April 22, 2019 – The U.S. is in a mortality crisis and you may be surprised by the population most affected. From 1992 to 2015, mortality rates among middle-aged non-Hispanic white Americans in the least educated 10 percent of the population increased by more than any other group, according to a Dartmouth-MIT study. Deaths from suicide, chronic liver disease and drug overdose explain only some of the change; the least educated white Americans are facing higher mortality from a range of other causes as well. Middle-aged whites without a high school education now have higher mortality than blacks of similar age and education. The findings provide new insights into who the mortality crisis is affecting most.
The research is summarized in a recently updated working paper by Paul Novosad, an assistant professor of economics at Dartmouth and Charlie Rafkin, a graduate student in economics at MIT, who began working with Novosad as a Dartmouth undergraduate.
Typically, when the mortality crisis in the U.S. is discussed, the focus is on how the mortality rate among whites in the bottom half of the education distribution—those with a high school education or less—has been increasing, based on earlier work by other researchers. This earlier work could not examine mortality rates among the individuals at the very bottom of the education distribution (i.e. high school dropouts) because of the statistical challenge of measuring mortality change in a group with substantial population changes due to rising education overall. The Dartmouth-MIT team developed a new statistical method to distinguish between mortality changes among the bottom 10 percent of the education distribution and the bottom 50 percent, and found that nearly all of the present mortality crisis is driven by changes at the very bottom.
To more accurately measure middle-age mortality rates in the U.S. population (from 1992 to 2015 in ages 25 to 69), the researchers drew on death records from the U.S. National Vital Statistics System of the National Center for Health Statistics and education data from the American Community Survey and the Current Population Survey. Building on previous research, their analysis focused — as do previous studies — on non-Hispanic blacks and whites. The team developed a methodology that examines mortality at the bottom 10 percent of the education distribution and solves the selection bias problem in earlier studies. They disaggregated the data into four groups representing different parts of the education distribution: the bottom 10 percent (roughly high school dropouts); percentiles 10-45 (high school completers), percentiles 45-70 (some college); and the top 30 percent (B.A. or higher). Most earlier researchers had lumped together mortality data on high school dropouts and high school graduates because of the statistical difficulty of estimating mortality change at the very bottom of the distribution.
There were several surprising findings. Mortality rates among high school dropouts have always been higher than among more educated groups, but the divergence between these groups from 1992–2015 was stark. The moratlity rate of white women aged 50 to 54 in the least educated 10 percent rose from 625–735 deaths per 100,000 in 1992–94 to over 1,300 in 2013–2015, an increase of close to 100 percent, or about 3 percent per year. A mortality increase like this for any subset of the population is almost unheard of in recent U.S. history. In contrast, mortality among the most educated 30 percent of whites and blacks declined at a steady rate that is comparable or better than total population mortality declines in Western Europe and Canada.
The figure shows change in mortality and average education rank for non-Hispanic white women aged 50-54 at different levels of education, from 1992-1994 to 2013-2015. Each point represents the average number of deaths per 100,000 people among people with one of four levels of education: No High School, High School, Some College, and a B.A. or Higher. The X coordinate of each point represents the average education percentile among people with the given level of educational completion. Education percentiles were computed using the American Communities Survey and Current Population Survey, and death records were obtained from the National Center for Health Statistics.
This chart is one of four charts from Figure 1 of the paper. Image provided by Paul Novosad, Dartmouth and Charlie Rafkin, MIT.
The study also found that mortality rates among blacks have improved across all education groups except for those in the lowest 10 percent. For individuals in the least educated 10 percent, white and black mortality rates have converged, and at some ages, the trend lines have crossed. The researchers write, “At the very bottom of the education distribution, white men over the age of 50 now have higher mortality than black men as do the least educated white women over the age of 40 relative to similarly-aged black women.”
Prior work on mortality change among whites has focused on poisoning, drug overdose, suicide and chronic liver disease, the so-called “deaths of despair” which are understood to be closely related to the opioid crisis. The authors confirmed that deaths of despair are rising substantially among almost all middle-age, gender and race groups, but they found that changes in these deaths account for less than half of mortality change among middle-aged whites. The least educated middle-aged white women in particular faced rising mortality risk from cancer, heart disease, respiratory diseases and many other causes in addition to deaths from despair.
“Our findings on U.S. mortality rates illustrate that this crisis is much more focused at the very bottom of the socioeconomic ladder and is thus more concentrated than was previously thought,” explained Novosad. “Low education is a marker of permanent socioeconomic disadvantage; our work charts a path for researchers to better understand which people are in the center of the current mortality crisis and why,” he explained.
An earlier version of the working paper was posted in October 2018. The updated version was posted in April 2019. In this press release and in the paper, “white” refers to non-Hispanic whites and “black” refers to non-Hispanic blacks.
Novosad is available for comments at: firstname.lastname@example.org.