Understanding differences in life expectancy inequality

The life expectancy gap at age 40 between high income and low income individuals is substantial. I explore how medical expenditures and unhealthy behaviors account for the life expectancy gap. The data reveals the following. First, low income individuals tend to spend more on healthcare than high income individuals at all ages. Moreover, health disparities by income is salient due to differences in unhealthy behaviors such as heavy smoking. To answer how much dierences in access to medical services and unhealthy behaviors can explain in light of these stylized facts, I construct a life cycle model. The distinctive features of the model are that it flexibly incorporates unobserved, potentially correlated initial human and health capital stocks and embed unhealthy behaviors. Furthermore, the model includes two health systems: private health insurance and Medicare. The main findings are i) differences in access to medical care driven by income inequality potentially accounts for 12.5% of the life expectancy gap, ii) health insurance increases longevity for low income individuals, but modestly, iii) the health condition when young shapes the trend in average medical expenditures by income groups and iv) the impact of differences in unhealthy behaviors is predominant in understanding the life expectancy gap.

That is from Tomoaki Kotera, a job candidate from the University of Wisconsin, here is the paper itself.


Unhealthy behaviors, sounds like a morality play. If you ever have the misfortune of visiting a cancer ward in healthy places like Colorado or San Francisco or Washington DC (as say opposed to Mississippi) what will strike you is how many fit, relatively young people you see.

"Unhealthy behaviors, sounds like a morality play": do you think smoking is not unhealthy?

Smoking is only unhealthy for about 33% of lifetime smokers; the rest escape cancer from what I've read. It all depends on whether the tobacco plant you are smoking has polonium in it, which the plant can draw from the air and soil. Outside your body, radioactivity from such elements (I think they are alpha particles) can be stopped by your skin or a piece of paper or cloth, but once lodged inside your sensitive lung tissues, the particles, unless coughed out, will develop their surrounding tissues into cancer. The point being: it's not so much environmental as genetics and luck. Many cancer victims have a genetic propensity towards cancer.

Bonus trivia: Approximately 38.5 percent of men and women will be diagnosed with cancer of any site at some point during their lifetime, based on 2012-2014 data. Hopefully it's something mild like skin cancer, not melanoma, but something less harmful that is not even staged. Lung cancer is a nasty one, as is pancreatic and brain cancer.

There are other health issues that can arise from heavy smoking beyond cancer....

Also, 33% is a huge number!

There are other health issues that can arise from heavy smoking beyond cancer….

Yep, emphysema and cardiovascular disease ain't no walks in da park.

This claim certainly calls for at least a little bit of data.

I live in Denver.
Got to visit the chemotherapy unit at Kaiser every 2 weeks for 6 months.
At age 65, I was almost always, the youngest person there.
I realize that you were commenting about the cancer ward, but I would chemo to similar in demographics.

"differences in access to medical care driven by income inequality potentially accounts for 12.5% of the life expectancy gap": potentially?

"a health shock hits randomly in each period": does he mean that the shock is random wrt time period, or also wrt the individual, irrespective of his income?

"I estimate 25 parameters": that sounds like running elephant territory.

I constructed a model in which low income infants of less than 4 years old

Were unable to obtain pre-natal and pediatric healthcare services

Because of the selfish behavior of the donor class

In failing to re-enact or fund

The CHIPS program.

I know, I know, it doesn't matter anyway because of the unhealthy behavior of the rug rats

Who didn't pick the right parents

Who could have paid for their healthcare.

Those selfish bastards. The least they could do is take a second job so they could pay higher taxes to support those who choose not to work and take care of their own children.


Also, I hope bill gave extra on his taxes, since he cares so much.

Around here, there are signs on busses and billboards advertising free prenatal care. All the mother has to do is show up. And they have to advertise the hell out of it just to get them to do that.

I recently read Rubinstein's book "Economic Fables" which makes a strong case for the use of many kinds of models and especially informal, almost metaphorical models in economic thinking. (He also makes it clear that the hard nosed rationalist math-filled models often used by economists are a fable.)

What you are offering here, Bill, is a kind of fable. Let me model it in a slightly different way: (hysterical voice) "Won't someone think of the children?!"

It is one of the oldest heartstring-pulling fables in the book. So, nice manipulative rhetoric, but now can we go back to talking about how to reduce self-destructive behaviours?

Yeah, those children, they are like puppies. I am so glad that you can see through this and ignore the children. If we had used the over 65 population, like your grandma, it probably wouldn't have elicited as much sympathy.

Malthus is your man.

So, what this paper tells you is

That if you are wealthy, and do not have bad habits,

You do not need health insurance, do not need regular check ups, do not need that mammogram, do not need to be prescribed medicines

Because the difference in outcomes in life expectancy between you and a poor person is only 12.5%.

Life is good.

How about reading the paper?

How do you get from

"The main findings are i) differences in access to medical
care driven by income inequality potentially accounts for 12.5% of the life expectancy gap,"


"Because the difference in outcomes in life expectancy between you and a poor person is only 12.5%."

How is your reading comprehension?

Viking, per you comment below, if your initial condition at 23 determines all but 12.5% of the difference associated re income inequality, one really does not need to buy insurance, have checkups, etc. Life is even better. Nothing to be gained in living longer by being wealthy or having higher income. At least at age 40.

It is sad that you think your absurd thoughts do anything but cause you to be labeled as a fool. Marginal Revolution needs a permanently ignore commenter function, as the comments are half-filled with this kind of nonsensical trolling.

As usual, this study doesn't enable an accurate prediction of the longevity any one individual, regardless of his lifestyle or medical regimen.

"The life expectancy gap at age 40 between high income and low income individuals is substantial"

How does this model define High & Low "income" ? How does it correctly measure those factors on a very large population? How does it account for individuals shifting between income groups? What relevant variables does the model not consider? What is the margin of error in the methodology, measurements, and analysis?

Have you considered reading the linked paper? Or do you prefer shouting questions into the abyss?


This paper should be re-entitled:

"If it doesn't kill you, It doesn't count."

So, if you don't get medical care, and then can't walk, or breath easily, or see properly,

You will luckily live only 12.5% less long than a person who can walk, breath easily, and see properly.

In some cases, longevity isn't everything.

It's living the moment without pain or disability that could have been avoided with adequate healthcare.

All papers can't measure all things

Just because it doesn't cover everything, doesn't mean it's not a good paper or that it can't tell us useful information. The main thing you want to know is whether lifestyle choices lead to both low income and poor health or whether low income tends to precede poor lifestyle choices. It is not just about a morality play, but an actual question of how much marginally increasing the purchasing power or income stability of low income earners will improve their health. If actual dollar income tends to drive poor lifestyle choice instead of underlying personal attributes driving both independently but in a correlated way then government spending can help - if it's the latter you only get the 12.% difference identified here.

Regarding your point, I doubt someone can suffer from pain or disability without this also showing up statistically in their lifespan. If you can't move around and be active, your body isn't going to last as long. There is a pretty limited set of unpleasant medical conditions that won't also meaningfully impact lifespan.

The big problem with Bill's point is that he is confusing 12.5% of the life duration with 12.5% of the life duration difference. Somebody unable to grasp such a simple clear point is not qualified to criticize a paper.

Actually, if you wanted to critique the paper you would start with the starting age of the study at 23 ("The model starts when individuals are at the age of 23".), the correlated variables (wealth and income, for example)( "These moments are related to i)survival rates, ii) health status, iii) fraction of heavy smokers, iv) medical expenditures, v) earnings,
vi) net wealth, and vii) the ratio of buying private health insurance")., the starting point of the model which begins at 23, when a person in childhood could have developed a condition which continued after 23, which could account for differences later as the condition worsened.

But, you are correct: the difference is the 12.5% of the difference of age between two 40 year old groups.

Also, wealth or relative wealth would be expected to change over the duration of the model.

And, initial conditions matter as well with respect to both income and wealth so the change you would expect for each variable would be dependent on initial conditions.

The difference imputed to limited access to medical care is 1.2 years. The difference imputed to health behavior and health at 23 is 8.3 years.

None of what you says criticizes the paper at all. The point of the paper is that more health care spending is not going to solve the whole 9.5 year gap identified by Chetty (2016).

Re correlation of life expecancy and income, there are at least 2 additional factors to consider:

1. Low income jobs are often blue collar jobs. Which tend to be more dangerous than office jobs.

2. Maybe the causation is the other way around:it's not so much that wealth favours good health. But rather that good health favours higher income. (Chronically sich and disabled people seldom get high paying Jobs).

Maybe being in some way "well made" means it's likelier that you'll be both richer and healthier. "Well made" might - conjecturally - include carrying fewer undesirable mutations, and having avoided various toxins and infections in the womb or in childhood.

Clearly this is part of the effect.

However, it is heresy.

America has become a feudal state, its citizens have become serfs of the plutocrats.

"i) differences in access to medical care driven by income inequality potentially accounts for 12.5% of the life expectancy gap," and "First, low income individuals tend to spend more on healthcare than high income individuals at all ages." seem to contradict each other. Either medicare delivers fewer services for more dollars or spending doesn't equate to health.

Your access to care could be limited to certain types of care, and that limitation could somehow decrease your life expectancy. That does not prevent you from consuming more of the care you have access to.

"Second, if I eliminate the health insurance market, medical expenditures for low income individuals
would increase. Nonetheless, this change would affect their longevity only modestly."

Interesting statement. Supports the Oregon study, and that maybe health insurance even saves money. First thought is it saves by denying care, but spending on low income is higher than high income, not lower.

This paper, which is based on Chetty et. al., may be an interesting snapshot of the past but of course the statement toward the beginning that "life expectancy for males in the 90th income percentile is 4.3 years whereas the corresponding indicator in the 10th percentile is only 38.5" is no longer close to the true life expectancies of 40 year olds in those groups today.

Interesting that a ten-year difference is considered substantial in this context. The difference between men and women overall is generally about seven years, and it attracts no attention.

At ages 1 - 20, there is a 5 year gap but at ages 30 and 40, there is a 4 year difference. At 60 there is a 3 year difference and by 70 a 2 year difference.

And at 110 there is a no difference at all.

The gap in life expectancy, starting at age 40, between the 90th and the 10 income percentile:
Unhealty behaviors: 3.7 years
Access to medical services: 1.0 years
Health insurance vs. no health insurance: 0.3 years

Seems like the trick is getting to 40 years of age.

That's the goal. Never mind what happens later. Like if you ever get to 65.

If you were poor and made it to 40, maybe your best years were behind you. Of course, we don't know, because the train stopped at 40. And, as we all know, if you had a health condition before 40 it disappears after 40.

Having poor health habits ranging from smoking (not quitting), excessive drinking, poor diets, lack of exercise, obesity and not following "doctors orders" or medication procedures are also characteristics that will result in higher poverty. Having "poor health habits" may also go with "poor work habits" in a nasty feedback loop. A random health problem arises, it depletes financial reserves (often small with poor work habits) with extra expenses and time off, increasing poverty and possible loss of jobs and now we see a correlation between health outcomes and incomes.

However, we still don't know which way causation actually flows or there is no "causation", just coupled feedback systems. Only fixing both poor health and work habits will stop the feedback loop. However, our "Social Science/humanities" experts have no idea how to fix sick local cultures and throwing money an unstable feedback loop is a total waste of resources.

How do societies with less income inequality and wealth differences vary in healthcare outcomes. Sweden. Denmark. Norway.

Depending on source Denmark is either OK or bad. Even just a few years ago it was comparable to the U.S. and worse for conditional life expectancy once you reached your teens since IIRC their neonatal mortality is much lower. I think the WHO numbers still show them ~80, probably still the worst in western Europe. Sweden and Norway are solid but not standouts, a little better than the UK but behind leaders like Spain and Japan.

Which side are you arguing again? LOL 😀

Apso, Those are states with less income inequality and higher ages.

Steve had trouble saying it but he did.

Laughs on you.

It is also possible that people with chronic health problems earn less because they are sick, miss work, or can only hold non stressful jobs. This may be why the spend more on health care but are not healthier.

If assortative mating means whites marry whites, blacks marry blacks, then if divorce is more easy then people will remarry more often, which decreases the oddball white marries black marriage, statistically speaking, if you adjust for population aging. So it’s really much ado about nothing. Due to globalization you should be getting less assortative mating, not more, but possibly remarriage is skewing this trend. That’s my 10 second synopsis from the abstract, which is usually pretty accurate.

Inequalities are unhealthy
by Vicente Navarro

Johns Hopkins University - Bloomberg School of Public Health

The answer is that inequality is in itself bad, i.e., the distance among social groups and individuals and the lack of social cohesion that this distance creates is bad for people’s health and quality of life. Studies performed among civil servants in Great Britain have shown, for example, that life expectancy (the years that people can expect to live) among the top civil servants, grade 32, is longer than the life expectancy of civil servants of grade 31, who have longer life expectancy than civil servants of grade 30, and so on, reaching the lowest life expectancy at grade 1. There is no poverty among British civil servants, but there are significant differences in their life expectancies. The same finding has been replicated in other countries.


The long standing question is does poor health cause poverty or does whatever causes poor health also causes poverty? (The 3rd option, that lack of access to health care causing poor health seems to always come out as only a small factor.)

"The distinctive features of the model are that it flexibly incorporates unobserved, potentially correlated initial human and health capital stocks and embed unhealthy behaviors. "

So does the above mean they found evidence that it is more the latter?

Comments for this post are closed