Skip to main content

Long-Term Trends in Deaths of Despair

PDF version

Anne Case and Angus Deaton famously chronicled a dramatic rise among middle-aged non-Hispanic whites since 1999 in “deaths of despair”—deaths by suicide, drug and alcohol poisoning, and alcoholic liver disease and cirrhosis.1 The Social Capital Project has extended Case and Deaton’s research to cover the full American population as far back as available data permit: to 1900 in some cases, and to 1959 or 1968 in others. We present here a snapshot of the long-term trends in deaths of despair. We also attach our full dataset for use in future research, including results broken down by age, sex, and race.

Mortality from deaths of despair far surpasses anything seen in America since the dawn of the 20th century. (The trend for middle-aged whites reveals a more dramatic rise but only goes back continuously to 1959.) The recent increase has primarily been driven by an unprecedented epidemic of drug overdoses, but even excluding those deaths, the combined mortality rate from suicides and alcohol-related deaths is higher than at any point in more than 100 years. Suicides have not been so common since 1938, and one has to go back to the 1910s to find mortality from alcohol-related deaths as high as today’s.

At the same time, a long-term perspective reveals that while drug-related deaths have been rising since the late 1950s, the current increase in suicide and alcohol-related deaths began only around 2000, as the opioid crisis ramped up. Suicide and alcohol-related mortality trends track each other well over the past 45 years, and after accounting for the changing age distribution of the US, combined deaths from the two causes were as common in the mid-1970s as today.

Self-reported unhappiness probably has been on the rise since around 1990 (though not all sources agree). That predates the increase in deaths of “despair” by a decade. Moreover, unhappiness likely fell over the 25 years preceding 1990, while deaths of despair rose and then plateaued. And one data source suggests stable levels of unhappiness over the long run.

Rising unhappiness may have increased the demand for ways to numb or end despair, such that the cumulative effects show up years later in the form of higher death rates. But the proliferation of a uniquely addictive and deadly class of drugs has meant that the supply of despair relief has become more prevalent and more lethal, which would have increased mortality even absent an increase in despair. Given the lack of correspondence between trends in economic and social indicators, unhappiness, loneliness, and deaths of “despair,” it may be more productive for policymakers to focus on the overdose epidemic than on despair per se.

 

Definitions and Methods

All of our estimates are from data publicly available from the Centers for Disease Control and Prevention (CDC). In the analyses below, we modify the Case-Deaton definition of “deaths of despair” in several ways. (Our data file includes trends using their definition as well.) Alcohol-related mortality, in our analyses, includes only those liver disease deaths deemed to be from alcohol abuse. But unlike Case and Deaton we add in deaths from a number of other diseases not associated with the liver that are attributed to alcohol abuse, as well as deaths from mental health disorders attributed to alcohol dependency. We also categorize deaths from alcohol poisoning under alcohol-related deaths, rather than lumping them in with drug overdoses as Case-Deaton prefer. In our analyses, drug-related deaths include those overdose deaths not deemed suicides, as in the Case and Deaton research, but also deaths from mental health disorders attributed to drug addiction. In those of our estimates using modern-day data, we exclude deaths due to drugs administered in medical or surgical care (which are included in the Case-Deaton definition). Our suicide definition matches that used by Case and Deaton; it includes alcohol- and drug-related deaths deemed suicides. The methodological appendix provides additional detail, and detail is also provided in our data file.

The increase in deaths of despair has been so large among non-Hispanic whites between the ages of 45 and 54 that it has caused overall mortality in this group to rise since 1999. For this reason, Case and Deaton devote special attention to the group. We display trends for the overall population and for non-Hispanic whites in this midlife age range. (Prior to 1999, Hispanic whites and non-Hispanic whites cannot be separated, so we include all whites together. Our checks indicate this has a minimal impact on the trends and levels reported here.)2

We also show the component trends for men and women separately. Our dataset provides trends broken down for other age ranges and racial groups.

 

Deaths of Despair

Figure 1 displays trends in crude death rates going back to 1900.3 Mortality from deaths of despair fell dramatically between 1907 and 1920, rose during the 1920s, and reversed that increase during the 1930s and early 1940s. Deaths of despair then rose from the mid-1950s to the mid-1970s and stabilized before skyrocketing after 2000. In 2000, there were 22.7 deaths of despair per 100,000 Americans—not that different from the 1970 rate of 21.5. By 2017, the rate had doubled to 45.8 per 100,000.

 

Figure 1. Deaths of Despair and Its Components, 1900-2017, Crude Rates

Source: Social Capital Project analyses of CDC data. For details, see the appendix.

 

We estimate the previous historical high (33.6) to have occurred in 1907, a level surpassed in 2013. A full explanation for the patterns in Figure 1 is beyond the scope of this brief, but notable historical events that might explain some of the changes over time include the Pure Food and Drug Act (1906), the Panic of 1907 (1907-08), the Harrison Narcotics Tax Act (1914), World War I (1914-18, with US involvement from 1917-18), the flu epidemic of 1918 (1918-19), the deep Depression of 1920-21, Prohibition (1920-33), the Great Depression (1929-38), World War II (1939-45, with US involvement primarily 1942-45), and the counterculture revolution of the 1960s. We return to some of these events in the discussion of subcomponent trends below.

Figure 2 provides trends using age-adjusted mortality rates. The CDC has estimated rates that hold constant 11 age groups at their 2000 shares of the population, so that the changes in rates over time are unaffected by whether older or younger people are becoming more or less prevalent. The long-term patterns for deaths of despair are similar to those for the crude rates, but the estimates are available only back to 1959. Age-adjusted suicide rates go all the way back to 1900, and they indicate higher death rates than the crude rates early in the 20th century. This suggests that if the early-20th-century population had been as old as the 2000 population, the overall crude suicide rate would have been higher (as well as, in all likelihood, the crude rates for drug- and alcohol-related deaths). It is unclear that age-adjusted comparisons over such a long period are better than the crude comparisons, however; people live longer in 2000 than in 1900 because life is materially better and easier, so imposing that age distribution on the 1900 population is a somewhat artificial exercise. Nevertheless, it is likely that age-adjusted deaths of despair rates for the early 20th century would be higher than the crude rates shown in Figure 1 for the same period.

 

Figure 2. Deaths of Despair and Its Components, 1900-2017, Age-Adjusted Rates

Source: Social Capital Project analyses of CDC data. For details, see the appendix.

 

Age-adjustment makes more sense, however, when comparing more recent years. Figure 2 suggests that after controlling for changes in aging, suicide rates have not changed much over the past 50 years. The rate in 1959 was 12.3 per 100,000, compared with 14.0 in 2017. Both suicides and alcohol-related deaths were as common in the mid-1970s as in 2017; the combined death rate from both was 23.5 per 100,000 in 1975 and 23.6 in 2017.

Figure 3 shows the age-adjusted trend since 1959 for whites between the ages of 45 and 54 (non-Hispanic whites from 1998 forward). Among this group, the 1975 peak was followed by a large drop in deaths of despair, so that the 1988 rate was the lowest on record. Soon thereafter, the situation deteriorated dramatically. From that low of 32.6 deaths per 100,000, the rate rose to 48.5 in 2002 (exceeding the 1975 peak) and to 91.6 in 2017.

 

Figure 3. Deaths of Despair and Its Components, 1914-2017, Crude Rates, Non-Hispanic Whites Ages 45-54

Source: Social Capital Project analyses of CDC data. For details, see the appendix. Prior to 1999, the trend includes Hispanic whites.

 

Suicide

The suicide rate has risen steadily since the early 2000s, reaching 14.5 per 100,000 in 2017 (Figure 1). That was about the average between 1910 and 1919, and the average over 1930-39 was higher. Suicides actually became steadily rarer from 1986 to 2000 (from 1977 using age-adjusted rates), but the 2017 crude rate was the highest since 1938.

Suicides spiked with the onset of the Great Depression, but they were rising steadily throughout the 1920s. The declines after 1915 and 1938 are partly attributable to World Wars I and II. These drops do not so much reflect the substitution of war-related deaths for suicides: suicide fell among women during these periods too, and the declines began before Americans entered the conflicts. Rather, as Emile Durkheim first posited, the likely explanation is that wars promote social integration, which reduces despair.4 The Panic of 1907 may also have caused a spike in suicides, but there too the increase had begun years earlier. The influenza epidemic of 1918 substituted flu deaths for some suicides, lowering the suicide rate.

From 1904 to 1940, age-adjusted rates of suicide were above 15 per 100,000 in every year except 1920. They have never reached that that level since (Figure 2). The trend for middle-aged non-Hispanic whites has been similar, though rates have consistently been higher than for the general population, and the recent rise since 1999 has been steeper (Figure 3). The suicide rate for middle-aged non-Hispanic white women has approached its all-time high (Figure 4).

 

Figure 4. Crude Suicide Rates by Sex, Overall and Non-Hispanic Whites Ages 45-54, 1900-2017

Source: Social Capital Project analyses of CDC data. For details, see the appendix. Prior to 1999, the “non-Hispanic white” trend includes Hispanic whites.

 

Alcohol-Related Deaths

In 2017, there were 11.0 deaths related to alcohol per 100,000 Americans (Figure 1). That was higher than at any time since the start of World War I. These deaths had been declining from the mid-1970s to the early 2000s, following a pattern similar to suicides.

Figure 1 reveals that most of the large drop in deaths of despair in the years before 1920 was due to a decline in alcohol-related deaths. Prior to that drop, alcohol-related mortality was higher than it is today, but by 1920 it had fallen from its 1907 high of 15.1 deaths per 100,000 to just 1.1—lower than the number of deaths from drugs. This decline preceded Prohibition at the national level, and alcohol-related deaths actually rose through much of Prohibition. A number of states had enacted their own prohibition laws prior to 1920, but they tended to be rural, and the impact seems to have been too small to have affected national figures much.5 Furthermore, suicide death rates follow a similar trajectory between 1907 and 1920.

Most likely, World War I and the flu epidemic were the biggest factors behind the drop. Since many alcohol-related deaths reflect an accumulation of years of alcohol abuse, however, it is possible that Prohibition dampened growth in alcohol-related deaths in subsequent decades.6 In the 1930s and 1940s, alcohol-related deaths were much further below their pre-1920 high than were suicide deaths, even though alcohol consumption had risen nearly back to its old high by the mid-1940s.7 Alcohol consumption hit a new peak in 1980. That increase may account for much of the rise in alcohol-related deaths between the mid-1940s and the mid-1970s, during which time suicides rose much more slowly. Alcohol has become steadily more affordable since at least 1950, though consumption fell significantly after 1980.8

As shown in Figure 2, after age-adjusting, alcohol-related deaths actually peaked in 1974 at 10.2 per 100,000 (still lower than the suicide death rate in any year on record). The number of alcohol-related deaths for middle-aged non-Hispanic whites (Figure 3) was comparable to the number from suicide through much of the 1970s, but today’s rate of 24.3 per 100,000 is the highest on record. Among middle-aged non-Hispanic whites, women exceeded their previous high in 2011, while men did not exceed their previous high until 2016 (Figure 5).

 

Figure 5. Crude Alcohol-Related Death Rates by Sex, Overall and Non-Hispanic Whites Ages 45-54, 1959-2017

Source: Social Capital Project analyses of CDC data. For details, see the appendix. Prior to 1999, the “non-Hispanic white” trend includes Hispanic whites.

 

Drug-Related Deaths

Drug-related deaths have been rising at an accelerating rate since the late 1950s (Figure 1). The increase has been especially sharp over the past 20 years. This long-run increase was preceded by a long-run decline dating back at least to the early 1900s. Cocaine and heroin use increased dramatically during the late 19th and early 20th centuries, and they (and morphine) became controlled substances only in 1914.9 The rise in drug overdose deaths likely was boosted by the countercultural revolution of the 1960s, when illegal drug use increased.10 The 1980s saw the crack cocaine epidemic. And then came the opioids crisis.

The proliferation of opioid deaths was initially a result of oversupply and abuse of legal prescription narcotics. However, as awareness of the dangers of misuse grew, policy changes restricted the supply and form of prescribed opioids. The crisis then shifted toward illegal drugs—first heroin and then more lethal synthetic opioids like fentanyl.11 On an age-adjusted basis, drug-related deaths rose by over 20 percent in 2002 and by nearly 25 percent in 2016. The overdose crisis is following a very different trend than those for other “deaths of despair.”

Similar numbers of men and women died of drug overdoses when such deaths were rare. But male overdose deaths have been rising faster and are now over twice as common as female overdose deaths (Figure 6).

 

Figure 6. Crude Drug-Related Death Rates by Sex, Overall and Non-Hispanic Whites Ages 45-54, 1959-2017

Source: Social Capital Project analyses of CDC data. For details, see the appendix. Prior to 1999, the “non-Hispanic white” trend includes Hispanic whites.

 

Rising Despair?

In 2000, the age-adjusted mortality rate from deaths of despair was at the same level as the previous low in 1983, and only slightly higher than in 1968. On an age-adjusted basis, combined mortality from suicide and alcohol-related deaths were the lowest on record, going back to 1968. It is no wonder that no one spoke of “deaths of despair” at that time.

The age-adjusted mortality rate from drug-related causes in 2000 was 5.2 per 100,000, having doubled in ten years, over which time suicide and alcohol-related deaths were falling. By 2007, drug-related deaths had doubled again. By 2017, the rate had nearly doubled again, standing at 20.5 per 100,000.

As noted, even absent this dramatic acceleration in drug-related deaths, “deaths of despair” would be higher than at any point in the past one hundred years. But on an age-adjusted basis, deaths of despair other than drug-related deaths was essentially the same in 2017 as in 1975 (Figure 7).

 

Figure 7. Drug-Related Deaths versus Other Deaths of Despair, 1959-2017, Age-Adjusted Death Rates

Source: Social Capital Project analyses of CDC data. For details, see the appendix.

 

Figures 1 and 2 showed that alcohol-related deaths and suicides track each other well over the past 45 years, but the same is not true of drug-related deaths. One possible explanation of the patterns discussed in this study is that while the suicide and alcohol-related mortality trends primarily reflect a “demand-side” problem—a desire to numb or end despair—the drug-related mortality trend also incorporates a “supply-side” problem. That is, the rise in drug overdoses not only reflects an increase in despair, but a change in the supply, addictiveness, and lethality of drugs that may be taken to numb despair.

Even the trends in suicide and alcohol-related deaths however, presumably reflect factors other than changes in despair. Figure 8 shows trends in self-reported unhappiness from four sources.12 Together, three of the four tell a consistent story of falling and then rising unhappiness. But while deaths of despair rose between 1965 and 1975 and then leveled off over the next 15 years, unhappiness fell over the period (with a temporary increase during the double-dip recession of the early 1980s). Unhappiness then rose, but the upward march of deaths of despair began only with a ten-year delay, starting in 2000. Furthermore, if the Gallup Organization trend in unhappiness is correct, unhappiness was flat to declining even in the 1990s and 2000s. Previous research by the Social Capital Project has found little evidence that loneliness has changed much over the long run.13

 

Figure 8. Trends in Unhappiness (Percent “Not Too Happy”), 1963-2018

Source: Social Capital Project analyses of public opinion data. For details, see footnote 12 and the appendix.

 

Finally, it is worth emphasizing how challenging the trends in this paper are for theories that explain rising “despair” by referring to either economic trends or social capital trends. It is very difficult to find such trends that improve over the 1970s and 1980s, then worsen after either 1990 or 2000.14 Case and Deaton have drawn attention to an important public health phenomenon, but we have far to go before understanding its implications for public policy and the health of our economic, community, and family life. In the meantime, apart from the question of whether or why despair may be on the rise, we clearly remain in the grip of a national opioid crisis that requires the attention of policymakers.

 

PDF version



1 Anne Case and Angus Deaton (2017). “Mortality and Morbidity in the 21st Century.” Brookings Papers on Economic Activity. Spring 2017. https://www.brookings.edu/wp-content/uploads/2017/08/casetextsp17bpea.pdf.

2 We looked separately at trends for all whites (including Hispanics) from 1999 to 2017. The difference in levels in 1999 is fairly small, and the rise thereafter is only somewhat less steep than when Hispanics are excluded.

3 Mortality data are available for the entire continental United States beginning in 1933, and data for Alaska and Hawaii are included beginning in 1959 and 1960, respectively. Our analyses indicate that the changing number of states included in the data does not meaningfully affect the long-term trends we estimate. For instance, in 1933, the overall crude death rate for the lower 48 states was 1,068 per 100,000, while it was 1,122 per 100,000 for the 11 states included in the data in both 1900 and 1933. Similarly, the death rates from suicide, alcoholism, and cirrhosis of the liver were 16, 3, and 7 per 100,000 for the lower 48 states and 18, 4, and 9 per 100,000 for the original 11 states. (Sources: https://www.cdc.gov/nchs/data/vsus/vsrates1900_40.pdf,

https://www.cdc.gov/nchs/data/vsushistorical/mortstatsh_1900-1904.pdf, https://www.cdc.gov/nchs/data/vsushistorical/mortstatsh_1933.pdf.) The 1962 and 1963 rates for whites and by race exclude New Jersey.

4 Emile Durkheim (1897). Le suicide. (Paris: Felix Alcan). For an analysis of suicide and the two world wars, see David Lester (1994). “Suicide Rates before, during, and after the World Wars.” European Psychiatry 9: 262-64.

5 Angela K. Dills and Jeffrey K. Miron (2004). “Alcohol Prohibition and Cirrhosis.” American Law and Economics Review 6(2): 285-318.

6 Ibid.

7 For alcohol consumption trends, see Sarah P. Haughwout, Robin A. LaVallee, and I-Jen P. Castle (2015). “Apparent Per Capita Alcohol Consumption: National, State, and Regional Trends, 1977-2013.” National Institute on Alcohol Abuse and Alcoholism, Surveillance Report #102, Table 1. https://pubs.niaaa.nih.gov/publications/surveillance102/CONS13.pdf. Originally accessed at Hannah Ritchie and Max Roser (2018). “Alcohol Consumption.” Our World in Data website, “Alcohol consumption in the United States” section. https://ourworldindata.org/alcohol-consumption.

8 Alcohol affordability trends are based on Social Capital Project estimates comparing the Consumer Price Index (CPI-U) for alcoholic beverages (available beginning in 1953 from the Bureau of Labor Statistics at https://www.bls.gov/cpi/data.htm) to nominal per capita disposable personal income (from the Bureau of Economic Analysis, National Income and Product Accounts Table 2.1, at https://apps.bea.gov/iTable/index_nipa.cfm). See also William C. Kerr, Deidre Paterson, Thomas K. Greenfield, Alison Snow Jones, Kerry Anne McGeary, Joseph V. Terza, and Christopher J. Ruhm (2013). “U.S. Alcohol Affordability and Real Tax Rates, 1950-2011.” American Journal of Preventive Medicine 44(5): 459-64. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3631317/.

9 David T. Courtwright (2001). Forces of Habit: Drugs and the Making of the Modern World (Cambridge, MA: Harvard University Press). David F. Musto (1999). The American Disease: Origins of Narcotic Control (Oxford: Oxford University Press).

10 Courtwright (2001). Gopal Das (1993). “Cocaine Abuse in North America: A Milestone in History.” Journal of Clinical Pharmacology 33(4): 296-310.

11 Social Capital Project (2018). “The Numbers behind the Opioid Crisis: Revised Utah Edition.” United States Joint Economic Committee, Office of the Vice Chairman. https://www.jec.senate.gov/public/index.cfm/republicans/analysis?ID=56F20AB6-A971-4EAC-A495-DE552F4F03C5.

12 The question wording and response options for all four sources is the same: “Taken all together, how would you say things are these days—would you say that you are very happy, pretty happy, or not too happy?” Figure 8 displays the percent responding “not too happy.” The 1963 and 1965 estimates are from Survey Research Service Amalgam and the 1972 through 2016 estimates are from the General Social Survey. The National Opinion Research Center (NORC) conducted the surveys for both SRSA and the GSS. Pew Research Center estimates run from 1996 to 2018. Gallup estimates run from 1977 to 2008. All estimates except from the GSS were obtained from the Roper Center’s iPoll database. The Gallup estimate for 1981 averages estimates from three surveys that year. The Pew estimates are averages from two surveys each in 2003, 2005, 2008, 2009, 2012, and 2016; from four surveys in 2011; and from five in 2014. GSS estimates were obtained from the GSS Data Explorer online tool on NORC’s website. We replace GSS estimates for 1972, 1980, and 1985-87 with ones from Betsey Stevenson and Justin Wolfers (2008). “Happiness Inequality in the United States.” Journal of Legal Studies 37: S33-S79, Table A1. Their estimates correct for the effects of survey changes that artificially affected responses to the happiness question.

13 Social Capital Project (2018). “All the Lonely Americans?” US Joint Economic Committee, Office of the Vice Chairman. https://www.jec.senate.gov/public/index.cfm/republicans/2018/8/all-the-lonely-americans.

14 For social capital trends, see Social Capital Project (2017). “What We Do Together: The State of Associational Life in America.” US Joint Economic Committee, Office of the Vice Chairman. https://www.lee.senate.gov/public/_cache/files/b5f224ce-98f7-40f6-a814-8602696714d8/what-we-do-together.pdf.

PDF version

Latest News