Are Middle-Aged White Men Dying from Despair?

Economists Anne Case and Angus Deaton have a couple of well-known papers, here and here, identifying a marked increase in death rates for middle-aged white American males without a college degree due to alcohol (both from overdose and from long-term abuse), drugs, and suicide. The upturn started around the year 2000, and the increase has been striking. Case and Deaton term the phenomenon “deaths of despair.” And, given the age of the men involved, it’s difficult not to think of Arthur Miller’s Willy Loman in association with the findings. But there is a lot to be cautious of regarding both the asserted American exceptionalism of the phenomenon, as well as in identifying the causes.

To be sure, Case and Deaton are more theoretically circumspect in the 2017 Brookings paper than they were in the 2015 PNAS paper. But before we get to the discussion of theoretical cause, the data warrant a closer look. Case and Deaton’s argument holds out the distinctive increase in alcohol, drug, and suicide mortality for this American age cohort relative to similarly aged males in Europe, and relative to similarly-aged African American and Hispanic men in the U.S. But while the magnitude of the phenomenon for this cohort of white American men is distinctive, Case and Deaton’s data don’t seem to indicate the phenomenon itself is — with apologies for a qualified superlative — quite as unique as they suggest.

First, in their Brookings study, the “Average Annual Percent Change in Mortality for Age 50–54 by Cause, 1999–2015,” Table 2 indicates an annualized increase in mortality from drugs, alcohol or suicide for U.S. white, non-Hispanic men during that period of 5.4 percent.

While that’s the largest increase among the comparison groups (national, ethnic, and racial) reported in the table, contrary to many of the narratives about their data, some of the comparison groups witness increases in these forms of mortality as well. To wit, Irish, Canadian, and Australian men in this age cohort also saw increases in mortality due to alcohol, drugs or suicide in the range of 3.0 percent, 2.5 percent, and 2.5 percent, respectively. Men in the U.K., Sweden, and Denmark also saw increases in these forms of mortality, even if the rates of increase were significantly lower.

Indeed, fully half of the fourteen nations reported on by Case and Deaton saw an increase in mortality from these causes for men in the middle-aged cohort. Again, the rate of increase is highest in the United States, but the fact that seven out of the fourteen countries reported also saw an increase in these forms of mortality take some of the shine off the claim that the phenomenon is exceptionally American.

At the very least, these other countries provide the possibility of comparative leverage for otherwise nation-specific theories for the mortality increase in the United States. This is a possibility that Case and Deaton don’t follow up on (albeit, because of data limitations, not because of their lack of desire).

Turning from comparative national to intranational comparisons, U.S. Hispanic men and U.S. black men in the age cohort have saw increases in these forms of mortality. Again, the rates of increase have been much lower than white men, but the fact of the increase itself also takes a bit of the shine off the uniqueness narrative Case and Deaton, and others, want to tell.

As I mentioned above, Case and Deaton punt their cross-national comparisons (for lack of data) as they delve into explanation. That is unfortunate because the cross-national data likely could provide substantial leverage on a number of the hypotheses they float as possible explanations for the data.

In the last half of the Brookings paper, Case and Deaton discuss various possible hypotheses, seemingly rejecting one hypothesis (income decline), but then recovering even that one by the end of the paper. They even provide a respectful nod to Charles Murray’s theory in Coming Apart.

By the end of the paper, however, they are left with positing “slow-acting social forces” as the most plausible causes, but it’s not entirely clear what that entails. While they want to endorse economic factors, they’re careful enough scholars to avoid committing to a specific cause given the significant ambiguity in the data. They even provide a nod to the isolating effects of the sexual revolution as a possible cause, even if they want to “emphasize” economic causes:

In our account here, we emphasize the labor market, globalization, and technical change as the fundamental forces, and put less focus on any loss of virtue, though we certainly accept that the latter could be a consequence of the former. Virtue is easier to maintain when it is rewarded. Yet there is surely general agreement on the roles played by changing beliefs and attitudes, particularly the acceptance of cohabitation, and of the rearing of children in unstable cohabiting unions.

A little later, however, even the “virtue” hypothesis returns with a bit more plausibility as Case and Deaton note the difficulty of disentangling causality from statistical indicators that all move in the same direction.

What our data show is that the patterns of mortality and morbidity for WNHs without a college degree move together over birth cohorts, and that they move in tandem with other social dysfunctions, including the decline of marriage, social isolation, and detachment from the labor force.

One last empirical whine, though: I do wonder about combining alcohol abuse, drug abuse and suicide in a single measure of “despair.” Despair is all about demand, yet for both drugs and alcohol, I think there are plausible supply-side theories to consider as well.

On the drug abuse front, the last couple of decades have seen a huge shift on the supply side, particularly in predominantly white communities, with the significant increase in access to and use of opioids, as well as meth and other drugs in rural communities. “Despair” in these communities could be constant while supply changes account for the variation in mortality figures. Similarly, alcohol mortality includes deaths from long-term alcohol abuse as well as alcohol overdose. Yet increased availability of alcohol in grocery stores and gas stations became widespread in the 70s and 80s, a couple of decades prior to observation of a spike in the long-term effects of alcohol abuse.

I am open to the “despair” hypothesis, but an increase in despair isn’t required, at least for the drug and alcohol mortality figures, if there is a supply side story. (And, to be sure, the story could be both a demand and a supply side story.) The comparative national data, and comparative ethnic and racial data, in the Brookings paper, however, make me uncertain the phenomenon Case and Deaton are observing is unique to white middle-aged American men. And, whether it’s uniquely American or not, I still wouldn’t take any bets yet on the identity of the culprit.