In 2015, economists Anne Case and Angus Deaton grabbed public attention by reporting that overall death rates for middle-aged White Americans had risen since the 1990s, sharply breaking with prior trends; they counted hundreds of thousands of new “deaths of despair.” Much news coverage, contention, and controversy followed.[1] By 2020, there was enough confusion that some people accepted the surge of deaths as fact and some dismissed it as just another academic kerfuffle.
This is a good time to revisit the claim. Last year, Case and Deaton produced a best-selling popular treatment, Deaths of Despair and the Future of Capitalism, and late this year they released two academic updates (here and here). Last March, the National Academies of Sciences, Engineering, and Medicine published a draft “Consensus Study Report” on High and Rising Mortality Rates Among Working-Age Adults. Meanwhile, the tide of studies that followed the initial controversy just keeps rising. What do we know?
We know that about three generations ago, newborn girls could be expected, on average, to live to about 71, newborn boys to about 65.This graph shows what happened then.
Lifespans rose and then they didn’t. Because more Americans in their prime years were dying, life expectancies leveled off for women and turned down for men,[2] despite continuing advances in health care and medical treatments. What happened?
Here are key questions: Who has suffered the rise of premature death rates? What kinds of deaths account for this trend? What distinguishes the individuals who were most vulnerable? And what social causes might explain their vulnerability in this era?
The tl;dr version of this post is that: Yes, there has been an unusual surge in American deaths in the last two decades; it has been concentrated among working-age, working-class Whites; “deaths of despair” is too expansive a description and explanation; the surge was basically due to the opioid epidemic and rising obesity; the opioid epidemic was mostly due to increased supply; economic and social dislocations mark the communities that have been most vulnerable to the opioid epidemic and thus premature deaths.The surge is real; Big Pharma and cardio health are most responsible; “left behind” communities have been the most vulnerable.
Who Has Suffered the Death Surge?
Native-Born Americans. As this graph from the National Academies report, shows, average American life expectancy (red line) has in the last 70 years–especially in the last 20–fallen behind that of 16 “peer countries” (blue line).[3] In the 1950s and ’60s, U.S.and peer lifespans were comparable; then the gap really widened, starting in the 1990s for women and in the 2000s for men. One study concludes that “no Western European country experienced the increases in mortality reported in the United States,” not even among the lower classes.
To be more precise, the death surge has been concentrated among native-born Americans of native-born parents. A just-published study found that “had it not been for the foreign-born, [the U.S. stagnation in life expectancy] would have instead been a steep decline . . . .” Life expectancy for the native-born in 2017 would have been “knocked back to its 2003 levels, [with] male life expectancy . . .roughly that of modern-day Tunisia.”
Working-Age Americans. It is the death rates for Americans aged roughly 25 to 64 that stopped dropping or started rising. While death rates for children, youths, and seniors have dropped sharply since the turn of the century,[4] those in between youth and old age experienced no change or an increase in death rates. Moreover, it is precisely among the middle-aged that native-born Americans did so poorly compared to immigrants.
Whites. Death rates of Black Americans have been and still are substantially higher than those of Whites, which are, in turn, higher than those of Hispanics. But the issue here is what has changed. The graph below shows, by ethnicity and age, how death rates changed since 1990: It shows the ratio of death rates for three-year periods each compared to the base period of 1990-1993. (For simplicity, I focus on males and leave out the 55-to-64 year olds; trajectories for females and for the 55-to-64 are much the same.[5]) The flat blue lines—Whites—stand out from the dropping death rates for Blacks and Hispanics.
The focus, then, is on working-age Whites, not because they are at greater risk than Blacks, but because of the unexpected change of direction: On average, Whites’ risk of premature death stayed flat or even rose, while those for Blacks and Hispanics declined (see also here).
Less-Educated Whites. The evidence here is sparser and complex, but sufficient for the National Academies to conclude that there has been generally “widening disparities” in mortality by level of education—but only among Whites (p. 3-8).[6] The next graph shows the trends from 1990 to almost 2020 in terms of life expectancy, how many more years of life 25-to-75 year-old women could expect (the male story is similar).
Contrast the two lines labeled Whites. Those lines clearly diverge; the dashed one (having a BA degree) keeps adding years of life, the solid one (no B.A. degree) loses years of life. For Blacks, both educational groups ascend in rough parallel until the last decade. (Black women without a B.A. experienced a drop in life expectancy, a development which some argue is from the fentanyl epidemic.) This widening of the education gap is distinctive to the U.S.
Especially in Non-Metropolitan Areas. The next chart shows an example of the geographical pattern. In 1990-93, non-Hispanic White men aged 45 to 64 died at about the same rates whether they lived in big urban centers or outside metropolitan areas. By 2015-17, death rates for the most urban of these men had consistently dropped; death rates for the least urban had dropped but then risen.
In a similar fashion, death rates across the states have widened significantly, mainly for the less-educated, with states in the South and Midwest showing notable increases mortality among the less-educated. Also, richer states have in the last 40 years notably reduced their death rates, poorer states not as much.
Other distinctions also identify whose risk of death rose, for example, unmarried white men. Other attributes do not. Notably, men and women have been affected about equally, a point to which I will return. The basic risk factors for succumbing to early death over the last 20 and especially the last 10 years are clear; it is being American, between 25 and 65 years old, non-Hispanic White, native-born, not a college graduate, and living in a more rural place. These are the people who have most centrally contributed to the death surge.
What are These People Dying Of? The Medical Causes
In trying to identify the immediate medical causes of the “surplus” deaths, the National Academies “>report focused on increases in drug overdoses, alcohol-related afflictions, and suicide starting in the 1990s, and also on trends in “cardio-metabolic” diseases (more hypertension, slowing declines in heart disease) during the 2010s. Of all these, the major driver has clearly been:
Opioid Addiction. One 2021 study put it plainly: “Without the opioid epidemic, American life expectancy would not have declined prior to 2020,” prior to Covid-19 (see especially fig. 1). America stands out in drug deaths. And drug deaths stand out statistically in accounting for changes in overall death rates.[7] The scale of the drug trauma and its predominant role is evident in several ways.
For one, opioid addiction skyrocketed; deaths followed. When some limits were imposed on opioid prescriptions, many users turned to (less expensive) heroin and then to fentanyl. More deaths followed. The graphs below from the National Academies report (p. 7-2,3) illustrate the trend with just a couple of the gender-age groups. Deaths by “drug poisoning” per 100,000 White men aged 25 to 44 rose about six-fold and per 100,000 White women aged 55 to 64 rose more than four-fold between the early 1990s and later 2010s. The increases among non-Hispanic Whites in drug deaths were much steeper, more consistent, and more widespread than were any upswings in alcohol deaths or in suicides.
More evidence: The drug epidemic was so powerful that it spread to other groups too, notably non-Hispanic Blacks. Although total death rates for Blacks had dropped steadily over many years, drug-induced deaths among Blacks just increased.[8] Growing addiction and overdosing help explain why the U.S. fell farther and farther behind peer countries in life expectancy. Before 2000, drug death rates were not especially high in the U.S. (they were highest in the Nordic countries), but now the U.S. has a far higher death rate from drugs than our peers.[9] Similarly, drug deaths account for the widening mortality gap between college graduates and non-graduates much more than alcohol deaths or suicide deaths do. Broader drug dependency has also skyrocketed; many more Americans reported to the Gallup Poll that drug use has caused trouble in their families—from 19% in 1995 to 30% in 2018 (see, also, Pew). Deaths by overdose kept growing even into the Covid years (while deaths by suicide declined).
Even so, drug abuse’s role in the death surge may be underestimated, because it contributes to deaths from disease, accident, and suicide. One complex estimate is that, in total, actual “drug use has left an imprint on US mortality that is roughly double that implied by drug-coded deaths [i.e., official counts] alone.”
A final note about the centrality of drug abuse: Over the last century, rates of death by suicide and by alcohol abuse have moved up and down, such that in 2017 they were at about the same level that they were in 1917 (an era of general under-counting). In contrast, overdose deaths rose at an unprecedented pace in the 1990s and now are about five to ten times higher than they were in 1917 (see Figures 1-3 here).
Alcohol a Bit, Perhaps. Alcohol-related deaths increased in the last couple of decades, only among Whites, and mostly among the less-educated (pp. 7-13ff in the National Academies report). The patterns are similar to drug overdosing, which may in part be simply because many victims abused both. However, the increase in deaths by alcohol was much less dramatic and its contribution to the upsurge in deaths far smaller than that of drug abuse (see endnote 7).
Suicide a Bit, Perhaps. Suicide rates rose in recent decades, mostly for Whites, especially White men. The next graph shows the trends from 1950 on that are worth noting. (I left out Hispanics and Black Women because they had very low and pretty flat suicide rates throughout.) There is a lot going on in the graph, but the key take-aways are: (1) Suicide rates have fluctuated up and down over the last 70 years; only among White men, 25-44 years old, has there been a long-term increase; (2) suicide rates did go up for almost all of these groups after 2000.
Suicide rates rose more in smaller communities (National Academies, Fig. 8-2) and among those with less education. These patterns parallel those of drug overdoses. (Indeed, Case and Deaton consider overdosing as a sort of slow suicide.) That parallel, along with an intuition about the causes of each, encourages lumping suicides together with drug overdoses as “deaths of despair.” However, today’s rates of drug overdose deaths are unprecedented while the suicide rates are not. Moreover, analyses of how much each contributed to the spike in overall middle-age White mortality (see endnote 7) show that the suicide trends contributed far less than did drug overdoses.
Heart Disease, a Lot. A key but oft-ignored contributor to the slowdown in extending American life expectancy is the slowdown in controlling heart disease, especially hardening of the arteries. While deaths from heart disease have long been in decline, that trend flattened out in the 2010s, especially for males and especially for Whites.[10] Were it not for this unusual development–one not experienced by peer countries–there would not have been an overall drop in life expectancy in the last decade. A trio of researchers asserted in 2020 that “a stagnating decline in cardiovascular disease (CVD) mortality was the main culprit, outpacing and overshadowing the effects of all other causes” in the U.S. life expectancy trend stalling after 2010. The next graph is their key evidence. The red line represents the U.S.; the black line represents the total for comparable countries. The brackets that I inserted show that in 2000, the American death rate for cardiovascular disease was lower than that of ten other leading nations; by 2015, it was higher.
Thus, separate from and greater than the “deaths of despair” trio, the overall blow to growing life expectancy in the U.S. resulted from a failure, especially since 2010, to make as much progress against heart disease as other nations have made. The profile of who actually suffered increases in heart-disease deaths is roughly similar to those discussed before: disproportionately White and living in non-urban places (National Academies, figs. 9-2, 9-5).
To wrap up this section, Case and Deaton pool drug, alcohol, and suicide deaths as “deaths of despair” and seek a common explanation for increases in the three. Other scholars demur, arguing that “suicide and alcohol-related causes contribut[ed] negligibly to . . .the increases for most working-age groups” and that trends for the three differ in timing, location, or key population (National Academies, p. 7-36). If we want to understand why White, Non-Hispanic, native-born, 25-to-64-year-old, heavily rural Americans started dying at unexpectedly high rates—the folks we saw earlier who are distinctively at risk—the focus should be on drug abuse and cardio health.
What are the Personal Causes of the Medical Causes of Increased Deaths?
The Cardio Slowdown: Probably Obesity. The number of Americans who could have been saved from mortal circulatory disease but weren’t mounted up after 2010 but, given the nature of that disease, the origins were probably many years earlier. A stunning rise in obesity–more than tripling over the last 60 years and still increasing recently–was likely the major one, especially among White men, whose obesity rates doubled in just the last 20 years,[11] and among the less educated. Although rising obesity is a global phenomenon, the U.S. has easily outpaced western European countries in getting heavy (and, no doubt, fallen behind in health care available to deal with that). The profile of the newly obese American is similar in some ways to that of drug use (the two can be conflated), but significantly different in others (e.g., the growing obesity rates of Black Americans).
Drug Abuse: Demand?. Some researchers argue that the major cause of the drug epidemic was the vast increase in the supply; I address that below. Here, I review the central discussion in the booming research on the death surge: What has been driving more Americans to demand opioids and their addictive successors? The usual answer, Case and Deaton’s answer, is the desire to ease growing distress, to pursue “euphoria or numbness” (p. 122 here). What sort of distress could that be (and did it grow)?
Physical Pain. This is the distress that opioids were meant to ease and that many doctors thought had been woefully under-treated. Reports of pain have increased. In one study, for example, 13% of men aged 55 to 61 said in 1992 that they had severe or limiting pain, 22% said so in 2016; for women, such reports grew from 18% to 28%. Those experiencing more pain are likelier to overuse drugs. But the correlation between pain and the death surge is not tight. True, the least educated Americans reported the most pain increase, but so did the oldest Americans and seniors experienced neither increasing drug abuse nor a mortality surge. Furthermore, the scale of the pain increase does not match the much greater increase in drug overdoses. Finally, the pain account begs yet another question, Why the pain increase? Obesity, in part. Greater work demands? Worse self-care? Psychosomatic responses?
Psychological Pain. Some studies suggest that over the last 20 years, increasingly more Americans—especially the kind of Americans highlighted earlier—reported psychological distress. In one, the percentage of Americans who said that their mental health had been “not good” in every day in the previous 30 days rose from about 4 percent in 1993 to 6 percent in 2019, rising much more among Whites 35-44 without any college experience (see also here). The percentage of White respondents to the General Social Survey who said that they were not very happy rose between the early 1990s and the late 2010s from about 9 to about 15 percent for non-graduates, while it stayed unchanged at about 7 percent for college graduates.[12] But some of the data are not consistent with the picture of growing psychological pain and some researchers question whether the scale of the change in emotions suffices to explain the drug abuse boom.
Social Disconnection. Increasingly many less-educated Whites , especially men, seem disconnected–outside the labor force, outside of marriage, and outside of church.[13] A few well-noted books have described these people as in crisis, including books by liberal political scientist Robert Putnam and conservative political scientist Charles Murray. Substance abuse seems to come with this disconnection, as consequence, cause, or just correlate.
Economic Dislocation. The most researched form of distress that might spur drug abuse is that produced by employment and financial problems. However sensible this explanation might be, the evidence connecting individual economic setbacks to drug abuse and mortality is mixed.[14] (Update: It is stronger for suicide.) In recent writings (e.g., here and here), Case and Deaton have stressed that they are not claiming that there is a direct and immediate causal effect of economic problems on drug abuse or mortality, but, instead are claiming that long-term community decline is to blame, that decline being due to economic setbacks (see below).
Connecting Distress to Drug Demand and Mortality. How might these personal conditions–physical pain, psychological pain, social disconnection, and economic dislocation–lead to drug abuse and early death? That intolerable physical pain can lead to pain-killer use and more addictions makes sense. But, claiming that the other distresses cause drug use (or obesity) typically assumes that distressed people use drugs to escape emotional pain or to self-medicate, to “numbness” in Case and Deaton’s terms. Other explanations for the correlation should, however, be acknowledged. Along the lines of Murray, for example, cultural changes in society may have induced certain people–notably, young working-class men–into self-destructive behavior, say, into playing video-games and using recreational drugs rather than working. Their economic woes may have followed the drugs rather than preceded them (although the research seems dubious).
All this accumulating scholarship and a just-published analysis by Berkeley’s David Cutler and Harvard’s Edward Glaeser [15] yield the following reasonable conclusion: that these distresses can help identify which Americans—disproportionately the middle-aged, less-educated, pain-ridden, economically-stressed, socially-disconnected, rural Whites—were likeliest to become addicted to prescribed opioids (and their later drug replacements), but that these personal factors cannot satisfactorily explain the surge in drug deaths from the 1990s to the 2010s. They cannot do so, in part, because, aside from reports of physical pain, the trends for these factors (e.g., measures of depression) are mixed, and in any case too modest in scale to account for the explosive growth in drug deaths. The surge in addiction and death is mostly not about a rise in demand, but about a tsunami of supply. (And: let’s not forget obesity in the bigger picture.)
What are the Social Causes of the Personal Causes of the Medical Causes of Increased Deaths?
Drug Abuse: Supply. Princeton health scholar Janet Currie and colleagues charged this year that “The Opioid Epidemic Was Not Caused by Economic Distress,” but by the aggressive marketing of pain-killers, especially of Oxycontin by Purdue Pharma, distributed by unscrupulous providers who prescribed boatloads of pills, and aided by a sincere health movement to better treat pain (see also here and here). [Update: A new study “suggests that the introduction of OxyContin explains 79% of the rise in the overdose death rate since 1996.”] Stanford demographer Matt Kiang fills out this story by describing how White users took up prescribed opioids from the mid-1990s to 2010, but as authorities put some limits on prescriptions, drug dealers showed up with heroin and then fentanyl. The drug part of the death surge is all about supply, about how America’s lax pharmaceutical system put us at the top of the drug death rankings. (Some would suggest that rising obesity is also story about greater supplies of cheap calories, but I’ll leave that aside.)
Still, many analysts question the supply account as insufficient: Why were particular groups and communities especially susceptible to drug addiction when the opioid tap turned on? Why, for example, Case and Deaton ask, did the opioid epidemic not infect the college graduates?
Community Vulnerability. Neither personal economic setbacks nor short-term economic developments seem to satisfactorily explain who succumbed. Case and Deaton and others argue that long-term community unraveling set many middle-aged, less-educated Whites adrift and thus made them vulnerable. Researchers have found that low levels of “social capital” in a locality–weakened associations, civic life, churches, etc.–are associated with higher local drug use and mortality (e.g., here). Whether those adrift are also in “despair” or not, it is classic sociology to view the detached as vulnerable to engaging in “deviant” behavior (classically including suicide).
Focusing on the community is an important step. Even if particular individuals are not pushed to drug abuse by personal economic setbacks or even by family crises, they are subject to the deterioration of the community institutions around them (especially so, perhaps, in rural places). Unemployed friends can induce an employed person to join them in recreational drugs; the shuttering of local entertainments can leave young men too idle. This perspective may also explain why there is no overall gender difference in the death surge. If local men become disconnected, if they increasingly fail as partners and fathers, local women face discouraging prospects for companionship and co-parenting, a disconnection even if the women are otherwise doing well. (Community-level factors might also explain why rural Americans are more obese than urban ones, a difference that has grown slightly). And, given that Americans are moving much less often, are staying closer to home, community conditions are that much more consequential to individuals.
And so the death surge was–is?–real, not because of a surge in despair, but because of two events: the flooding of the country with opioids and the stalling of progress against heart disease, probably because of greater obesity. The communities most damaged were, unsurprisingly, disadvantaged ones, but in this case the non-urban ones, inhabited largely by less-educated Whites, places that economic trends have left behind.
UPDATE, August 1, 2022:
Lawrence King, Gábor Scheiring, and Elias Nosrati published an article in the Annual Review of Sociology re-making the case for the economic distress explanation of opioid addiction. In particular, they draw parallels with a surge of middle-age deaths in Eastern Europe in the 1990s during the collapse of the USSR and of state socialism and the introduction of capitalism.
UPDATE, December 29, 2022:
American life expectancy continued to fall in 2021, across a wider range of ages, with major causes being Covid and fentanyl-fueled drug overdoses.
UPDATE, December 30, 2022:
A study published in The Quarterly Journal of Economics by Abby Alpert and colleagues reinforces the supply-side explanation for opioid overdoses. As the abstract describes, the authors compare what happened in those states that made OxyContin prescription in 1996 more cumbersome than other states did. Although the states with the greater hurdle originally had higher rates of addiction, they ended up with much lower addiction rates and deaths. The “results show,” write the authors, “that the introduction and marketing of OxyContin explain a substantial share of overdose deaths over the past two decades.” And a study published in the American Economic Review by Aljoscha Janssen and Xuan Zhang found that independent pharmacists were much more likely to provide OxyContin than were chain pharmacies, presumably because of financial pressures or incentives to do so.
NOTES———————————————————–
[1] Examples of coverage: N.Y. Times and New Yorker; of controversy: Vox; Gelman, Harris, and Smith.
[2] National Academies report, Figure 2-1.
[3] This is also from the National Academies Report, Figure 2-1. “Peer countries” are basically western European nations plus Australia, Canada, and Japan.
[4] Between 1999/2000 and 2015/16, down 15% for under-25 Americans and by 20% for those 65-plus (calculated from CDC WONDER Online Database, Compressed Mortality File 1999-2016 Series 20 No. 2U, 2016, accessed at http://wonder.cdc.gov/cmf-icd10.html).
[5] Blacks and Hispanics experienced a continuing decline in death rates, but Whites experienced a decline and then a rise after 2011.
[6] There is a technical issue in comparing educational groups over time, that the least-educated are a shrinking group, an increasingly disadvantaged one, but the conclusion stands nonetheless. (Geronimus et al use a technique to get around that problem; see also here.)
[7] Table 4-1 of the National Academies report decomposes the all-cause mortality, by race and age, into its many components and provides estimates of how changes in each specific cause contributed to the overall change in mortality rates, in deaths per 100,000, between 1990-93 and 2015-17. Here is an attribution of how much of the total mortality increases were due to which factors—for non-Hispanic Whites:
Male 25-44 Male 45-54 Male 55-64 // Fem 25-44 Fem 45-54 Fem 55-64
Total Change + 5 - 5 -19 +37 + 7 -14
Percent of any upward shift due to.....
Drugs 74% 38% 18% 57% 36% 20%
Alcohol 3 7 7 5 9 6
Suicide 13 11 5 7 4 3
The major contributors to reducing overall mortality were declines in cancer, “Ischemic & Other Circulatory System,” and HIV-AIDS.deaths.
[8] Table 4-2 in the National Academies Report. [Update, 2/18/22: Since the studies used here, the rates of overdoses for Blacks and Latinos have soared. The reason has less to do with rates of addiction and more to do with the introduction of highly fatal fentanyl and similar synthetic compounds into the doses being sold on the street.]
[9] A chart from Ho (2019):
[10] From 1980 to 2009, the age-adjusted death rate for ischemic heart disease among males (Whites) declined 10.0 (7.6) deaths per 100,000 each year, but from 2010 to 2018, it slowed to a 3.3 (2.7) rate, a more drastic slowdown than for females, Blacks, or Hispanics (calculated from https://www.cdc.gov/nchs/hus/contents2019.htm#Table-005).
[11] Obesity rates for non-Hispanic White men rose from 20% in the 1988-1994 years to 41% in the 2015-18 years; for women, 23% to 39%. There were comparable roughly 20-point increases for Blacks. Data for Hispanics starts in 2007-10, but the trends are similar. I took these numbers from https://www.cdc.gov/nchs/hus/contents2019.htm#Table-026.
[12] My analysis of the GSS. See also here.
[13] Many have written about these changes. On work (and other connections), see, e.g., here and here. On marriage, see also here. On religion, the graph below, which I calculated from the GSS, illustrates how less-educated Whites became church non-attenders at a higher pace than did white college graduates:
[14] Scores, perhaps hundreds, of studies emerged in the several years since the original Case and Deaton article, all trying to figure out the connections among economics, drugs, and death. A number of studies seem to show a pretty straightforward connection–e.g., an auto plant closing in a county tended to be followed by an increase in opioid mortality. Others are less clear–e.g., changes in states’ minimum wages seemed not to influence drug deaths; ; and drug abuse seems to precede being outside the labor force rather than to follow it. The National Academies report states (p.S-6): “[T]he evidence is mixed on the causal effect of relatively short-term economic changes on substance-related mortality. . . . The best interpretation of current knowledge about the broader relationship between economic well-being and mortality suggests that economic hardship is associated with higher mortality, especially in the context of widespread availability of potent and life-threatening medications. However, the overall impact of the direct economic shocks that have been examined (i.e., short-term changes in economic circumstances) appears to be modest. (P. S-6).” For example, it is hard to connect the business cycle to drug deaths. [Update: Indeed, in the 2010s, the white working-class overall did relatively well economically.] But, later (p. 10-2), the report says that a long-term explanation based on cumulative, community-level economic disadvantage is more supported.
[15] See this article and its online supplement.
MADE IN AMERICA 