U.S. life expectancy in perspective

From Avik Roy:

A few years back, Robert Ohsfeldt of Texas A&M and John Schneider of the University of Iowa asked the obvious question: what happens if you remove deaths from fatal injuries from the life expectancy tables? Among the 29 members of the OECD, the U.S. vaults from 19th place to…you guessed it…first. Japan, on the same adjustment, drops from first to ninth.

Here is more.  Arnold Kling comments.


Does this study account for the fact that our amazingly efficient health-care system is "diagnosing" people with cancers that couldn't *possibly* kill them given their age and the rate of growth (just so they can bill medicare)? IE is it "life expectancy at diagnosis with stage 4 cancer" or with any "cancer"?

What totally benign cancers that can't kill people given age/rate of growth are being diagnosed so that Medicare can be billed? I'd be surprised if Medicare was never billed for cancer treatments that were excessive, but I highly doubt this is a thing.

It may also surprise you that cancer can cause problems before it kills you.

Most prostate cancers will not kill you but are treated as death sentences. The extensive use of PSA testing results in many interventions at very early stages for cancers that would not be fatal. Statisitics on survival post intervention are worthless. Look at per capita mortality from the disease.

Also look at average age at death from prostate cancer.

Kling also mentions that possible confounding factor -- that early diagnosis may skew cancer survival rates. But the life expectancy data isn't vulnerable to that problem. If removing injury deaths from the stats pushes the U.S. to #1, we have to conclude that either the U.S. health system is effective (albeit expensive) or alternately that there's something about the American diet or lifestyle that promotes long lives. I'll leave it to you to speculate on which of those two explanations is more plausible.

Or third conclusion: those who die from injury early tend to be a group that would still have a short natural life expectancy. So completely excluding them isn't totally fair, because those same people in other countries would have lived but probably died earlier regardless.

The conclusion is that comparing "natural (non-injurious) life expectancy" isn't quite so simple. It may require advanced statistical methods, like IV to look at exogenous shocks to the injury rate and the subsequent impact on natural life-expectancy.

alternately that there’s something about the American diet or lifestyle that promotes long lives. I’ll leave it to you to speculate on which of those two explanations is more plausible.

Believe it or not it could be because we are fatter. See the CDC report: https://www.creators.com/health/david-lipschitz-lifelong-health/longer-life-one-outcome-of-being-pleasantly-plump.html

Also note Hispanics live longer in the USA that the rest of the population and blacks at 85 years old have greater life expectancy that whites at 85 years old but die much more from homicide and accidents indicating if you take out accidents and homicides blacks may live longer than whites, so believe it or not ethnicity could be a factor in favor of the USA.

I forgot the link for black life expectancy at age 85. here it is:

Most of the homicide mortality rate is heavily concentrated among men. Higher homicide mortality means that a higher proportion of 85 year olds in the population will be women. 85 year old women have higher life expectancy than 85 year old men. So it's not surprising that black 85 year olds as a whole have longer average life expectancy since a higher percentage are women. They could still overall have worse health.

Like I said in my last post, you can't just assume that early deaths are random representatives of the broader population. A simple comparison doesn't work. You must adjust for early death population dynamics. And that's damn hard to do.

This is a very important question and it should be branded on the left hand of everyone debating about health care.

Note that the study quoted did avoid the weird question of early-versus-late diagnosis by simply subtracting out fatal injuries. It should make everyone suspicious of putting too much weight on using life-expectancy measures as the sine qua non of measurements since just taking out something that really shouldn't depend too much on our health care system shows such a big swing.

But, yes, measuring "how well does this health care system deal with people diagnosed with cancer" is something subject to a lot more inputs than just the health care system. See the Will Rogers Phenomenon for another pickle.

Totally unrelated- diagnosing a fatal illness earlier in its natural history may make your therapy look more efficacious, but it will not effect aggregate life expectancy at all.

I can't find how they define "fatal injury," but for the OECD a "fatal occupational injury" includes injuries that result in a death within one year. I would be surprised if the metric they are using includes only accidents that do not give the victim any chance of receiving healthcare. This statistic could have at least something to do with overburdened emergency rooms, which, many would argue, is a symptom of our poorly designed healthcare system. But maybe someone can find out exactly where this statistic comes from and how it is defined?

More on the Ohsfeldt and Schneider results here, including a dynamic table for ranking the nations' reported life expectancies.

Oh, by the way, both the regressions they show for the relationship between health expenditures per capita and GDP per capita are wrong. (As a caveat, what they show is okay for the data they have, but the fitted relationships they, and other researchers, have found break down when you start getting into subnational units).

The "fatal injury" context might be different in the reports from different countries. It's interesting to find that after this adjustment, the life expectation for Japan decreases from 78 to 76. It implies that average age of death caused by fatal injury in Japan is more concentrated in older group. In the US, it is totally different, life expectation increases. Therefore, the group of people come across such fatal injuries are younger. Overall, this adjustment doesn't yield any meaningful conclusions.

That may mean that in Japan fatal injuries are spread pretty evenly throughout a lifetime, whereas in the US they maybe heavily concentrated among the young.

I expect it is common for "fatal injuries" to increase with old age, because people are more likely to fall over, and less likely to survive when they do. To have the opposite effect means that gang-murders and car-crashes are kill more people that bathtub falls.

I suspect in most countries, the car-crashes are enough. At least for males.

I saw the study and what I don't understand is the following: According to the actual mean life expectance the US is 19th with a life expectancy of 75.3 years. When the fatal injuries are removed and the standardized mean is taken the US life expectancy rises to 76.9 and the US vaults to 1st place. However the life expectancy as per the standardized method of the countries above the US declines rather than rises. i.e Japan's life expectancy for Japan declines from 78.7, according to the actual mean, to 76, according to the standradized mean after you removed the fatal injuries. can someone explain this difference.

I'm guessing it has to do with the tails of the distribution with the distribution bounded at 0 and ~100. In the US, homicides, car accidents, etc are the loss of the young. We have lots of young people who die due to lethal injuries.

In other countries those who have fatal accidents tend to be on the right hand side of the distribution and removing them moves the average down.

What is unclear to me is what the cause of this is? I haven't read deeper than the blurb on this page, but it could simply reflect demographics if the population has lots of old or young people, right? If you have large numbers of one cohort and you remove those that die for any reason it will alter the average significantly. I have no idea if this is the case, and am too lazy to look it up!

As an actuary who studied mortality table for living once, I can assure you that if you reduce the mortality rate, the life expectance should increase. The result for Japan doesn’t make sense at all. This makes me very suspicious of the result.

The only potential explanation is that the author added the global average mortality due to accident etc to each individual country. Hence US got a lower mortality table and Japan got a higher mortality table.

'The result for Japan doesn’t make sense at all.'


Sure the result makes sense - see, we are number 1, at least when the data is properly manipulated. And there is a bit of (mockery worthy) explanation at the end of the Forbes article.

Maybe the people were removed from both the numerator and the denominator?

I applaud the effort, at extracting injury. It might lead to constructive change, say a greater effort toward emergency response, and perhaps reduced effort on low causes of mortality, etc.

But, I really can't be convinced by an adjustment that removes causes and results in a lower expectancy. Speaking directly to Weber, if you remove "injured Japanese" you should be removing mostly young Japanese. I'm not getting how removing young deaths lowers expectancy.

Yu Feng's answer is more mathematically intuitive.

That is my impression. They are not reducing the mortality rate, per se. They are pretending these people don't exist as an excluded outlier. If you have a large number of people at the extremes of a distribution and you simply remove them, the mean can change quite a bit. I'm not convinced this is what they did, but it is possible.

I find it interesting, but not particularly useful without knowing what the distribution looked like to start with. Plus this too can mask issues. We all know the young have higher risks because they do stupid things. But poor healthcare in the elderly can make them much more prone to injuries that are potentially life threatening. Excluding them would be inappropriate in such a case. Heck, just being elderly is a risk factor for fatal accidents, regardless of care. Knowing nothing of Japanese demographics, could this be the case with the strange Japan result?

I looked at the paper--basically a power point--and had the same questions. I wish the had described in detail how they adjusted data, what they defined as accident, how gun deaths or drugs enter into the mix, or whether emergency room or hospital access affected death rates as well for the excluded category. We will never know from this non replicable PowerPoint.

The thing is I don't see why these people need to be excluded, and I agree with the others that this could potentially skew the results. If I were doing this, I would split deaths into two decrements, natural and accidental. Then, rather than just throw out the data for "accidental" I would leave them in as censored data points. For example, if someone gets shot at 18, we would assign them a natural age of death of "at least 18" since we don't know when they would have died, but it would have been greater than 18. This is a basic procedure in survival studies when participants in the initial cohort leave the study early.

Only way I can account for that table is that they're regressing life expectancy against fatality rate or something, then calculating the result if everyone has the average fatality rate.

Is it possible that rather than just removing injuries, they are using some control adjustment? In other words, I don't think they are stripping out auto injuries, but rather trying to hold them constant. Is there a direct link to the study?

Looks like many older people are part of the fatalities.
Euthanasia being counted? Is it more prevalent in Japan?

If one excludes non-natural deaths in calculating mean life expectancy for any country, the life expectancy for that country should go up -- not down.

Not necessarily. If there is a form of non-natural death that skew old, then subtracting them out might reduce the calculated life expectancy. So imagine there was a suicide epidemic among the elderly in Japan:


If suicides include more people who've already passed the mean age of death than those younger, subtracting out suicides out could reduce expectancy. But that would be a pretty perverse result.

Ah, I see. Yes, if you remove mostly-old people and not mostly-young, you do the opposite of a survivership bias. You make a non-survivor bias.

I'm surprised though, I'd think that the young get out more and die due to accidents more. An old person, such as my 90 YO grandfather, might fall, break a hip, and die later from complications at surgery ... but that was age more than accident, IMO.


I don't think it makes sense that excluding fatal injury reduces Japan's life expectancy.

Something is fishy. Very fishy.

In the comments, the author states that these figures are normalized (with no mention as to how), hence only useful for relative comparison of nations.

I figured as much.

The frequency and nature of fatal injuries is so different in the US than other OECD countries?

Yes. The U.S. has far more automobile accidents than many OECD nations. And also a much higher homicide rate among its criminal populations.

The U.S. is tied with Belgium in the table you linked to above.

On the other hand, Ohsfeldt and Schneider show that the U.S.'s life expectancy jumps 1.6 years when they make their adjustment while Belgium's increases only 0.3 years. Their data are from 1980-1999 so it might be the case that the discrepancy was larger then (and also for homicides) but I share other people's desire for these guys to show their work.

We drive alot more than most european countries, probably density related. Additionally, we have certain cities and areas that are particularly troublesome when it comes to homicide, ie organzied crime, gangs, drugs, etc. For example, those 500 Chicago murders of young men a year really make your life expectancy go down and hence look a ton worse from a health care prospective.

You think "alot" is a word.

Why not just make up the numbers you want, Jan? If the frequency and nature are bad, change them. Or blame Republicans for not being able to.

What a cool, fireworks-on-the-Fourth-of-July, patriotically fun fact! It reminds me of the Memorial Day weekend some years back when three people died on Lake Mead (near Las Vegas) in separate accidents. When I read that in the paper Tuesday morning, I thought, "Wow, it sounds like a lot of people were having a great time."

Avik Roy says "It's great the Japanese eat more sushi", but that result from removing fatal accidents indicates maybe that's not such an important difference.

Does eating improperly-prepared fugu sushi count as "accidental death"?

The cited study makes untenable assumptions about competing risks and uses low quality data. For a more updated and believable study, see the recent Health Affairs article by Jessica Ho: http://content.healthaffairs.org/content/32/3/459.abstract. While injuries are important and eliminating them should be a policy goal, this obviously won't push us to the top of the heap.

From the Jessica Ho article;
"The major causes of death responsible for the below-fifty trends are unintentional injuries, including drug overdose—a fact that constitutes the most striking finding from this study; noncommunicable diseases; perinatal conditions, such as pregnancy complications and birth trauma; and homicide."

My initial comment was simply to note that if one decides to adjust certain data, one can make that data look however it should in the eyes of the person doing the exclusion.

Thankfully, the Forbes author was kindly enough to note how that works, in an update - 'A number of mathematically astute readers have asked why some countries have increased average life expectancies once you take out fatal injuries. I asked Robert Ohsfeldt about this, who responded that the adjustment factor was based on fatal injury rates relative to the average. Hence, the adjusted numbers shouldn’t be seen as hard numerical estimates of life expectancy, but rather as a way of understanding the true relative ranking of the various countries on life expectancy excluding fatal injuries.'

Let me note that again - the 'adjusted numbers shouldn’t be seen as hard numerical estimates of life expectancy, but rather as a way of understanding the true relative ranking of the various countries....'

The new social sciences - always providing the truly relative, even when it comes to the only true human universal - death.

Why it crazy to remove death outcomes in which the health care system could not under any circumstance make a difference? This statistic is repeatedly used by single payer advocates, suggesting that our system fails. In reality we are looking at perhaps a problem with the transportation system and an extremely small, violent subsection of the US population (gangs, organized crime, drug trade, ghettos).

Thank you. I was having a hard time parsing through the discussion here but you cut to the chase.

A number of mathematically astute readers have asked why some countries have increased average life expectancies once you take out fatal injuries.

Seriously? People are asking why life expectancy is higher if you remove early deaths?

It is the flaw in making comparisons between countries. We see something similar in crime stats. The usual suspects prattle on about crime rates in Europe versus the US. Then someone rudely observes that adjusted for race, age and sex, crime rates are not all that different. Across the Anglosphere, they become almost identical. Oddly, the people the most fond of comparing America to other countries know the least about those other countries.

As this relates to health care, it just shows that once you address the basics like infant mortality and common treatable diseases, life spans grow to their genetic maximum, less misadventure and social engineering.

My understanding is that people who choose race are finding a correlation, but people choosing poverty and education are finding one as well. Frankly, I find the latter to be more constructive efforts. I mean, the race thing is a search for a dead end, an answer with no solution.

There are solutions. More importantly though, may not be the solutions, but the non-causes.

Your understanding is incorrect, and it's embarrassing that you could have it. Hispanics have long lifespans, despite being poor, uneducated, and fat.

BTW, congrats on proving the general rule that "race" can fit into any MR discussion.

John, can you explain why you prefer to consider income and educational attainment to the exclusion of race without using emotional or political reasons based on your priors?

My prior is definitely that people are people, and that while we all vary dramatically, the interracial variance isn't any more interesting than intraracial variance. Danish soccer players have more in common with Tunisian soccer players than they do with people in the Danish philharmonic. Or, borrowing from another thread, white car-stealing meth-abusers have more in common with minority car-stealing meth-abusers than they do with neurosurgeons of any ethnicity.

Yes, but do you not see a lot of psychological investment in people blaming the gaps on racism? If race is relevant, it means there isn't really a big problem of racism.

That's an argument for never comparing averages of anything. After all, who cares if "Americans" die younger than "Japanese"? Americans who die at 80 are more like Japanese who die at 80 than they are like Americans who die at 17.

What gaps?

Nevermind, if any society can produce dumb jocks, rock stars, or rocket scientists(*), then we deserve no prejudice as the jock, musician, or scientist prove themselves in their domain.

* - pretty much any society produces great cooks, and the world has a lot of great cuisines. Note that many great cuisines come from societies that racists would rank as inferior. I find that sad, in the sense that the racists are closing themselves off from some great and available accomplishments.

no mike, it is not. It is a lesson in statistics. The average or mean are reductions, always, of overall distributions.

(The wider the distribution the less meaning average and mean have.)

You try to imply there is no genetic differences between races, which is, of course, false.
Grownups can discuss this without being racists.

I said, TMC, the interracial variance isn’t any more interesting than intraracial variance.

You know, a am reminded of another (early) "prior." When I was a teenager I scored at the 98.5th percentile on the standardized tests. It was readily apparent to me that I could not take that as any measure of worth. Yes, of every 100 people I met, 1 might have scored higher, 1 with me, and 98 scored worse ... but success and happiness in High School correlated poorly to that result.

So yes, I have been prepared from a young age to have an open mind, and to look for contributions from anyone anywhere, rather than to take some statistical truth that I am better than (almost) everyone else.

(The astute observer will note that I have "peers" from all races at the 98-99% level. I might relate more to them than the dummies who free-ride on being the same race as me.)

Or to continue the cruelty ... I might be just the sort of person who drags up "your" curve, but I don't like you.

If there's one thing that's absolutely clear in your muddled thinking, it's that you personally hate people who disagree with you. No need to reiterate that.

I'm just trying to break though your "the average matter" nonsense, mike. If I'm way up at the 98.5th percentile, shouldn't I just think it sad, pathetic, that people way down at the average look around and say "my average is better than your average?" Your average isn't even "yours." You are sponging off me.

I mean, the fact that there are white supremacist gangs in prison should be enough to destroy the idea that the average matters ... but think about it ... those gangs in prison are using people like me to make their case that they are better than other inmates.

(Switching off cruelty mode, we all have gifts and burdens. If we are humane we will be sympathetic to people who have both gifts and burdens different than our own. This average nonsense, and comparing medians of bell (or long tail) curves would be abandoned. People are people. Tom Cruz and Carl Sagan bring different things to the party, but they both make it better.)

So you really are arguing against ever calculting averages. You are a complete moron.

Averages in what, mike? I think it was Taleb who observed that some averages work better than others, for predictive power. Consider height and net worth. You can guess that Bill Gates is 5' 10.2" tall (2008 average for US males), and not be far off. You'd be off by a small percentage of the guess. But if you guessed that Bill Gates made $30,513 (2004 average for US males) you'd be off by a long shot.

Averages can matter, but not on wide data distributions, as I say above.

Here Persona explains how he acts like an idiot

What I find personally disappointing is that the racists (who I define here as those who stubbornly want to use race as a principle distinguishing characteristic) just get so dumb and sullen ... without advancing their apparent understanding.

And it is really sad that the very people who believe standardized tests, as one of their proofs, can just decide that I (who score high on them) am a "moron" or an "idiot." Why? Because I want to be open to people and their gifts, because I want to be fair, and without prejudice.

the interracial variance isn’t any more interesting than intraracial variance

But intraracial variance is fascinating and talked about all the time.

Are you sure that sane, adult, non-racists are only talking about intraracial variance?

They might be talking about the wonder of human diversity (spanning all variance). Indeed the key here is that racists have a little self-referential feedback loop. Because they keep looking at race, they keep seeing race, as "their thing." It being their thing is what makes them racist.

If you accept human diversity, and that no "race" is angel or devil, the distinction of race becomes kind of stupid and boring.

Are you sure that sane, adult, non-racists are only talking about intraracial variance?

I'm not only not sure of it, I didn't write anything like it and don't think it's true.

Intraracial variance is more interesting than interracial. That doesn't justify you shutting down your brain on a different subject.

OK, I'll give you a chance here to prove that "intraracial variance" is interesting. First I think you better define "race." I mean, there are thousands upon thousands of individual ethnicities around the globe. Doesn't the typical racist make a stupid reduction? Ie. Swedish-American and a Italian-American are "white," while a Tunisian-American and a Kenyan-American are just "black?"

Or, to give you one more chance in a different way .... Hollywood is full of actors who are not exactly rocket scientists. Despite not being rocket scientists, they all enhance our world. They come from many races (ethnicities). Why should we care about the race bit? They make the world a better place.

BTW, it is completely fair for me to mention that much of the cruelty in WWII came because the Japanese thought that they were a different "race" than the Chinese, or Koreans. And of course the Germans thought that they were a different race than the Jews. Modern racists really do have to wear that history of vanishing distinctions, as they group Japanese and Chinese as "asians" and German and Jews as "white."

OK, I’ll give you a chance here to prove that “intraracial variance” is interesting

err... I'm sort of flummoxed here. It's such an absurd question, it's hard to grapple with, I'm not going to try to prove that it's interesting, I'll just let the opinion of every human on Earth including you stand for itself.

we all vary dramatically,

This is you answering your own question.

A day before you asked it.

I did typo intra for inter there, but it really works either way. This thread has been about racism, and the defense of racists that race (as conceived by themselves) is some meaningful thing.

It should be another short disproof that "race" as seen by racists is a stupid and reductionist thing. It is based on gross phenotype, what you look like to me. Since you look white, or black, or asian, I'll just call you that and assign you to that group. God help you if you individually span a few "races" or come from an area between some "racial" boundary (Ulan Bator?). The racist will just put you in one bucket or another, and go from there.

Meanwhile, despite your incredulity, it will continue to work.

What works? The bucket sort works for the stupid sort of little racist who thinks the bucket sort works? That's just sad and self-referential.

Across the Anglosphere, they become almost identical.

This comment keeps getting repeated by the usual suspects ("lets just make a statement, keep asserting its true without ever providing evidence. Works everytime!")

Anyways, long story short, crime rate in comparable European and Anglo countries is between .5 and 2.0 per 100K. US crime rate adjusted for rate is between 2.5 and 3.0 per 100K. Notice how a similar adjustment for European countries (where it is often claimed the immigrants commit anywhere from 50 to 95% of the crimes) is never proposed either.

What are gun deaths between populations and how does this affect the death rate.

Thanks for proving my point by way of example. Not adjusting for race, age and sex makes such comparisons meaningless.

That would be an example of the racist feedback loop. If it is your prior that race is a principle distinction between people (ie. racist), then you must claim race as a character in any study of human beings.

TheAJ, do the people in the countries with a crime rate of 2 per 100k castigate themselves endlessly for being horrible backwards violent places compared to the countries with a crime rate of .5 per 100k? Because that's a bigger difference than the difference between the white American crime rate and the white European crime rate.

So basically the US gets about the same adjusted life expectancy than Canada, while spending twice as much on healthcare. Diminishing returns at their extreme... Gotta use that one in class.

And Canadians have to wait longer to die in a fatal car wreck. See, the free market works. USA! USA! USA!

Not as simple as you suggest, from the forbes comments:

That is misleading as well. For instance, we subsidize prescription
drug costs for much of the world. Other countries will put price
limits on drugs which means the money to pay for development of that
drug, and the others that failed, have to be divided amongst fewer
people… mainly Americans. There are other ways they force the costs
onto the US as well but price controls are one common mechanism. This
also applies to the drugs they develop themselves. They recoup their
costs in the US. Europeans are notorious free riders for more than
just their defense.

Most treatments other than drugs are also developed here so the
cutting edge will always be more expensive. Those results you see from
other countries would be much worse were the US not shouldering the
development costs. By the time treatments are exported, efficiencies
have been gained. If we copy European systems then there will be no
one left for us to free load off of. The Canadians in particular will
suffer since we are the safety valve that makes their system

Also, you have to take into account what is spent but not recorded. In
Japan for instance it is expected that you will pay the doctor under
the table if you want decent service because the official remuneration
is a joke. All of the hospitals would otherwise be shuttered. There is
no way to calculate that figure. Likewise you bring your own linen and
do your own laundry when you go to the hospital etc… In many countries
you pay twice by other methods that also end up off the books. SO
yeah, let’s copy the Japanese. At least then there will be fewer
middlemen and you can pay the doctor directly. You just have the
barnacles of the official system you will have to scrape off the top.
But then that won’t look much like what you are advocating will it?

Finally, costs are cut in ways that would be unacceptable in the US.
We take heroic and costly measures at the end of life because we are
rich and our system, until recently, was relatively free of
governmental interference in these decisions. Whereas the NHS for
instance just lets patients starve to death in similar circumstances
and denies them treatments that are common here. That is certainly a
cheaper way to go and the results are not all that different. But it
is a hell of a lot less humane and more importantly, you learn nothing
going forward.

And it isn’t just the end of life care that suffers. My friend in
France broke his collarbone. No one was there to set the bone and he
was told to come back. By the time someone could be scheduled to see
him many days had passed and the bone had knitted improperly. That was
certainly a savings that would not show up as a bad outcome since he
was told there was nothing to be done at that point. You have to
realize that the people giving you these statistics are the same ones
that regularly under count rapes and assaults so as to make it look
like the crime rate is going down. A businessman at least has a bottom
line that has to reflect reality even if nothing else does. He has to
compete for customers. Socialist systems acknowledge no reality but
what is politically convenient and everyone is compelled to
participate. If you read the Soviet production reports you would have
thought they had built a utopia. The reality was quite different,
though the readers of the NY Times would never have known it.

But even putting all of that aside, why the hell do you think the
people who got everything wrong about outcomes would be any more
accurate about costs once we delve into them? Were they ignorant or
did they lie? What do you think the efficiency and outcomes of other
countries would be like without the US funding their R&D for them? The
fact is we don’t know squat about how these things are calculated

It is a good comment, but I've always wondered about people who write such long comments.

I once crashed my mountain bike, and given how much it hurt, I went to the ER. The doctor there told me nothing was broken and so I went home and toughed it out. It turns out I healed a fractured hip the old fashioned way - with no expense or surgery.

Not sure what health plan would send people home who could heal on their own.

(I was later told that surgery would have been "just" to shoot a screw in, to secure the bone.)

There was a study out of Columbia that reported 60% of increases in longevity were due to medications. Most new drugs come from the US.

Canada - by forcing the price down to allow a supplier to make a profit on the manufacturing cost but not the R&D cost - receives a massive subsidy from US consumers.

That study's ridiculous. Hence, the adjusted numbers shouldn’t be seen as hard numerical estimates of life expectancy, but rather as a way of understanding the true relative ranking of the various countries on life expectancy excluding fatal injuries.? In other words, Ohsfeldt is pulling a number out of the back of his lap, and since the good old USA comes out on top, Forbes is happy to lap it up. The update is even worse, unless increased is a misprint for decreased. A number of mathematically astute readers have asked why some countries have increased average life expectancies once you take out fatal injuries.? Why of course the average life expectancies will increase if you take out fatal injuries! Younger people get killed and run over more often than older people, so they would tend to drag down the average unless the society was so safe that the contribution from oldsters falling down the stairs outweighs the former one.

Nowhere did he say he was pulling numbers out of air, just that the figures are normalized and shouldn't be used as hard and fast life expectancy averages, just as a relative rank between compared nations.

For this to matter as much as a lot of people want it too, the death from injury rate has to be independent of the quality of the health-care system, which it obviously is not. The degree of dependence is critical here.

This can be controlled for if the data is available (I don't know that it is or isn't for sure). If you die at the scene of the accident or before you reach the ER, the event is independent of the health-care system. Surely there are stats about mortality rates from classes of accidents that reach the ER.

Now, having written that, I would be literally shocked if it really were the case that American emergency trauma treatment in urban areas really wasn't near the top, or at the top, of the world rankings, and those regions where it really is subpar are likely in the sparsely populated rural areas.

Everything I've seen from the USA on this topic seems to try to explain away rather than explain. The pattern is that you pick some phenomenon that might let you make the USA top of the pops, you do some arithmetic, you draw the desired conclusion, and you don't examine evidence that might conceivably point in the other direction.

Learning anything useful from comparing figures across countries is hard; it gets no easier when you suspect that most of the writing on the subject is dominated by national amour propre. Anyway, everyone knows that all the Japanese figures are bent by people continuing to claim the pensions of their long-dead parents, don't they?

I suppose I ought to declare my own views on this topic: I'm not really convinced that life expectancy, however cleverly and honestly computed, is necessarily the best, or at least sole, measure of how well your "healthcare system" is doing its job. Personally, I put some weight on the simple observation that no country has taken it into its head to copy either the American or the British system, from which I infer that both are widely viewed as pretty poor stuff. Judging by Obamacare, they may even be unreformably poor stuff.

"The pattern is that you pick some phenomenon that might let you make the USA top of the pops, you do some arithmetic, you draw the desired conclusion, and you don’t examine evidence that might conceivably point in the other direction."

Cherry picking, even unusual metrics, to get the desired result? This is exactly what you did, here:

"I suppose I ought to declare my own views on this topic: I’m not really convinced that life expectancy, however cleverly and honestly computed, is necessarily the best, or at least sole, measure of how well your “healthcare system” is doing its job. Personally, I put some weight on the simple observation that no country has taken it into its head to copy either the American or the British system, from which I infer that both are widely viewed as pretty poor stuff. Judging by Obamacare, they may even be unreformably poor stuff."

Actually, he only 'cherry picked' the two extremes - the NHS is fairly unique (at least without considering communist nations like Cuba), as is the U.S.

Picking the two opposite poles of industrial world health care systems is not cherry picking, at least in the normal sense of the term.

"Learning anything useful from comparing figures across countries is hard; it gets no easier when you suspect that most of the writing on the subject is dominated by national amour propre."

It seems like most of the writing I see comparing America to other countries is motivated by a desire to criticize America, but I agree with the general point that these types of comparisons are extremely susceptible to bias.

The US has a different health care system than most places.

The US also has a different population than most places.

Comparing the life expectancy of Japanese to that of Americans will yield no useful data about the American health care system. The reason why the life expectancy is high in Japan is almost entirely explained by the fact that Japan is full of Japanese.

It is only comparing the life expectancy of Japanese to that of Japanese-Americans that yields useful data. Japanese-Americans do indeed live longer than Japanese.

The question is: Are there countries where the residents have significantly higher lifespans than the equivalent immigrant population in the United States? A quick google of a few large OECD countries says no.

Once you start playing this game, someone could point out that you might also want to adjust for other influences, such as differences in rates of smoking. Of course, smoking rates are probably somewhat endogenous. But the same is true of accidents and even homicide - the quality of care and how quickly you receive it have an influence on whether you die. For homicide, see here: http://people.wku.edu/james.kanan/Murder%20and%20Medicine.pdf

How is it that by removing fatal injuries some countries' life expectancy FALLS? Take Japan. They went from 78.7 to 76.0 in Avik Roy's chart. I would have assumed that these injuries are mostly to young people though... Is there a string of fatal injuries being caused to Japan's elderly? What else could cause that?

Japanese people are relatively less violent. They also don't seem to be into risk taking. In any country, an old person falling down can be a fatal injury.

Because of auto accidents and homicide, the US is probably the one developed country in which fatal accidents skew young.

@Jake -- agreed that this is a problem. And presumably what the authors talked about in the comment mocked by @AnonymousCoward above (though he jumps on the use of "increased" to mean "decreased", rather than ask the obvious next question).

Also compare Mexico and Turkey -- somehow the Turkish life expectancy increases by more than 7 years when excluding fatal accidents, while Mexico's changes by less than 2. This is implausible.

Without a lot more support and explanation of methodology, this study has to be dismissed as Yet Another Result in the Social Sciences.

"A few years back, Robert Ohsfeldt of Texas A&M and John Schneider of the University of Iowa asked the obvious question: what happens if you remove deaths from fatal injuries from the life expectancy tables? Among the 29 members of the OECD, the U.S. vaults from 19th place to…you guessed it…first. Japan, on the same adjustment, drops from first to ninth."

Is there any point to this exercise other than to make the US look better?

'Is there any point to this exercise other than to make the US look better?'

What nobler goal in life is there? Especially considering just how much work is involved.

I believe it is meant as a response to the common meme frequently parroted by one side of the political spectrum that the US pays more for healthcare with worse results than other OECD countries and point to life expectancy as their evidence of choice.

Personally I agree with Sam above, that comparing life expectancy across entire nations is a bit simplistic since different nations are full of different people and it would be more effective to compare Danes with Danish-Americans and Japanese with Japanese-Americans but this is vulnerable to selection bias unless controlled for (only rich, healthy people emigrate perhaps), but the result is the same, the US frequently comes out on top or near to it, which doesn't agree with the priors of some people and hence must be BS.

'the US pays more for healthcare'

It does - by a truly noticeable margin. 'Total national spending on health care in the United States by both the public and private sectors was $2.6 trillion in 2010, or nearly 18 percent of total economic output (GDP).' - http://pgpf.org/Chart-Archive/0006_health-care-oecd

'parroted by one side of the political spectrum that the US pays more for healthcare with worse results than other OECD countries'

Well, it is true that the Peter G. Peterson Foundation was the source of the above quote, and they even wrote this - 'Americans currently pay about twice as much per capita on health care as our peers do in other advanced nations, yet our health outcomes are no better.' Though the PGPF does claim to be non-partisan, generally most billionaires (or former Nixon appointees) interested in cutting Social Security, Medicaid, and Medicare are not considered to be on the 'left.'

Oddly, no one in the rest of the world has any interest in this putatively better American system of health care. Possibly because people in other countries have experience with a functioning health care system that doesn't bankrupt them.

The debate isn't whether we spend more, that is somewhat easily measured, but that the outcomes are worse as well.

Also, where is your evidence that the rest of the world isn't interested in our system. It is my understanding that along with Japan, the US is a common destination for advanced or risky treatments.

Oddly, so is Germany, especially Heidelberg's quite extensive network of institutes, research centers, and treatment facilities. The thing is, there are certain places which tend to attract high paying international patients, and it is not precisely based on where the facilities are located in terms of national boundaries.

On the other hand, I don't think too many international patients are travelling to rural America for the best in medical care. Or to a typical American city hospital, for that matter.

"Oddly, no one in the rest of the world has any interest in this putatively better American system of health care."

This is wildly inaccurate. The rest of the developed world looks to the US for best practices, guidelines, protocols, research and quality improvement results, etc. While other countries may not follow the US's lead in financing healthcare, they absolutely do in providing it, particularly in acute care.

"Is there any point to this exercise other than to make the US look better?"
To figure out how the healthcare system is performing?

I'm not sure cancer survival duration is a great metric for health care strength. It completely overlooks preventative health.

Has preventative health been shown to be a great metric in any way?

Not smoking seems to be effective.

...and that has what to do with the medical system?


Interesting. Why filter out death from injuries? Are hospitals not expected to treat the injured? How about procedures that cost more in the U.S. for the same results? There are all kinds of reasons for all kinds of things but if I get sick what matters to me is whether I can expect to get efficient affordable treatment.

In the end, rather than trying to manipulate the data to come up with a rationalization we're more comfortable with, how about applying our data analysis skills to figuring out what works better and cheaper?

The purpose is to rebut those who claim that Japan's higher life expectancy means that Japan's medical care regime "works better and cheaper." Rebutting false and widespread claims is at least as important as advancing new theories. Oh and the reason it's valid: if some ghetto hood thug bashes your skull into a concrete sidewalk until you're dead then it's not the healthcare system's fault.

Maybe if you didn't exclude death from injuries you might have to look at gun deaths in both populations?

Automobile deaths is a greater contributor, but I'm not sure we're NOT looking at gun deaths, just watch Pierce Morgan's show after a shooting, it seems like that is all some people are looking at.

It is death from fatal injuries, not treatable ones.

This part I missed. So you are saying the it is only people who never reached the hospital, or literally coudn't have been saved by anything other than an act of God?

Sure, hospitals are expected to treat traumatic injuries, but if in doing a cross country comparison you are comparing aggregate figures of place with lots of traumatic injuries vs much less traumatic injuries, it is a disingenuous comparison. That so many commenters are miffed if not indignant that the study attempted to control variables is lost on me. This is pretty common stuff in statistical analysis.

More to the point, while the U.S. may have more trauma deaths than its peers because there are more trauma incidents, the areas where U.S. mortality is exceptional (trauma, infant mortality) are also the areas where the affected populations have the least access to quality health care. http://washparkprophet.blogspot.com/2013/11/trauma-is-main-factor-driving-down-us.html

There are a lot of negative things correlated with the areas of high mortality that have little to do with healthcare, unless you're going to draw a causal link it helps little to bring them up.

You'd like to capture the effect of the way we have our healthcare system organized, without also capturing the effect of, say, how much we drive, or what our ethnic mix looks like, or how many guns we have.

Something similar comes up with crime rates vs laws (gun control, harshness of sentences, drug laws, etc)--it's quite hard to separate the effects of the laws from the effects of other stuff when looking at, say, murder rates across countries.

The same sort of issue comes up with assessing educational systems. Are our schools all that much worse than Finland's schools? It probably depends on whether you take ethnic mix into account--white and Asian kids in the US apparently do pretty well on those tests, but blacks and hispanics do a lot worse.

And this raises the really hard question: how do you know when you have accounted for all the important differences? It's easy to just keep looking for different ways to analyze your data till you come to a desired conclusion. What's not so obvious is when you should stop trying to reanalyze your data. (Think of the schooling example. Looking only at whites' test scores eliminates the effect of the racial mix differing across countries, but also means you lose the ability to detect a school system that needlessly shafts minority kids.)

Never really, but at least regression analysis I'd say when your coefficients are high (explain 80-90+% of variation in data) and your confidence interval is also high (95+%). This is not really a regression analysis study so much as just raw data normalization.

A couple of points:

1. Is it possible that the fatal injuries might not have been fatal had the US had better health care for all?
2. The author says that private insurance is better than public insurance in the US. Maybe so, but that is because the rich can afford to spend money on themselves. Even so, if it is so good, why haven't we figured out how to get everyone private insurance? The Republicans have no answer. The Democrats have tried. IMHO, it is better to have everyone with some health coverage rather than the rich with great coverage and the poor with none...

It may be possible, but that's just hypothesizing at this point. As far as I can discern, this study didn't factor in a comparison of injury treatment efficacy. With that said, I don't see how payment structures would really pertain to what would most likely be ER situations. In other words, what does having a single payer system or private third payer, or no third payer system have to do with a gunshot victim on the operating table?

Lastly, I doubt other countries have better ERs, but it could be.

I think the general assumption in the minds of people who use the Orwellian term "single payer" is that if we just adopted "single payer" then every person everywhere all the time would get the best doctor ever and all the coolest medicines and devices and rainbows and ice cream and unicorns and gay people getting married and everyone all across the world loving and caring and holding hands and singing in perfect harmony the most beautiful song you've ever heard and a cute little baby smiling and laughing

I think I figured it out. Rather than "taking out" fatal injuries, he's controlling for fatal injuries by giving all countries the "average rate" of fatal injuries. So for countries that in reality had below-normal rates of fatal injuries, he's actually postulating that some people would die younger than they did.

To summarize from many of the comments above,,

This is a piece of uncontrolled BS

Unless you are inclined to agree with it.

I wonder what reality is and how you could actually construct a rigorous study rather than a PowerPoint presentation that gets picked up in blogs.

It may or may not be BS, but it appears that the real problem is that Avik Roy either didn't understand the precise methodology or described it incorrectly. Specifically saying "take out" rather than "control for".

Mike seems more close to the truth than Bill here. The study doesn't seem to be the problem so much as the journalist's interpretation, which is SOP.

....How can one asking for the truth be closer to the truth than another ?

I think Russia should get life expectancy numbers controlled for vodka.

If the purpose of the statistic is to gauge the effectiveness of the medical system, then yes

Oh, I do not like to come on a member of the endogeneity taliban, but this study cited by Roy screams out for an intervention. It looks to me as pretty obvious that there would be a negative correlation between expected life expectancy without death be accident or gunfire, but this is not accounted for. A big confound would be liquor use, with half of fatal automobile accident deaths involving DUI. Presumably the people causing these are more likely to be drinking to excess and thus more likely to die younger if they managed to avoid dying in a car accident they caused. Also, gun deaths, many of which involve either suicides or criminal activities. Suicides are generally depressed, and even if they avoid suicide, depresson tends to shorten life expectancy. Likewise, criminals are more likely to be absing either or both alcohol or drugs, thus also likely to have shorter life expectancies if they survived their gun battles.

This is pretty obvious, and that apparently this paper made no correction for this academically disingenous and shameful, typical of Roy, who seems to be mostly a third rate propagandist.

Whether Avik Roy's data is accurate or not, life expectancy as a proxy for health care system efficacy is still hugely flawed. We know this because even within countries there are huge differences. Last I checked the difference in life expectancy between Hawaii and Washington DC was 10 years. Life expectancy between England and Scotland is 2-3 years depending on gender, and in Canada life expectancy differs by 3 years between Newfoundland/Labrador and British Columbia -- anyone want to tell me this is because of the health care system?

As a matter of fact, Colin, there are a lot of systemic differences between states in terms of the quality of health care. Just to pick one obvious one, there are huge variations in the rates health insurance coverage. So, about 98% of the population in Massachusetts is covered because of Romneycare, whereas in Texas only about 74% of the population is covered. There are lots of other differences.

In other nations, where there is universal coverage, there are many fewer such differences.

These differences predate Romneycare. And we can take it to a more granular level. In Baldwin County, Alabama female life expectancy is 80.4 years while in Dallas County, AL it's 75.6. In Kern County, California the average is 79 years for females, while in Santa Clara it's 84.6. Again, you want to tell me these are due to the difference in health care systems? As for universal systems, let's note that male life expectancy in Glasgow, Scotland is 71 while in London's *worst* borough, Islington, it's 75.4 (and 84.4 in Kensington, it's best borough). I think it's fairly obvious at this point that life expectancy is, at the very least, a hugely problematic proxy for health care system efficacy.

In any case it's always good to hear from a former professor of mine.

Always a pleasure.
I most certainly never said (and do not believe) that nature of health care system is the top determinant of health outcomes. But your data is easily explained by a well-known variable that does invfluence them, namely per capita income levels, at least for the CA counties. I know nothing (and did not check) about the AL ones.

As it is, this is part of what is so embarrassing. Here we are, one of the highest per capita income nations in the world and spending far more than anybody else (gist of Arnold Kling's comment), but we are getting lousy health outcomes, unless, that is, you believe the unpublished paper by Ohsfeldt and Schneider that is fraught with screaming methodological errors that severely bias their results.

But your data is easily explained by a well-known variable that does invfluence them, namely per capita income levels, at least for the CA counties

That doesn't seem clear at all to me. Maryland has the highest per capita income at around $70K, yet ranks only #26 for life expectancy. The second-richest state, Alaska, ranks #34, and the fourth-richest, DC (Ok, not a state) at #4 (tied with CT) is #43 for life expectancy.

Here's something else that's interesting:


Note that the 214 "million middle Americans" have a per capita income of $24,640 -- about $3K higher than Asian Americans -- and yet their average life expectancy of 78 is a full seven years less than those same Asians with lower per capita incomes. Also note that Native Americans live two years longer than Southern poor blacks despite having slightly lower incomes.

In Canada, Toronto's median total income is $69.7K vs. $68.9K for Vancouver (source: http://www.statcan.gc.ca/tables-tableaux/sum-som/l01/cst01/famil107a-eng.htm) yet per my googling Vancouverites live two years longer at 81.1 years vs. 79 for Toronto. To place that two year gap in perspective, if the US added two years to its life expectancy we would jump from #33 to tied for 17th along with such countries as Netherlands, Norway and Germany. Now, coincidentally -- or perhaps not -- Vancouver's East Asian population is 28% vs. around 12.5% for Toronto. Is it also coincidence that the US state with the highest percentage of Asian Americans, Hawaii, also has the highest life expectancy?

Seems to me that demographics play a pretty significant role in determining life expectancy as well, and given the wildly different demographic profile the US has vs. countries like Japan and Iceland (both tied for #1 on life expectancy at 83 years -- about the same as Hawaii at 82.7 years) it surprises me not at all that our life expectancy is different.

Again, the point of all this is that life expectancy is a pretty poor proxy for health care system efficacy. Given your stated belief that health care systems are not the top determinant of health outcomes -- which logically means you think there are other factors at play -- then I am guessing you must agree with me on this point.

As for the US health care system, I believe that it is wildly inefficient, but it is less clear to me we are getting lousy health outcomes (at least in absolute terms -- on a relative cost effectiveness basis I have no problem believing we are lousy).

BTW, I looked at the paper, but I saw no sign that it has been published. Perhaps editors and referees at journals the authors have submitted it to have noticed some of the endogeneity problems I pointed out above, and that is why the paper is not published, that serious academics recognize it is seriously flawed, which it is.

I made it look like Roy did the study, but he did not. But he has a bad track record of publicizing garbage studies like this one.

1) Tyler, why do you point us to a powerpoint ? a PPT is a aid to a speech, not something to be read
2) This powerpoint is agood example of how to lie with statistics; they point us to cancer (we are good) but not diabetic amputations (we are bad)
3) the authors leave japan out of table 1-6 (cancer survival rates)...wonder why (Hint: it doesn't make the us look good)
4) anyway you slice it, we have more dead babies. anyway you slice it, a kid under 5 dying ain't good
5) several people say we subsidize the world with pharmaceutical RnD; I personally think this is nonsense, cause most of the new drugs don't work that well, and other countries do a lot of RnD

Lots of foreign companies do R&D in their home countries and recoup their costs selling to the American market.

"It has been claimed (Ohsfeldt and Schneider, 2006) that adjusting for the higher death rate from accident or injury in the United States over 1980-99 than the OECD average would increase US life expectancy at birth from 18th out of 29 OECD countries to the highest. In fact, what the panel regression estimated by these authors shows is that predicted life expectancy at birth based on US GDP per capita and OECD average death rates from these causes is the highest in the OECD. The adjustment for the gap in injury death rates between the United States and the OECD average alone only increases life expectancy at birth marginally, from 19th among 29 countries on average over 1980-1999 to 17th. Hence, the high ranking of adjusted life expectancy at birth mainly reflects high US GDP per capita, not the effects of unusually high death rates from accident of injury. For information, the most recent data (which were used to make these calculations) on average standardised death rates per 100 000 population from accident or injury over 1980-1999 for land transport, suicides, homicides, and falls, respectively are 17.4, 11.4, 9.2, and 4.6 for the United States and 15.5, 13.3, 3.0, and 10.4 for the OECD average (OECD Health Data, 2008). Life expectancy at birth on average over 1980-1999 was 75.3 years for both the United States and the OECD average (29 countries) (OECD Health Data, 2008)."

Carey, D., B. Herring and P. Lenain (2009), “Health Care Reform in the United States”, OECD Economics Department Working Papers, No. 665, OECD Publishing.

OK, folks, I went and checked. This 2006 paper by Ohsfeldt and Schneider has never been published anywhere, although its "findings" have been quoted far and wide in places like the WSJ and Commentary, blah blah. They did publish an article in 2007 the Cumberland Law Review on the role of competition in health care and also a book on that topic in 2006, but never the paper that is the basis of this whole post. BTW, Ohsfleldt has a very extensive publication record, including many papers with numerous coauthors, sometimes including also Schneider. So, I seriously doubt that the failure to publish this paper is due to some lack of effort on their part. The paper has very serious problems, which I at least here have pointed out. Garbage squared.

I am going to hammer this worthless study one last time. There is reason to believe that they in fact structured it to make a propagandistic point and overstate the supposed effectiveness of the US health care system.

A highlight is the cancer survival rate chart, where the US is tops. Indeed, the US health care system is excellent in certain areas, with cancer being the most important. The US is tops overall. However, the second best performer is France. Look at that table. France is not on there. Why is that? Well, leaving France off sure makes the US look even more spectacularly good than it is. Did the authors consciously leave France off to make such an exaggerated appearance? I do not know. But, I do not think they are stupid.

Oh, to Colin, no, of course I do not think income explains everything although it looks like I said that. It explains a lot, and your isolated anecdotes picking this city or state do not disprove it. Looking at all data and accounting for the many other factors involved, of which there are many, shows that per capita income is on average an important independent element, even if there are outliers. And, no, I do not believe that there is one single explanation for health. Colin, you of all people should know I am not that stupid.

What is pathetic is how many people here were taking this stupid and apparently unpublishable in a serious journal study seriously. Gag. Be ashamed of yourselves, you all!

Outliers? Colin listed the the top 4 states by income and showed that they don't appear near the highest on life expectancy. Marginalizing these as "outliers" doesn't make you correct unless you want to actually provide any data to back up your side whatsoever. What Colin is saying is that life expectancy is more complicated than just correlating with income and calling it a day, and perhaps demographic makeup may be more of a contributor to life expectancy outcomes when comparing two nations or areas within nations.

Sorry, Jay, but the studies on this are legion. Try this widely studied paper from 2008 in the Journal of Economic Perspectives by Angus Deaton (giving all this info if link does not work), "Income, Health, and Well-Being around the World: Evidence from the Gallup Poll," with there being lots of other studies confirming the basic findings. This is just boringly silly and ridiculous. http://www.ncbi.nlm.nih.gov/pmc/article/PMC2680297 . Really guys, do not make fools of yourselves by pushing silly nonsense. Outliers exist even within strong relationships, and the income-health is one of those strong relationships.

Comments for this post are closed