Data Avengers

Are death rates really soaring for middle-aged white Americans?

It's hard to argue with a paper written by a Nobel Laureate. But statistician Andrew Gelman is, in a way, doing just that.

focus group 3 1024x878 copy
Credit: Andrew Gelman

Data Avengers

Show More

It's hard to argue with a paper written by a Nobel Laureate. But statistician Andrew Gelman is -- if not debunking (as he points out in a blog post), at least clarifying -- much-reported research that death rates are soaring for middle-aged white Americans.

"Unlike every other age group, unlike every other racial and ethnic group, unlike their counterparts in other rich countries, death rates in this group have been rising, not falling," wrote the New York Times in fairly typical coverage of recently published research by Princeton economists Angus Deaton and Anne Case. (Deaton won the 2015 Nobel Prize in economics for his analysis of consumption patterns). The rise, this research found, was due to drug abuse.

But are the death rates really rising?

Gelman, director of Columbia University's Applied Statistics Center, points out two interesting data tidbits.

1. The "45- to 54-year-old" age group is not uniform throughout the years. And in the U.S., "what with the baby boom generation moving through, the average age in the 45-54 group crept up." It hasn't gone up a lot, Gelman admitted; but since mortality does increase substantially with age, even that 0.4 year rise can have an effect. According to Gelman, once you adjust for that change in average age, death rates have flattened since 2005.

2. While some commentators leapt to the conclusion that white American men are being affected by this trend ("What's killing middle-aged white dudes?" asked Will Bunch at the Philadelphia News), in fact death rates for men in this group have been falling while death rates for non-Hisptanic white women in that age bracket are rising.

What is really happening, he said after analyzing the gender and year across time data, is "increasing mortality among women aged 52 and younger -- nothing special about the 45-54 group, and nothing much consistently going on among men."

A main point of the paper, Gelman notes, compares unusually high death rates in the group with other countries; and "that holds up just fine. The place where everyone is confused is about the trends among middle-aged non-Hispanic white Americans."

Lessons here: Make sure the groups you're comparing over time are in fact similar, and take a dive into subgroups if you're looking for important trends.

Computerworld's IT Salary Survey 2017 results
Shop Tech Products at Amazon