If You’re Poor, Where You Live Matters

Originally featured on WantHealthcareLLC.com.

In a recent article in the Journal of the American Medical Association, economists from Stanford and Massachusetts Institute of Technology (MIT) did a very interesting thing. There is a general assumption that the richer you are, the longer you live, on average. This turns out to be true. But why does increasing income mean increasing lifespan? These economists tested that hypothesis by using two related data sources: tax returns and Social Security Administration death records. Using these two databases, they were able to map life expectancy by income level across different “commuting zones” around the country, over a five year period. This involved roughly 1.4 billion tax records over that time.

The results are fascinating. Some of the conclusions:

• If you’re wealthy, geography matters less. Pretty much across the country, if you’re a man in the top 1% of income, you live to be about 87; a woman in the same income bracket lives to be almost 89.
• If you’re poor, where you live matters a lot. The difference for the bottom quartile in income is 4.5 years between the best areas of the country and the worst. This is roughly the difference we’d see in population life expectancy if we were able to cure cancer (i.e., not a small difference!).
• The study did not support four popular theories about the causal link between income and longevity.
  • First, access to medical care didn’t correlate with living longer, measured by percent uninsured, Medicare spending, 30 day hospital mortality rates, and use of preventive care.
  • Environmental factors didn’t seem to matter. This was measured pretty indirectly, looking at the degree of income segregation in a commuting zone. The reasoning goes like this: if you have money, you move away from environmental hazard, and therefore more highly income-segregated areas would have a bigger difference between rich and poor life expectancy. Contrary to what one might expect, those regions that were more income-segregated had better longevity for the poor, not worse.
  • Greater income inequality was not correlated with shorter lifespans for the poor. It was correlated with shorter lifespans for the rich. Social cohesion was not correlated with lifespan.
  • Finally, local labor market conditions did not correlate with lifespan among the poor. The theory that the best thing we can do for health is to get people jobs wasn’t supported in this study.
• These were some of the characteristics that were associated with areas with higher life expectancy for the poor: higher percentage of immigrants, higher incomes, higher governmental expenditures, higher population density, and higher college graduate rates.

What else correlated with increased lifespan for the poor? In addition to the factors above, they were the tried and true behavioral factors we already know about: lower smoking and obesity rates and higher exercise rates.

There is much more in this study, and I will continue to ponder its findings. But in an oversimplified way, this tells me that the best way to improve the health of communities may not be to get them more health care. The best way might be to alter the health habits of low income residents. There may also be sound societal reasons for trying to clean up the environment, reduce income inequality, and reduce unemployment, but improving the health of the population isn’t one of them, at least according to the correlations found in this study.

The other thing we should notice is that this study took massive amounts of computing power. Manipulating 1.4 billion records is no small feat, but we are likely to see more studies like this because of the emergent force of big data. We should see these kinds of insights with increasing frequency as analytics get ever more powerful.

*Editor’s Note:
The Colorado All Payer Claims Database (CO APCD) was recently updated with Medicaid and commercial health insurance information through 2014, and has new data on chronic disease prevalence (diabetes, asthma, hypertension, and depression) by county and other new measures like breast, cervical and colorectal cancer screening rates. The CO APCD combined with other social determinants of health data provide a rich source of information for Colorado to begin investigating opportunities to improve health by evaluating variation across regions and demographics.

About the Author: Dr. Jay Want is CIVHC’s Chief Medical Officer. Contact him at jwant@civhc.org.