There is a large literature that documents a positive correlation between income and a variety of measures of good health. This correlation may reflect causality in both directions and may also reflect omitted "third variables" that are positively related to income and health. In my research, I employ an exogenous negative shock to income due to a natural disaster to estimate the true causal impact of income on health, focusing on diabetes. The shock I will use is Hurricane Katrina, which severely damaged counties in Alabama, Mississippi, and Louisiana in August 2005. I use these treatment counties and a variety of alternative sets of control counties in a difference-in-differences (DD) research design. From the Behavioral Risk Factor Surveillance System I obtain measures of income and health for residents of these counties before and after the date on which the hurricane struck. I estimate DD regressions in which income or health is the dependent variable to obtain the impact of the hurricane on each outcome in the treated counties net of other factors. I then use the interaction between an indicator for residents of the treatment counties and an indicator for the period after the hurricane struck as an instrument for income in a two-stage least squares regression of diabetes on income.
Jang Wook Lee. "The Effect of Income on Diabetes after Hurricane Katrina." Proceedings of the New York State Economics Association. vol. 7, October 2014, p. 88-100
BibTeX entry download