communicable mortality rates first increase for per capita income between 1,000 and 12,000 pesos, then decreases for incomes between 12,000 and 30,000 and then increases again for incomes above 30,000.
Regarding non-communicable diseases we also find an statistically significant relationship consistent with the cubic approach. However, for non-communicable causes we find that mortality first decreases for income between 1,000 and 17,000 pesos per capita, increases for income between 15,000 and 35,000 and finally decreases for income above 30,000.
For injuries we find that the relationship is statistically significant and quadratic with a bell-shape form. That is, mortality rates for injuries first increases with income and later decreases with the peak at around 35,000 pesos per capita.
Thus, our results suggest that for some mortality categories may be a mortality Kuznets curve.