Income Inequality and Later Life Health: Estimating Life Course Treatment Effects

I propose mechanisms through which exposure to income inequality early in life might be functionally related to later-life health. I first frame and heuristically model how inequality might theoretically affect health. The model explicitly recognizes that health results from short- and long-run production processes. The model allows for the possibility that experienced income inequality a) directly affects health; b) changes the availability and price of publicly financed health inputs; and c) has no effect on health inputs. As candidates underlying the first mechanisms, researchers have proposed that inequality affects levels of stress people experience - that then degrades an individual's health (Sapolsky et al 1997; Abbott et al. 2003) and/or that inequality leads to lower health because it increases the relative deprivation/social position a person occupies (Wilkinson 1996). The best evidence for the second mechanism is from Araujo et al. (2008) who shows that income inequality is related to how municipalities allocate resources (in favor of health inputs). The latter channel allows for the spurious correlation because income inequality happens to co-vary with a third factor that does affect health. Candidates for this mechanism are early-life disease conditions and/or omitted measures of medical technologies that developed in ways that mirror the trend in income inequality.

The model builds on theory and empirical evidence that suggests that inequality experienced in critical parts of life might matter. The model’s predictions rest on these temporal effects. The empirical implementation of this approach demands much data. I develop the required data. I use Panel Study of Income Dynamics (PSID) data. The PSID is a longitudinal survey that follows individuals for up to 49 years. I construct and map to individual PSID respondents measures of income inequality experienced every year over each person's whole lifetime. I measure state income inequality using the Gini coefficient. I estimate the Gini coefficient for each state in each year using observed state-year Ginis in the decennial Census years 1940 and later and from the CPS from 1976-2008. I then backcast to 1929 each state Gini coefficient using Bureau of Economic Analysis data on employment and earnings by industry. I also control for state-specific spending on factors that plausibly affect health – hospitals, welfare, etc. I examine the correlation between outcomes and lifetime exposure to inequality and to inequality experienced during theoretically critical periods. Finally, I examine the extent to which income inequality proxies for systematic cross‑state differences in other determinants of health. Early life income inequality matters – but not always in the way a casual observer might guess.