Correlates of Middle-Aged White Mortality

Anne Case and Angus Deaton’s paper on middle-aged white mortality was a bit of a three days wonder– why was mortality rising for middle aged American whites when it was falling for American non-whites and for Europeans?– that collapsed into an argument about the degree to which the observed increase in all-cause mortality was driven by changes in age composition of the cohort, as white Baby Boomers age. (Eg, Statistician Andrew Gelman and sociologist Philip Cohen  )

The paper itself is short and very clearly and convincingly written.

Gelman and Cohen’s age composition critique may or may not cast doubt on the idea of a *general* increase in mortality across all whites, particularly among men, but I don’t see how anybody can look at the dramatic increases in suicide across all age groups (Figure 4) in poisoning and suicide in all regions of the country (Figure 3) in the ratio of mortality causes between lower and higher education groups (Table 1), and in reported pain morbidity and disability (Table 2) and think that lower-education whites are in a good way. The single most striking figure is not Figure 1, which has been reproduced in many outlets and is what Gelman takes issue with, but Figure 4, which shows that the collective mortality due to suicide, alcohol, and drugs has doubled across *all* age cohorts of whites, clearly not due to age composition.

I did a little poking around with different data sources to see what county-level death rates for this cohort (45-54 YO non-Hispanic Whites) was associated with in 2013 (not looking at changes in death rates since 1999, so as to obviate the age composition issue.) Not surprisingly, richer counties have lower 45-54 YO NHW death rates:

As they say, money can’t buy happiness, but the log of money can.

Obesity (the CDC’s 2012 numbers) has a lot of explanatory power for middle-aged white death rates in 2013, both by itself:

And even if you control for median household income, the elasticity of death rates with obesity, controlling for median household income, is a little under 0.4; if a county increases its obesity rate from 24 percent to 25 percent, holding income constant, it is associated with an increase in death rates for 45-54 YO Non-Hispanic Whites of 1 percent.

The CDC says that obesity rates for women didn’t increase that much in the last decade: even as death rates rose more for women than for men, but perhaps we’re seeing a delayed effect of the huge climb prior to 2000, with increased pain and reduced mobility being partially responsible for the large increases in drug abuse, alcohol poisoning, and suicide that Case & Deaton observe. On the other hand, obesity could be just correlated with other social changes that are the real drivers of the increased deaths, though it would have to be in a way that didn’t disappear when you control for income.

Disability data isn’t publicly available for all counties in every year, since the CPS doesn’t sample every county in every year. That means you lose some sample size, but even so the rate receiving SSI-Disability in a county is the single best predictor of death rates in this cohort (45-54 YO Non-Hispanic Whites in 2013) of anything.

You can tell some of this isn’t just the effects of poorer, more rural counties also having higher disability rates, since the sign on county population reverses itself once you control for disability and income. While more rural counties (those with lower population) definitely have higher death rates:

Even so, the association between population and death rates disappears once you control for obesity, disability, and income, which by themselves explain about 66 percent of the county-level variation.

To me, this all suggests that Case & Deaton’s strong implications that increasing obesity, morbidity, and disability (aside from any economic changes) are important drivers behind the rising mortality rates seems right, perhaps with pain of various sorts (and uncontrolled pain management) being an important mediator. Other social changes might also be important- eg, marriage rates appear to explain a slight amount of the variation, though no more than an additional percentage point when controlling for the above factors, and focusing on employment rates as opposed to income doesn’t seem to make a difference. And there remains the question of why increasing obesity and disability rates have not produced the same surge in mortality rates among other groups as they have in non-Hispanic whites.



Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s