Aggregate age-adjusted trends in death rates for non-Hispanic whites and minorities in the U.S.

Following up on our recent Slate article, Jonathan Auerbach made some graphs of mortality rate trends by sex, ethnicity, and age group, aggregating over the entire country.

Earlier we’d graphed the trends within each state but there was so much going on there, it was hard to see the big picture.

All our graphs are age adjusted.

Also, each graph is on a different scale. Graphs are scaled just to include the data. So look at the y-axes carefully if you want to compare different plots. As always, there’s a tradeoff between the unambiguousness of a common scale or the higher resolution of different scales. In these graphs we went for resolution. Soon we’ll post all our data and R code and then you can easily play with the code and make your own graph. Or you can go to CDC Wonder, download the data, and make your own graphs right now.

All trends are from 1999-2014.

Summary

Minorities used to have higher death rates than whites among almost all age groups. But now, death rates for whites and minorities are close to equal in most age categories.

During the past decade and a half, death rates for in the different minority groups have steadily declined in most age categories, while death rates for whites have declined more slowly and actually increased in a few age categories (25-55-year-old women and 25-34-year-old men).

As demonstrated in our other document, these patterns vary a lot by state and region of the country.

The graphs below compare non-Hispanic whites to others for the U.S. as a whole.

Non-Hispanic whites vs. all minorities

Non-Hispanic whites vs. blacks vs. Hispanics vs. others

The controversy

Economists Anne Case and Angus Deaton wrote two papers on mortality rate trends. These two papers did the service of getting these trends discussed in the news media and in the scholarly community more broadly. Unfortunately most of the discussion of this in the news media, back in 2015 and again now, began and ended with Case and Deaton, not giving other perspectives.

I think the best recent news article on the topic is this piece by Malcolm Harris, who went to the trouble to read the literature and interview some demographers who work in this area. Harris goes into detail on the problems with Case and Deaton’s comparisons of trends in different education categories. It’s not that such comparisons shouldn’t be done, but you have to be careful about the interpretation, because of selection bias.

The rest of the news media (NYT, NPR, etc) pretty much punted on this one and just did straight Case-Deaton with no alternative perspectives. Kinda frustrating, but this is the tradition in science reporting, I guess. As I wrote earlier, I’m not saying these journalists had to talk with me; the appropriate people to interview are various actuaries, demographers, and public health researchers who understand these numbers inside and out.

Obligatory criticism of the graphical display

I’d prefer a different color scheme—it’s kinda weird that the color for white people changes from the first to the second set of graphs. Also, I’d take advantage of the scrolling feature of html and display the graphs in a 10 x 4 grid: that’s 10 age categories x 4 batches of graphs (women whites vs. minorities, men whites vs. minorities, women whites vs. blacks vs. hispanics vs. others, men whites vs. blacks vs. hispanics vs. others).

Or maybe, how’s this for an idea: 10 x 5, each graph has one line for men and one for women, and the 5 columns are whites, blacks, hispanics, others, and all minorities. That could work! At least, it’s worth a try. I don’t like the current graphs with 4 different colored lines; it’s too hard for me to keep these clear in my head. I keep finding myself going back and forth between the plots and the color code, and that’s not good.

57 thoughts on “Aggregate age-adjusted trends in death rates for non-Hispanic whites and minorities in the U.S.

    • Steve:

      Case and Deaton definitely have a point. Their original claim of “a marked increase in the all-cause mortality of middle-aged white non-Hispanic men and women in the United States between 1999 and 2013” was wrong, but they were right that the flat trend for middle-aged non-Hispanic whites was much different from the decline in other age groups, other ethnic groups, and other countries. Their comparison by education level is interesting too, although I suspect their analysis overstating the differences because of selection bias. I’m glad the news media reported the story that Case and Deaton wrote about, and I’d like the media to do the next step and talk with expert actuaries, demographers, and public health researchers to understand this all better.

      • Rahul:

        I don’t actually know who’s calling it “White Death”; I don’t recall that phrase from Case and Deaton. More accurate would be “White non-decline in mortality rate (except among young children and the elderly.” Then again, I’m a statistician so I like to give very accurate but boring descriptions.

        • Andrew:

          Steve Sailer is calling it “White Death”. See his comment above.

          Isn’t the dominant theme just natural saturation, in most cases? Decreasing marginal returns to effort?

          i.e. When mortalities are high (e.g. minorities) there’s many low hanging fruit but as you get them lower (whites) it gets very difficult to get any further declines.

          Barring a few cohorts, that seems to be the dominant story here?

          The only intriguing bits I see are (a) the recent spike in non-hispanic White females aged 5-15 Almost looks like a data artifact.

          (b) The (slight) upsurge in both male and female whites aged 25 to 35.

        • Rahul:

          1. I agree with Sailer that Case and Deaton have a point. I don’t agree with Sailer that White Death is an appropriate description.

          2. No on saturation: death rates appear to continue to trend down in many European countries.

          3. There are lots and lots of intriguing bits if you go to the state level data.

        • Re. #3:

          But in’t that just a symptom of the fact that data mining can always find *some* features given enough data?

        • Rahul:

          Sure, you look at the data and find interesting features; that’s how you learn. These are 15-year trends so it’s not like we’re talking about weird things that just happened in one year. We’ll want to understand these unexpected trends in different states, which could be explainable by some mixture of measurement artifacts, statistical biases, and actual trends in deaths.

        • In a general sense this question bothers me. How does one distinguish an artifact from a lasting trend. Don’t people always warn that with data mining large datasets one will always hit many artifacts?

          Is the only solution to go after every trend and whatever doesn’t endure we forget? What’s an efficient strategy to avoid unproductive dead ends.

        • People have quickly noticed all sorts of subtle things about baseball statistics.

          But it turns out America in the 21st Century suffered hundreds of thousands of early deaths by white people without that fact becoming a Thing in the press until November 2015 (and then perhaps only because Angus Deaton had just won a Nobel Prize).

          Why?

          Perhaps because there are no respectable organizations that have a mission of scanning statistics to look out for the well-being of white people? To do that is tantamount to getting you on the SPLC’s list of people to hate. So few respectable people do that.

  1. I named it The White Death on the model of The Black Death. It seems to involve white-colored drugs (heroin and prescription opiates) and, so far, it mostly seems to kill white people (and maybe American Indians).

  2. Here’s a question:

    How many incremental deaths were there in 2014 if death rates among white women 25-54 had stayed as low as they had been in 1999?

    How many incremental death were there in 2014 if death rates among white women 25-54 had declined as much as they did among Hispanics and Asians?

      • Mortality rates for 25-54 year old white men and Hispanic men were about equal in 1999, but now the white men have mortality rates 25% higher.

        If 2014 white male death rates were as low as 2014 Hispanic death rates, about 27,500 fewer youngish white men would have died that year. (These are very rough calculations but they are in the ballpark.) That’s substantially higher than the highest annual death toll in the Vietnam War.

        Or, just look at the absolute rise in death rates among white women 25-54: the absolutely higher death rate among white women 25-54 in 2014 meant about 7,500 deaths that year that wouldn’t have happened with the lower white female mortality rates of 1999. Considering the improvement in medical care, public health, and safety, it would seem like there were about 15,000 incremental deaths in 2014 among white women 25-54 than there should have been.

        Add 25-54 white men and white women together and it looks likes 40,000 or more incremental deaths in 2014.

        That’s bad.

  3. One question this does leave me with is how this white mortality trend measures up to previous trends during other decades. Is this increase in deaths as large as other trends in deaths prior to 2000? I have no prior to assume one way or another. Also not trying to ignore the problem with the rising rates in deaths. Just curious how big this problem is compared to others historically. Would also be useful for a paper on how media coverage is a function of the mortality rate for specific subgroups of the population.

    • The big increase in black male mortality rates in the late 1980s/early 1990s due to crack and Aids and crack murders got a lot of publicity fairly quickly, although the total white population is about an order of magnitude greater than the black male population.

  4. What’s the point of having small multiples if you don’t have a common y-axis? At the moment, the logic of “higher resolution” eludes me.

  5. Steve Sailer’s question as to why this wasn’t noticed earlier is a good one. I don’t think the answer is just political correctness. True, there are no interest groups scanning the data each year for press release material, but that’s true for most data outside of what’s useful for business forecasting or political polls. And while newspapers and journals aren’t as interested in “white death” as “black death”, if that were the explanation we’d still expect to see the fact being well known in the background, to demographers and suchlike.

    It’s still a good question, though, and one highly important to we scholars. It’s like, “Why didn’t economists notice what was going to happen to financial markets because of the 2007 housing bubble?” Sometimes there’s really interesting, robust, conclusions not too deep in public data, and it takes a long time before anyone notices. Partly it’s because the most interesting answers are to questions we don’t ask because we think we already know the answer (“White death rates are falling like they always do”, “Housing has bubbles, but Wall Street won’t end up taking the losses”), so we don’t even ask the question consciously or check the answer. But is that all?

    (If I may flatter, there’s a micro-level, Steve Gelman answer: lack of good graphics, so you can’t see what is obvious when you eyeball the data in good graphs.)

    • Eric,

      I think that lots of actuaries and demographers did know about these trends, but I agree that this knowledge didn’t break out into the mainstream. I’d not heard of it, for example, until the Case and Deaton thing. So I do think Case and Deaton deserve a lot of credit for bringing these trends to general attention—even if they had to exaggerate to do so. Indeed, without the exaggeration these important trends might well have not cracked the news media attention barrier.

      The other interesting part of this communication story is that it seems to have taken a published journal article by a Nobel prize winner for the trend to get noticed. After all, the data were public, and I assume that Case and Deaton were telling people about their work, maybe even giving talks on it, long before their article was published. So, at any point, anyone could’ve called up NYT, NPR, etc, and said, Hey, there’s this big trend! But that wouldn’t’ve been enough. The published article created the conditions for the news story.

      I guess my colleagues and I should’ve published Red State Blue State in PNAS rather than a political science journal, and released the results all at once rather than gradually blogging it as we learned more things. Then maybe it would’ve been perceived as a more newsworthy event and it would’ve been reported and discussed more.

      • Looking at my blog, I started vaguely tracking white mortality rate stories in the fall of 2012:

        http://www.unz.com/isteve/charles-murrays-coming-apart-vindicated/?highlight=mortality

        Charles Murray’s “Coming Apart” Vindicated
        STEVE SAILER • SEPTEMBER 20, 2012 • 400 WORDS • 151 COMMENTS [=0]

        From the NYT:

        Reversing Trend, Life Span Shrinks for Some Whites

        By SABRINA TAVERNISE

        For generations of Americans, it was a given that children would live longer than their parents. But there is now mounting evidence that this enduring trend has reversed itself for the country’s least-educated whites, an increasingly troubled group whose life expectancy has fallen by four years since 1990.

        • But the White Death wasn’t really a Thing in my head until Case and Deaton in November 2015.

          I’m a big believer in the Orwell “Newspeak” weak version of the Sapir-Whorf hypothesis: that having a verbal meme in your head makes it easier to notice patterns that are out there in the buzzing, blooming confusion of life. Now I have the meme “The White Death,” while others have a less obvious meme involving the name Deaton, but at least they have something to grasp onto.

        • > weak version of the Sapir-Whorf hypothesis
          I always thought the weak version was sensible.

          But analyzing challenging data sets is easier in 2015 than 2010 and especially 1990 and that’s a similar kind of resistance that has to be overcome to realize things.

        • Perhaps, but we had the Internet and personal computers ten years ago.

          In fact, maybe it’s getting harder to notice realities because the conventional wisdom is getting more extremist and disconnected from reality. You’ll notice that during these years of The White Death, we kept hearing more and more about White Privilege. Perhaps that’s not a coincidence?

        • “You’ll notice that during these years of The White Death, we kept hearing more and more about White Privilege. Perhaps that’s not a coincidence?”

          … or maybe people can read graphs and notice that a) death rates among blacks are still much, much higher than among whites; and b) there has been no appreciable convergence in racial wealth inequality over the period either.

          http://www.pewsocialtrends.org/2016/06/27/1-demographic-trends-and-economic-well-being/st_2016-06-27_race-inequality-ch1-05/

        • There’s no question to me that there exists a “White Privilege” in the US, and jrc’s comments are right on the money about some of the effects. But, this could well be a reason why we didn’t notice uptrends in white death rates. I mean, the assumption could be “gee white people are doing fine, it’s black people who are having a hard time” and so no one is really looking for trends within white person data.

          On the other hand, it’s also entirely possible that “we” *did* notice the trend. For example, in popular culture the TV show “Blue Bloods” focuses on at least one character who is an Iraq war veteran police officer, and has several episodes where other war veterans are involved, either as suicide cases, or as homeless homicide victims or whatever. Also the popular TV show “Longmire” has episodes involving war veteran suicides and/or treatment for mental illness associated with PTSD with risk of suicide. Furthermore, I have read in the paper multiple accounts of unusual cases of PTSD related behavior, such as the ex marine who got in a car accident in Oregon, and then fled the scene on foot wearing sandals into the snowy woods, dug himself into a hidden position and hid there until he was found 48 hours later:

          http://www.oregonlive.com/pacific-northwest-news/index.ssf/2012/02/former_us_marine_sniper_spends.html

          Also there’s been a fair amount of publicity on efforts by the military to stem the number of veteran related suicides.

          So, for example, *if* much of the uptick is related to military suicides, or veterans who are doing poorly in life (addiction, homelessness etc), then we did in fact know about it and it *was* on the radar.

          On the other hand, I don’t have specific information about how many of these deaths are suicides by veterans, and I don’t have an explanation about why they might be affecting white people more (other than, of course, that the military is predominantly white even if the other races are a few percentage points over-represented vs their general population prevalence). I could speculate that it might be because black and hispanic veterans might have started out in harder conditions, and have developed better coping skills pre-military, so that the return from military service doesn’t seem like such a decline in quality of life… but again this is speculation that would need to be tested against data on veteran suicides.

        • Daniel:

          One of the frustrating things about much of the discussion of that original Case and Deaton paper was that: (a) the data showed a clear increase in death rates among middle-aged white women, but a slight decrease among middle-aged-white men; (b) Case and Deaton obscured this important difference by pooling the sexes and not age adjusting, which is just a weird thing to do, because everybody knows to disaggregate mortality data by sex and everybody knows to age adjust; and then (c) much of the news media decided to fit this into the existing “white men are suffering” story.

          As a white middle-aged man myself, I have no problem with concern about white middle-aged men. But as a statistician and social scientist, it makes me want to scream that a statistical pattern among women was immediately reframed as part of an existing story about men.

          I have the same problem with lots of the discussion of “working class,” which has a bit of a gendered aspect to it as it evokes images of men working in factories, not so much women cleaning bedpans.

          Public attention to mortality rate trends may be new, but we’ve been hearing about the struggles of the middle-aged white American for several decades now. Just about all my life, actually.

        • jrc’s comment illustrates my point: that tens of thousands more white people started dying annually than had been expected, but nobody much noticed for 10 or 15 years because, as jrc writes:

          “… or maybe people can read graphs and notice that a) death rates among blacks are still much, much higher than among whites; and b) there has been no appreciable convergence in racial wealth inequality over the period either.”

        • My comment wasn’t about why no one noticed the uptick in mortality rates for middle-aged white women. It was about your speculative connection between missing this demographic trend and social commentary on white privilege.

          Here it sounds like you are arguing that people willfully ignored discussing the trend and, on top of that, increased their rhetoric about white privilege to cover up what was happening to middle-aged white women in mostly Eastern states. In other places, though, you seem to argue that no one noticed because it wasn’t cool or PC to notice*. I don’t see how it can be both.

          My comment above was pointing out that, despite this change in relative mortality rates across racial groups among middle-aged women, median household wealth is still an order of magnitude higher in white households than black households (where it is frighteningly close to 0) and the gap hasn’t really changed over the last decade. And so it seems like it still makes quite a bit of sense to be discussing white privilege, at least if we want to focus on aggregate-level demographic/economic trends. And so I think your comment was basically just your usual relatively-high-level Sailer trolling, and I figured it was my turn to point that out. Next month it will be someone else’s turn to remind the comment-readers here how selective your framing and arguments are.

          * http://statmodeling.stat.columbia.edu/2017/03/30/aggregate-age-adjusted-trends-death-rates-non-hispanic-whites-minorities-u-s/#comment-452477

        • Andrew: Yes, I agree that Case and Deaton’s aggregation across sexes was unfortunate, and also their age adjustment issue. But one thing I think your comment at least suggests is that you haven’t noticed in your own graphs that it’s not just about middle aged white women in the south… there’s also a very definite trend in both white women and men between age 25 and 35.

          Now age 25 to 35 is not a time when you expect a lot of internal-cause mortality (ie. stroke, heart attack, meningitis, pneumonia etc) what you expect for this age group is either accidents, suicides, overdoses, or the like.

          My own personal prior/bias is that we’re seeing a combination of war vet suicides (which we know are way up since 9/11 since the military has made a big deal about it… but I don’t know exactly how much because I don’t have the data) and also suicide due to what is in essence economic hopelessness caused by over-borrowing for skyrocketing higher education costs, skyrocketing housing costs, all together with wage stagnation and job loss that hit the recently-graduated hardest, and combined with the elimination of bankruptcy as a way to dispose student loan debt. Imagine the horror when you realize that after getting a bachelors degree in whatever from medium grade private small liberal arts college that your parents encouraged you to go to, you discover that after rent + food + student loans your net disposable income is negative $350 a month for the next 30 years with no possibility of bankruptcy. That’s gotta drive a lot of despair. I don’t know how many people that is, but I suspect it’s more than a trivial number.

        • Daniel:

          I never said, nor did I ever mean to imply, that there were no interesting trends in other age groups. I wrote a lot about age 45-55 just because that’s what Case and Deaton had written about, and that’s what everyone was talking about. My recent document with Auerbach showed all age groups and all ethnic groups.

        • Andrew: it’s great how you’ve put together the graphs, and the document you have all broken down is very nice. I’m glad you weren’t minimizing this effect on 25-35 year olds. I think it would be very interesting to look at the group white 21-40 on a year-by-year basis, and also by educational attainment. Within this group, is it the “vulnerable” non-HS grads or is it the high achievers despairing of how they can’t seem to get ahead or how the industries they are in are all rent seeking bastards? Perhaps it’s both, depending on location:

          https://www.theatlantic.com/magazine/archive/2015/12/the-silicon-valley-suicides/413140/

          https://www.theregister.co.uk/2016/07/07/ian_murdock_autopsy/

          http://business.time.com/2013/01/14/mit-orders-review-of-aaron-swartz-suicide-as-soul-searching-begins/

          are there associations with anti-depressant medications?

          etc etc.

          The story “desperate poor people more likely to die young” isn’t that surprising, the story “by all measures potentially affluent people actually living in near poverty due to lasting effects of terrible economic policy and despairing of how the only way to get ahead financially is to work for rent seeking bastards” is another much more sinister type of story.

          I am not saying I have the answers, but I certainly have these questions.

  6. @Daniel:

    >>>I mean, the assumption could be “gee white people are doing fine, it’s black people who are having a hard time” and so no one is really looking for trends within white person data.<<<

    Isn't this another framing of the age-old debate whether levels matter more or rates? e.g. Growth Rate vs Absolute GDP level etc.

    Sure White death rates may have started increasing but if they are still (say) half of the corresponding cohort's black death rate, then the question a public policy maker faces is which does he pay more attention to?

    • absolute numbers matter. I mean if 10% of the population is dying at rate x, and 90% of the population is dying at x/5 then there are a lot more people involved in the x/5 rate than the x rate….

      It also matters *why* the deaths are occurring. If the x rate is due to say genetic disorders among a minority population and the x/5 rate is due to say violent crime, you can affect violent crime, you can’t affect genetic disorders…. so it’s not so clear cut that you should look at the group with the highest rate, you should look at the group that you have the largest ability to reduce absolute number of QUALYs lost.

        • Which, it seems there’s a prima facie argument that this is what policy makers should be doing. I’m not saying you couldn’t argue otherwise, just that …. gee that seems like a really good idea.

        • If your objective were indeed to merely maximize QUALY’s per dollar spent, shouldn’t we be scrapping most of the research-funding we spend investigating cures for rare diseases?

        • Unless of course your research proposal is a way to virtually cure this rare disease that’s pretty cheap to implement.

          I see this all the time indirectly in biomed research. people get tons of money to investigate say ALS which affects 2/100k per year and then something like say nonunion bone fracture affects something like 50 or 100 people per 100k person years and the first gets hundreds of times the money as the second. or whatever. The way we allocate funds for research is not very good from a “expected QUALY saved per dollar” type calculation.

        • Perhaps some of us think that paying absolute obeisance to the altar of ““expected QUALY saved per dollar” type calculation” is a bad idea?

          i.e. Feature, not bug?

        • In other words, I don’t want to be the guy telling the kid with a rare genetic disease, “Hey, there’s only a few hundred kids with this syndrome, so screw you, I have better use for my public-health-spending dollars”

          Sometimes, efficiency isn’t everything.

        • I’m fine with spreading the money around a bit, on the off chance we are mis-estimating things… but it seems to me like we should apportion funds within approximately an order of magnitude or so of estimated QUALY saved / dollar because for every “special” case that’s highly visible and sounds really terrible and gets 800 times as much funds on a QUALY/dollar basis, there’s necessarily a whole bunch of people who are suffering because their “less exciting” but more common or more likely to be cured illness is under-funded. It’s the availability bias of not “seeing” those suffering people that is the hidden damage caused by not making at least some effort to spend based on expected QUALY/dollar.

        • In practice, in the absence of a QALY/dollar type calculation I suspect the result is that more money is allocated to unusual disease affecting rich white people and less money is spent on things like preventing poor black kids in Baltimore from being poisoned by lead paint or solving problems associated to millions of terribly poor West African children or whatnot.

          When you put a concrete face on one side of the equation (say kids with rare genetic diseases like osteogenesis imperfecta or something that causes super brittle bones) and then leave the other side of the equation anonymous … it’s easy to ignore that anonymous side, but as soon as you put a face to that side, it’s hard to argue that somehow theoretically maybe helping a few hundred children with something rare is better than almost certainly helping a few tens of thousands of children with something common…

        • No, I’m not, I’m just saying that at first sight it seems people should evaluate things on this basis. Particularly people who are policy makers and have access to data and money decision making power.

  7. Here’s the graph I made from data I got directly from CDC Wonder data on suicides. I let CDC do the population rate adjustments, so I don’t know if they’re really correct, but it seems like they should at least know how to do that correctly.

    http://models.street-artists.org/2017/04/04/suicide-rates-in-the-us/

    Note that White rates are highest, and also trending upwards. So this isn’t even a matter of “we didn’t notice the White data because we were focusing on Black people who have a much higher rate” (of suicides). Black people may have a higher rate of overall deaths, but White people are the ones committing suicide by far (not only higher rate, but also a bigger population).

    Now I don’t break this down by age, only ages 15 to 65. I figured before age 15 people don’t much commit suicide, and after age 65 there are reasons like chronic untreatable pain or degenerative diseases or end-of-life choices which complicate the question. I figure if we prevent all of the suicides age 15 to 65 we are probably doing the world a lot of good, if we prevent suicides in sick people age 93… the question is more problematic, so that was the justification for my age range choice just to not have any question about whether prevention would be a good thing.

Leave a Reply to Rahul Cancel reply

Your email address will not be published. Required fields are marked *