Life Expectancy in the U.S. is Getting Shorter for the Least Educated

The late 20th century brought landmark public health movements to the United States, like the control of tobacco, and medical breakthroughs in the treatment of heart disease and cancer. Life expectancy surged overall, but today the lower educated are still stuck in a different era.

From a new paper in Health Affairs (behind a paywall): “We found that in 2008 US adult men and women with fewer than twelve years of education had life expectancies not much better than those of all adults in the 1950s and 1960s. When race and education are combined, the disparity is even more striking.”

Here’s a graphic showing the breakdown of life expectancy at birth just for white females. Notice three things. First, the current gap between most and least educated women is more than 10 years. Second, the gap actually widened by several years between 1990 and 2008. Third, the gap widened mainly because of a reduction in life expectancy at the bottom, rather than faster gains at the top.

This is truly sobering. I previously blogged about a paper showing that health behaviors do not explain the growing mortality gradient between less and more-educated Americans, but I did not put the scale of the problem in context.

These powerful educational differences are the primary driver of growing disparities between whites and African Americans. Because a similar gradient exists among African Americans, and the college graduation rate is so much lower among African Americans, they are falling behind white Americans.

How does education confer positive health benefits? The authors consider at least two compatible theories:

“Education exerts its direct beneficial effects on health through the adoption of healthier lifestyles, better ability to cope with stress, and more effective management of chronic diseases. However, the indirect effects of education through access to more privileged social position, better-paying jobs, and higher income are also profound. The absence of education and its related socioeconomic status benefits exert their direct harmful effects throughout the relatively shorter lives of those in less fortunate social positions (especially whites). This is one important reason why efforts to modify behavioral risk factors alone are not likely to have a major impact on disparities in longevity.”

Perhaps. But before we give up on the idea that modifying behavioral risk factors will not make a significant dent in mortality, we need to get a better handle on how much we can improve life expectancy by chipping away at the main exposures that afflict the lower educated groups. With smoking, it was a comprehensive campaign to change the image of smoking and increase education about the risks of tobacco use. With obesity and the diseases of sedentary lifestyle, we are still in the phase of developing interventions and changing the messaging around unhealthy eating. Some behavioral interventions will work, many will not, but it remains to be seen how much reduction in mortality can come from pairing more dietary education with a fundamentally different food delivery system.

The authors also single out increasing educational attainment, which happened steadily through the 1990s, especially for African Americans. That goal is now more imperiled than ever before due to cuts to funding for higher education and tight state budgets. Evidence like this study – showing that educational attainment is a driver of life expectancy needs to be injected more forcefully into conversations about cutting educational spending. For reasons that we are still trying to piece together, more school means that people live longer and healthier.

10 responses to “Life Expectancy in the U.S. is Getting Shorter for the Least Educated”

  1. But Brendan, we have very good evidence that educational interventions targeted at behaviorial modification have a tendency to increase health inequalities even where they improve overall population health. I agree that the explanations for why education correlates so strongly with mortality are complex and multivariate, but I am quite dubious of the idea that one of the important pathways between education and health is through modification of risky behaviors (especially because the distribution of such behaviors is strongly determined by social position to begin with).

    Other thought here is regarding the big debate among SDOHers regarding whether education is really a fundamental cause of disease. I tend to side with David Low’s view, which is that even if it is not a true root cause, it is a critical driver (and this study provides excellent further evidence for the point), so we’d do well to attend to it even while we are interested in more upstream factors.

    (I don’t know why I always seem to be chipping in here to dissent from your posts. I really like them — thoughtful, careful, etc. I think the fact that I enjoy them so much and that they make me think is partly what inspires me to comment . . .)

  2. Hi Daniel,

    Very illuminating comment. First, I’d like to see the “very good evidence that educational interventions targeted at behaviorial modification have a tendency to increase health inequalities even where they improve overall population health.” If an intervention increases everybody’s health, but increases the healthiest’s health the most, that doesn’t count as a reason to not implement the intervention in my book. It does raise some important questions. First, what can we do to increase the effectiveness of the intervention for those badly off? That is where I see education coming in. But let’s agree that not all education is equal.

    Education does many things, in addition to changing people’s social positions. Education changes people’s values, their time horizons (increasing future orientation), it may instill self-control, and it increases literacy and numeracy. If we had school curricula that placed health promotion on a par with learning skills for the SATs, we might begin to increase the returns of education to the health of the most disadvantaged. We have examples of this. In the United States in the early 20th century, and today in the developing world, children are often taught the basics of hygiene, which they can share with their families. We could do much more to harness the potential of children, teenagers, and college students as leaders for health change for healthy eating, exercise, alcohol use, safe driving, and treating depression.

    Are there models to make this happen on a large scale? There may well be, but I don’t know much about this. Of course, none of this is a refutation of the fundamental causes thesis, but it does suggest that the pathway between education, behavioral modification, and health can sometimes sidestep complex systems of social inequality, or at least work within the limits of structural inequalities.

    Does that make sense? What do you all think?


  3. Hi Brendan,

    What a fascinating dialogue! I do hope others will chime in as well.

    Shameless self-promotion: I address a lot of the commentary below in a paper of mine out in Public Health Ethics: But briefly:

    So, the broad framework is nicely addressed in a wonderful paper by Benach et al. that Paul Kelleher nicely unpacks in this post:

    I’ve relied on this paper in several of my own articles, but the key idea here is that there are two primary goals of ethically optimal public health policy: to improve overall population health and to compress health inequities. It is quite possible to have public health interventions that accomplish both, neither, or only one of these goals while exacerbating the other. The example I give of a classic public health intervention that maximizes both of these goals is sanitation. Sanitation increases overall population health and, because the least well-off bear disproportionate burdens of waterborne disease, it increases their health more than that of the affluent. Therefore sanitation compresses health inequities insofar as it is equally distributed across a population.

    So I would want to perhaps disagree with, or if not, refine your notion that a public health policy which increases population health even if it expands health inequities is copacetic. It might be acceptable, but it is suboptimal in comparison with a public health intervention that both improves overall health and compresses inequities. More to the point, we might accept a given intervention that improves health even if it expands inequities if the trade-off is small (i.e., if the increase in inequities is small and the gain in overall health is large), but I do not think we would wish the bundle of policies and practices we apply to a given health problem to be characterized by an approach which expands inequities. Better to work on approaches – which exist in the real world, such as, most obviously, Rose’s whole population approach – which promise to improve overall health and compress inequities.

    As to your specific question, we have pretty good evidence that education targeted at behaviorial interventions can expand health inequities. In fact, this has happened – smoking cessation programs disproportionately benefit the affluent. So even where overall health has improved because incidence has decreased, health inequities based on smoking-related diseases have increased along the social gradient. I discuss this in my paper, and cite quite a few sources, although I am perhaps most partial to Capewell and Graham’s lovely 2010 paper in PLoS Medicine. There they develop the concept of agentic interventions, or those which depend for their benefit on the resources and agency of the individuals to whom they are targeted. Honestly, I do not believe the finding is even particularly controversial – agentic interventions, especially characterized by those that target behaviorial change on the individual level, have a marked tendency to expand health inequities.

    Whew – as ever, too many words! Thanks for the great conversation.

    • Terrific! Because I agree with so much of what you say let me just take the opportunity to clarify the main ideas I am interested in:

      Let’s say that there are two different investments in children that policymakers are considering, but these investments are not completely mutually exclusive (they can do at least some of each). One is to invest in a healthy eating curriculum for elementary schoolers. Preliminary data suggest that this intervention will reduce future obesity, particularly for children who are strong readers, because these children will complete the supplementary activities at home. The second choice is to put more money into the education system, which could go to any of several programs ranging from physical education, to a reading program, to increasing performance on a college entrance exam.

      What’s important to notice is first, that each of these investments could increase population health, and second, that each intervention could either reduce or increase disparities depending on how the resources are targeted. For example, it may be that the obesity intervention, if implemented in the current environment, would yield little benefit to many disadvantaged children who do not have parents to help them with their homework. But now let’s assume that a complementary investment is made in a school reading program. The school reading program may have lots of positive consequences, but we are mainly interested in how the literacy intervention improves health. One hypothesis is that delivering reading skills in combination with a health intervention would improve the yield on the health intervention for the worst off group. Whether it is enough to compensate for other factors that limit that intervention’s effectiveness is an open question — it may not be enough to reduce disparities.

      This scenario is totally plausible to me, but unfortunately we often speak about “behavioral interventions” and “education” as if they are completely without context. The yield that one can realize on each may depend on what one is doing in other domains. We don’t have data to support a firm conclusion one way or the other, however. I am just saying that we should be careful not to put down behavioral interventions just because their effectiveness is right now hampered by the weaker agency (to paraphrase your term) of the least well off group. At least, not until we know whether the interventions are “tweakable” to the needs of specific subgroups.

      This is just my opinion — I would get behind these kinds of interventions even if they did little to change the underlying structure of disadvantage, because having a very targeted impact on population health could more than justify the investment. This is not a point that I am able to defend in great detail, but I just wanted to raise the idea that doing behavioral things may be “worth it”, whether it addresses the fundamental causes and indeed, whether or not it is a strong enough force to countervail the drivers of health disparities. At least a rising tide lifts all boats a bit.


  4. I think everything you say is quite reasonable, but my own view remains at least somewhat different. Look, there’s nothing wrong with parallel processing here. I’m certainly not opposed to giving people education in ways that we think might increase the uptake of healthy behaviors, even if I believe quite strongly that that is not the kind of education that is most likely to have a significant impact on overall health or on the compression of inequities. The false choice fallacy must be avoided; we can implement each of these kinds of interventions.

    But the ethical issue is one of relative priority — even if we ought to implement both, it does not follow that we ought to allocate the same level of resources to each category of intervention. And given the evidence, I have little hesitation in concluding that we ought to expend significantly greater resources on interventions targeting (more) upstream variables such as education in the broad sense rather than health education. And part of my frustration is the almost undeniable evidence that our current funding priorities are exactly the converse; we spend way too much time talking about and funding narrow health promotion and education efforts targeted at individual behavior change, and not nearly enough time and money targeting at increasing educational attainment at a broad level. The former have the added disadvantage of intensifying stigma directed at the least well-off, which is both socially corrosive and bad for one’s health.

    But there’s also an elephant in the living room that we have not really touched on here: most health education interventions focused on individual behaviorial change actually do not work. We are speaking of interventions that do work, and discussing how we ought to prioritize those relative to interventions targeted higher up the causal pathway. But that’s a counterfactual which is quite different from the world we live in, the latter being one in which good evidence — again, see my paper! — shows that many of these kinds of health education interventions simply do not work. And indeed, once we understand the ways in which deleterious social and economic conditions seem to facilitate risky health behaviors, we would predict exactly that lack of efficacy. That is, if it is something about these root conditions that drive the highly unequal distributions of risky health behaviors, then effective remedies would be ones that actually addressed those root conditions, as opposed to the methodological individualism that characterizes so much health education and health promotion in the U.S. at least.

    Sometimes I ask my students about this, and it seems to shock them: we know that risky health behaviors like smoking, promiscuity, poor diet & nutrition, are disproportionately prevalent among the least well-off. Is it because poor people are just ignorant? Do they simply not know any better? Or is it rather that there is something about the lived experiences of material deprivation that is somehow generative of risky health behaviors? And if it is indeed the latter – which I most certainly think it is – then we should not be surprised when simply seeking to educate individuals regarding health behaviors is ineffective.

    Does this make any sense?

  5. At least in the U.S., I would guess there is a strong correlation between educational attainment and availability of health insurance. Wouldn’t this be the most obvious reason for the gap in longevity?

  6. Hey Eric,

    Actually, there is only weak evidence that access to health insurance is a prime determinant of health inequalities in the U.S. A recent paper suggests in the case of Medicaid this might be so, but David Orentlicher correctly notes such a finding is anomalous with the great body of evidence on the subject. And in any event, I do not think the study stands for the conclusion asserted, at least in part because of the difficulty of adequately controlling for the intergenerational effects of deleterious conditions over the life course.

    So I would have to say that while the correlation between access to HI and education might be the most obvious reason for the gap, the evidence strongly suggests that it is not the primary reason for said gap.

  7. Thanks, Daniel. The conclusion seems counterintuitive. My knowledge on the topic is narrow, but in Alabama (which has limited Medicaid coverage), untreated diabetes has a major impact on life expectancy. I’m not sure how a study on the correlation between HI and life expectancy would be useful unless it spanned a decade or more, given the slow progression of diabetes to life-threatening cardiovascular events. Even a Medicaid study would need to factor in whether dialysis is covered, and whether it is offered in convenient locations. I’m using diabetes as an example, but I assume the same issue would come up with other progressive conditions.
    The disturbing undertone to all this is that lack of educational attainment and generational poverty are closely intertwined. Is it possible that unhealthy lifestyles (including chemically induced escapism) reflect a lack of enthusiasm for life? A statistically significant despondency?

  8. Well, the methodological problem you note here is IMO an extremely important one, but it cuts the other way as well (i.e., controlling for the effects of SES and other conditions requires such controls over the life course and perhaps even intergenerationally. The near-universal habit of using even good snapshot proxies for social status like neighborhood is insufficient as a control for SES because of the latter’s effects over multiple life courses.

    Lack of educational attainment and material deprivation is indeed strongly intertwined — why that might be the case is really important, although I am somewhat dubious how much of it is motivational (although there is evidence the crushing burdens of living with profound deprivation is generative of risky health behaviors, see Akers et al. 2010).

  9. I am sorry to say it but its false and myth. Why ? Because my grandma was 84 years old. She had no education. She can’t read and write at all. Her life span was fine without education. I have know many older people who have no education in my country have a longer life span. This is false story! Yeah, Its false.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: