The Spirit Level Delusion

The Spirit Level by Wilkinson and Pickett made a big splash a decade ago by showing many correlations between inequality and various problems. In a recent talk, Pickett summarized the thesis of the book with the graph at right.

Even at the time, however, there were peculiarities in the data–for example, some countries were dropped without explanation and data with different definitions were spliced together–as pointed out by Christopher Snowdon in the Spirit Level Delusion. Adding in a few more countries, for example, made many of the correlations disappear.

Snowdon now has a nice post with another test. Suppose we run the same or similar regressions using today’s data? If the relationships are robust we ought to see the same correlations or even stronger correlations given the increase in inequality.

Here, for example, is a key graph from The Spirit Level on inequality and life-expectancy.

If, however, we use the same countries but today’s most up-to-date figures on inequality and life-expectancy we find no correlation.

If we add the four countries that are unequivocally richer than Portugal and were excluded from The Spirit Level for no good reason (South Korea, Hong Kong, Slovenia and the Czech Republic (now known as Czechia)), there is a statistically significant association with inequality but it is in the opposite direction to that predicted by The Spirit Level hypothesis, with greater inequality correlating with longer life expectancy (r2=0.145, R=0.385, p-value=0.0495).

After examining a variety of data. Snowdon summarizes:

In summary, most of the biggest claims made by Wilkinson and Pickett in The Spirit Level look even weaker today than they did when the book was published. Only one of the six associations stand up under W & P’s own methodology and none of them stand up when the full range of countries is analysed. In the case of life expectancy – the very flagship of The Spirit Level – the statistical association is the opposite of what the hypothesis predicts.

If The Spirit Level hypothesis were correct, it would produce robust and consistent results over time as the underlying data changes. Instead, it seems to be extremely fragile, only working when a very specific set of statistics are applied to a carefully selected list of countries.

It’s certainly possible that inequality has a causal effect on various issues (both positive and negative) but it seems that such effects are small and subtle enough to require much more than cross-sectional country-level data to uncover.


Wonder what explains this - 'Life expectancy in Britain for both men and women has fallen by six months compared to last year, figures have revealed.

Statistics from the Institute of Faculty of Actuaries, which calculates life expectancy for pension providers, expects men to live until 86.9 years and women to 89.2 years.

This is a drop from 87.4 years for men and 89.7 years for women in 2017, the projections reveal-'

This is a forecast, not actual life expectancy. So a change in assumptions not a change in reality.

Good point - wonder if anyone who actually uses life expectancy as a measure who ever thought of it.

'So a change in assumptions not a change in reality.'

Well, it is fair to say that basically we all be dead before seeing whether the change is actually reality. However, a lot of real things are based on such assumptions - such as the cost of various forms of insurance and annuities.

Any Western country that brings in 3rd world people will experience a decrease in life expectancy.

Also it is a statistical error to compare large population countries with small population countries. It is also a mistake to compare countries with a single culture/ethnicity, like Japan, with a country with many and multiple cultures and ethnicities, like the U.S.

Usually when a study or a professor makes these statistical mistakes it is intentional and the intent is to misinform for political reasons. Often a study like this is followed up with some statement about obesity or diet, etc.

It's all balls, apparently, an artefact produced by a cocked up census. (How in God's name do you cock up a census? The first census in the neck of the woods where I grew up happened almost 2000 years ago.)

Herewith I link the linked link.

That first picture is absurd. You can’t have 1/2 a problem. It’s more a doodle than a graph.

The whole thing is absurd. They are measuring inequality in rich countries, which means they are measuring Number of Poor People.

Problems go up for poor people, eh? Groundbreaking stuff.

Indeed, there are problems with their assumptions, almost as though they picked their conclusion first, and then threw together something to justify it. Imagine a country with structure in the population, where one segment is high mean ability, and another low mean ability. We'd expect both inequality and poor health outcomes to be higher, than if the population were "medium ability". Imagine moving everyone in the Sudan to Japan.

I'm doubtful a multidimensional concept like "inequality" can be adequately summarized by a single factor, and certainly the policy implications are unclear. Welcome to modern economics, hordes of midwits wasting their times going through the ritual of writing papers. There are maybe 10 people alive doing worthwhile economic research.

“There are maybe 10 people alive doing worthwhile economic research.”

I’m in History, not Econ., so if you have any names, I’d be interested in reading several of this group of ~10.

But we won't know which ten until they are dead.

For at least 30 years...

Yes, it's a punchline presented as such.

The Spirit Level was rubbish from the start. But the behaviour of its Authors after publication really reeked of bad faith. If you didn't know it was rubbish at the start, you could tell it by the way Wilkinson and Pickett behaved.

But it was a very useful book. The book, and the fevered left-wing commentary around it, established 3 things:

1. The techno-policy elite of the Left was really bad / innumerate / incompetent at social science and statistics.
2. The Left didn't have any scientific support for it's vision of society.
3. Virtue signalling destroys minds and effective enquiry.

So it was a useful book, in a way...

1. The techno-policy elite of the Left was really bad / innumerate / incompetent at social science and statistics.

That may be. The trouble is, there is no alternative 'techno-policy elite' because dissenters are not welcome in graduate programs. The closest you get to a dissenter would be the pair who moderate this blog.

2. The Left didn't have any scientific support for it's vision of society.

Scientific support is only of cardinal interest when the source of a dispute centers on the practical implications of a given norm, though it certainly doesn't help your argument to be trafficking in spurious correlations. It's hard to think of anyone who is much good at arguing normative questions. Michael Walzer?

Come to Russia, in any University (I have Moscow in mind) you will be surprised how many techno-policy elites are very pro-authoritarianism. It will be refreshing.

All techno policy elites are authoritarian, the variation is all in how overt they are

Here is a left wing take down of 'The Spirit Level' by far left economist John Quiggin.

Quiggin points out that the effects of inequality can be looked at over time as well and this also undoes the conclusions of The Spirit Level.

I can not help noticing the American data point way below the other developed countries? Has a nuclear war taken place?

My understanding that the US numbers include very premature births as the point of life at birth and other countries don't. Babies born at < 1 kg (about 23 or so weeks) are at very high risk and whether they are counted as births would make a huge difference in the life span statistic. One near stillbirth (a C section got it out alive) will pull 9 people who live to 90 down to an 81 average.

It may be. I would like to have the numbers.

The vast majority of the life expectancy gap occurs before age 50. The major drivers are: homicide, transportation (e.g. MVC), poisoning (i.e. opiates), infectious disease (e.g. AIDS, gonorrhea, influenza), and deaths surrounding births.

Homicide is complicated, some folks like to say that it varies with gun ownership, but the R value is around 0.003 between US states and around 0.02 for European countries. Most likely, the ultimate driver for American homicide rates is cultural (e.g. perhaps a legacy of racism, perhaps a by product of urban underclass culture, perhaps a legacy of frontierism).

Transportation is pretty simply because Americans drive a lot more than Europeans. Deaths per billion km driven puts the US below France and Japan. You could argue that the US needs more trains and for Americans to live lives confined to dense urban areas ... but the current build out of American housing cannot support that.

Poisoning is driven almost entirely by opiates. I blame this largely on the "pain is a vital sign" nonsense and our utter destruction of a centuries old cultural-religious taboo regarding pain med addiction. Shockingly when we liberalized our cultural habits regarding use of addicting substances for pain control we saw a massive spike in addiction and overdose. When we beat back the prescription drugs, well fentanyl is hard to dose and easy to poison. It almost like we tried this exact same approach in the 1860s and 70s to the exact same results.

Infectious disease deaths are vastly more common among those with mental health issues. In the US it is vastly harder to commit some one for inpatient care against their will. So we now have freakishly easy to treat diseases (e.g. syphilis) running rampant among homeless individuals who refuse to get care. A lot of these chronic infections are not lethal per se, but then when you superimpose something like influenza your body is overwhelmed. This is made worse by all the perinatal stuff below.

Births are tricky. Who dies at birth is mostly a function of who lives to birth. In the US a huge number of mothers refuse any sort of "therapeutic abortion" for things like trisomies, anencephaly, and the like. These children range for dead with days with no help of medical improvement to dead in their 50s. In France, for instance, we have seen a giant drop in the number of children born with these issues. Given French maternal ages, they actually should have a much higher rate, but mothers in France are far more prone to get screening and not carry the pregnancy when these issues are diagnosed. Another major issue is that a number of countries will not code very low life expectancy births as births. Anencephaly is universally fatal within days at most, yet many countries do not even record it as a live birth and even when the official regs require it be recorded as such, many physicians still will not follow through. The relatively most dangerous time to be an American is during the first month of life, this in spite of the fact that America has the generally acknowledged best neonatal care in the world (i.e. every centrally planned health system in the world has lower rates of pediatric hospitals, specialists, and far less capital investment).

Add onto this the fact that Americans are more obese, have smoked more pack-years, and get less physical exercise ... the wonder isn't that Americans die sooner; it is why isn't the gap larger?

The U.S. has about the lowest percentage of smokers among the countries compare at 14%. In France, 27% smoke and Sweden is the only country in Europe where a fewer percentage of people smoke at 7%.
In Japan, 16% smoke and in South Korea, 18% smoke - about the same as in the U.K.

In terms of per capita consumption, the U.S. is at the low end there, too, compared to the EU, Japan and South Korea.

The relevant question is not who smokes now, but how many pack-years people used to smoke. Lung cancer has major lag with smoking.

US pack-years peaked sooner and higher than in Europe. Fairly soon the old US smokers will die off (e.g. all the GIs who started in WWII are mostly gone) and then Europe will take the lead. However, that day is not yet here.

Europe, Japan, and South Korea were all poor in the 50s so their peak smoking years are just beginning to enter the long term death statistics. The US was rich then and we are really paying for all the smokers from the 60s and 70s right now.

"Europe, Japan, and South Korea were all poor in the 50s so their peak smoking years are just beginning to enter the long term death statistics. The US was rich then and we are really paying for all the smokers from the 60s and 70s right now."

Neither part of this is correct. It also didn't matter how poor Japanese and Koreans were in the 1950s, they smoked like chimneys and health effects from that were showing up in the 1970s and more so in the 1980s.

Per capita cigarette consumption, empirically is at 1930s US levels in Japan in the 1960s, half the consumption (google is plentiful), and that probably concentrated in a subset of males. Sure seems correct in broad picture though would've been more so in 2000.

Smoking rates also seem different from what you suggest - Japan - 21.7, USA - 17.2, Sweden - 20.6, although France - 27.6 and South Korea - 26.7.

Your smoking rates data is old - look up 2018.

I see a graph that shows in 1958 to 1968, 80% of Japanese men smoked and between 10% and 15% of women smoked.
(fig. 2)

Do you have a historical graph for per capita? The percentage of smokers would have a larger affect on health. (Japanese women have always smoked less per day than Japanese men.)

Might be a difference in the rates over a few years (2015 v 2018), but historical rates are the main topic here. Though if you've got a 2018 cross-country comparison, I'll look at it, but not comfortable cobbling together different data sources with different definitions and methodology.

Data on number of % daily smokers only seems to go to 1980 -

Harder to keep track of than overall cigarette sales volumes I'd guess, there might not be the cross-country comparable survey dataset to work it out (international market research or international health organizations probably needed to get good shared definitions). I wouldn't expect the % daily smokers to diverge heavily from the per capita cigarette consumption, though it can (e.g. France - fair numbers of people smoking their Gauloise after dinner with a wine but consumption doesn't go even above 7 per day per capita even at peak).

Even in the case of France, cigarette consumption per adult has always been in the lower leagues, and continues to be, despite high rates of smokers -

The US lead the world in per capita cigarette consumption until 1983, at which time Japan overtook the US. It only fell below western Europe in per capita cigarette consumption in the late 90s. For most of of the elderly's lives, Americans smoke significantly more than anyone else. We also had much higher female smoking rates. The curve is bending so that in a few years the American cigarette burden will be below OECD average, but as said above, we are not there yet.

So it was about 45 years ago when Japan overtook the US in cigarette smoking - surely the effects of this on Japanese longevity would be obvious by now.

Myself I think the low US life expectancy vs other developed countries is a very multi-factor result. It is definitely not due to the problem of over recording of live births as many Americans seem to believe, life expectancy is lower at almost every age in the US. There is no plausible reason 10 year olds have lower life expectancy in the US vs other countries that are due to birth recording effect. Most likely the health system is part of the problem - it is obviously true that there are some people who do not go to the doctor in the US because they lack any kind of insurance so that is part of it. It is also true I think of the opposite problem - too much medical treatment due to bad incentives (hospitals are dangerous places lots of people die there).

Availability of drugs is another aspect of US society that is very different to other European ones, both legal and illegal - more drugs definitely leads to more deaths if only due to accidents.

Another aspect that I believe causes this result of lower US life expectancy is an unhealthy attitude to alcohol. In most European countries people abuse alcohol in public settings, in the US, because of the shame factor, they do it in private which makes it much more dangerous.

Finally obesity and general lack of exercise must be part of the reason, which I think is due to the car culture of the US vs European public transport systems, and the cheap restaurants. If you go from house to work in a car you are never as aware of your lack of fitness as if you have to run to catch a bus. And if most people around you are fat, then you will face less issues being fat.

People in the US eat out vastly more than in Europe I find and restaurant food is particularly bad for obesity as you don't control the portions, plus the food itself tends to be high fat/sugar.

Oh and guns are certainly part of the problem. Suicides and drive by shootings by guns are much higher in the US, no question.

Just on smoking, per capita deaths "caused by smoking" still seem to have US edging out Japan -

US population is slightly under 3x Japan, and "deaths caused by smoking" slightly over 3x Japan.

Adjust for older Japan than USA and US probably still has a bit more of an edge in deaths, still. But certainly they've evened out since 1990/2000.

Though less so than in 1990.

US has a higher % of deaths from cancer attributable to smoking though - Probably due to patterns in use (smoking patterns in US more cancer, Japan more emphysema), and greater stomach and liver cancer in Japan (in US lung, breast and prostate are relatively more significant).

36 years ago Japan overtook the US in smoking rate. This does not mean they overtook the US in pack-years at that time; pack-years are cumulative over a smoker's lifetime and they likely overtook the US a decade or two after they reached parity. Think of it like money, just becomes income are equal one year, it does not follow that wealth also equalizes that year.

Cancer lags smoking by another decade or two. So it looks like the peak relative risk of cancer for Americans from cigarettes peaked a few years ago (sometime from 2000 - 2010 would be my off the cuff guess) and we are reverting to the mean from a highly elevated relative risk.

At the present moment, we are still edging out Japan not just on lung cancer, but also on things like COPD. Outside of Japan, the US has a much larger gap when it comes to pack-year exposures, Western Europe peaked later, again, and at much lower year pack-year consumption rates.

Percent of the population that "smokes" is a pretty poor measure. The world's oldest recorded living person smoked something like a couple of cigarettes daily. Medically, we use "pack-years" because the chain smoker who goes through 2 packs a day really does have close to double a pack-a-day user and because a quarter pack-a-day user really does have 25% of the risk.

I agree that US life expectancy is multifactorial, hence why my link is directly to graphs of the most impactful causes of death. But part of that multifactorial analysis is indeed smoking.

Guns as suicide does not fit with the data. South Korea and Japan, for instance both have higher suicide rates with far fewer guns. Yes guns make impulse suicides far more likely to complete, but on the flip side The American suicide rate is not terribly different than that of France or Sweden.

For homicide, even if you eliminate all the gun deaths, the US still leads all its peer competitors without eliminating their gun deaths. Only about 1/2 - 2/3rds of murders in the US use firearms and the US homicide rate is such that merely the people who are stabbed or bludgeoned to death is higher than most peer countries murder rate.

Guns simply are not a cost effective thing to go after in the states. An Australian style buyback program would need to be multiplicatively more expense; use weapon stocks are valued in the hundred billions. For that type of cash we can purchase far more QALYs with just about anything.

The truth is that America has been a more violent place before anyone began heavily restricting guns and it continues to be today. Given that virtually all the GSWs I see come from people engaged in illegal behavior, I strongly suspect that even getting to Japanese level gun control will not move the homicide needle all that much until we also adopt Japanese norms regarding criminal violence (e.g. highly structured with strong prohibitions on endangering the innocent).

US low life expectancy is easy to understand: Americans eat too much and too much burgers, fries and soda. If you look at life expectancy of ethnic groups inside the US that do not share the mainstream American diet, their life expectancy is much higher: Asian Americans live longer than any country, at 86.7 years (, because they eat healthy food. Hispanic Americans also live way longer than the average thanks to their Latino eating habits.

That is a part, but when you look at causes of death and the actuarial impact the big leaders also include things like traffic fatalities. It is always hilarious to me that when you look at actual causes of death, the US stands out far more for boring stuff - like traffic fatalities, diabetes, and neonatal deaths - than the headline grabbing things like guns and maternal mortality.

"I strongly suspect that even getting to Japanese level gun control will not move the homicide needle all that much until we also adopt Japanese norms regarding criminal violence"

That's a reasonable claim, but would note for the posterity of comments (hey, they're probably being archived by USG) that Japanese norms are not Japanese norms since time immemorial - early 20th century overall rates are about the same, which probably means White American homicide rate was *lower* than Japanese. That makes it sort of tricky to take any of that back to "frontier culture" or the like.

Very interesting. I have always read the high infant fatality rate as a criticism for American medicine not providing adequate care to the poor, but it turns out it is almost exactly the opposite problem. Amazing.

It is pretty sad.

Thanks for taking the time to respond, Sure.

Great comment, Sure. Very informative.

+1 to Sure for that. Revised my opinion on US death data.

"Most research on life expectancy centers on the 50 and older population and their mortality. However, much of the expectancy gap between the U.S. and other countries is due to mortality at younger ages, which leads to many decades of potential life being lost. This early mortality is dominated by injury deaths ..."

If you don't mind my generalizing, addressing these jobs would require government action, but far too many people believe "freedom" is more important.

Free to die younger than teh socialists.


I wasn’t aware of the constituency looking to legalize homicide, assault and reckless driving.

If we cared about welfare over freedom we would move the bar a bit on guns, certainly.

Why not a shooter's license, analogous to a driver's license? Use it to buy guns and ammunition, have it handy when you are operating a gun, or suffer consequences.

There are approx 2,700,000 deaths in the United States annually.

About 9,000 of these were firearm homicides, so Approx 0.33% of deaths per year in the US.

So compared to over 650,000 for heart disease, or 70,000 for drug overdose, or 37,000 for vehicle accidents.

How many accidental gun deaths? How many gun suicides (yes I know suicides can find other means but there's pretty good literature on the fact that many impulsive gun suicides would actually not kill themselves without a gun around to make it easy)?

That said, guns are not the only low hanging fruit we could work on to reduce untimely death.

Numbers I’ve seen were 23k in 2016 for suicide-involving-firearm, and “<500 annual deaths” from accidents-involving-firearm.

Can’t find any stats which break down legal/illegal possession for accidents and suicides.

Sure, we can concentrate on 9,000 of the 2,700,000 deaths. The problem is, there’s no constituency for it.

You could stop and frisk likely illegal gun possessors, but that makes 50% of the country go crazy because it sends people to prison and there’s a race angle.

The liberal response instead is to make guns illegal, which infuriates 70% of the population since law abiding gun owners are punished because liberals don’t like stop and frisk.

My answer would be to:

1) make a law that says 3 net tax paying adults have to vouch for an individual to buy a gun. It’s private and carries no liability but can be subpoenaed. Any militia would require upstanding citizens to vouch for membership. If three net taxpayers don’t think you should own a gun, you probably shouldn’t own one.

2) stop and frisk in all high crime areas

Extend #1. Anyone can have a gun. But they have to have public indemnity insurance for it, like with cars.

Let the market figure out who can be trusted.

Maryland has had that since 2013. Murder rates in Baltimore are up since then.

"If we cared about welfare over freedom we would move the bar a bit on guns, certainly."

You left out the "not" at the end there

Such is life in Trump's America. So sending food and medical supplies as we are sending to Venezuela in ita time of need wouldn't help, I guess.

The kindle edition is priced at $10.08. Why eight cents? Why not a nickel or a dime? Is this Amazon's version of satire? Anyway, generalizations about the social consequences of high or low levels of inequality strike me as misguided. Since the correlation between inequality and life expectancy was an important finding in the book, is the declining life expectancy in the U.S. since the publication of the book due to a rising level of inequality or a rising use/abuse of alcohol and opioids? Is the rising use/abuse of alcohol and opioids due to the rising level of inequality or the de-industrialization of America (i.e., the loss of high paying jobs for less educated workers)? Is the de-industrialization of America due to rising inequality or the shifting of manufacturing to places with lower costs especially labor costs? Is the shifting of manufacturing to places with lower costs increasing or decreasing inequality? Global inequality has plummeted even as within country inequality has risen (even in places such as China). It's too complicated for generalizations about the social consequences of inequality. On the other hand, generalizations about the economic and political consequences of high levels of inequality are instructive. But that's not the subject of the book or Tabarrok's blog post.

Life expectancy in the US has increased since the publication of the book.

Sorry, not an artfully drafted sentence: I was referring to the decline in life expectancy in the past couple of years, which is since the publication of the book, but life expectancy today is not lower than life expectancy when the book was published. It's always good to have the grammar police around to clarify the intent.

Policing being misleading (whether you meant to or not) is different than policing grammar.

Who is misleading: me for pointing out that life expectancy in the US has been falling the past few years or Tabarrok for failing to mention it? Did I accuse Tabarrok of misleading his readers? That's ridiculous. My main point is to agree with the critics of this book: that generalizations about the social consequences of inequality are misguided.

Looking at all those updated charts, I think there's an elephant in this room.

I think the whole debate is flawed. The way inequality works is not between countries but within countries. What you must therefore plot is what happens to Czechia over time, as it becomes more unequal. Does life expectancy drop? You cannot add Czechia in one graph and omit it in another, over time. So it is error for the 2009 graph not to include Czechia and the 2019 graph to include it.

Bonus trivia: in terms of wealth rather than income, the top 1%, of which I'm a member, control 40% of the USA's wealth. And that's a good thing.

Just to underscore my point -- that you must look at each individual country over time, not compare countries (i.e., inequality is only relevant within a country not between countries, as is well known in economics, wealth is relative not absolute-- you can clearly see from the graph as countries become more unequal over time, they lose their "top spot" relative to the world, meaning their quality of life goes down (decelerates, 2nd derivative).

That's to say: look at Germany, Spain, Greece, and especially Japan, all have become less equal over time so people die earlier than they would have (trend line) had these countries been more equal.

Look at Japan: once, when their Gini coefficient was near 25, they were outliers in terms of longevity, but now, with a Gini coefficient near 33, they are "nothing special", though still #1 in long life (deceleration of their trend line). The thesis is if Japan's Gini coefficient had remained at 25, they'd be living to well over 100 and near the maximum estimated lifespan of 120, given current biological limits, rather than merely just ahead of the pack, as today.

Bonus trivia: the oldest person in the world is a Japanese woman.

Nobody in the past thought if Japan's gini coefficient had remained at 25, they would live to anywhere near 90 to say nothing of 100.

Of course, *now* medical technology is about to increase longevity around the world past 90 by the 2040s - Bangladesh is only five years behind the U.S. and China is just a year or so behind.

@Todd K- correct, my thesis is novel, granted. And your second sentence goes to the well-known phenomena of 'catch-up' growth where the trailing nations quickly adopt the cutting edge technology of the developed countries. The entire curve for age, for example, has shifted upwards (people living longer) because Third Worlders use (and don't pay for) patented medications and medical technologies. But my point still stands, arguably less inequality is better for productivity and a bunch of other metrics. Not that I'm personally for inequality BTW, being in the 1%.

I recall from years ago our host would argue that inequality wasn't rising but falling (a point he made in an interview in the NYT). And inequality was falling globally; indeed, inequality was falling rapidly and significantly. Then along came the anti-globalist Trump, and out went that argument in favor of the argument that, sure, inequality is high and rising but that's a good thing not a bad thing. It takes flexibility to be a successful lawyer and a economist economist. [This comment is in response to Ray's point that inequality should be viewed within countries not between countries. Indeed, global inequality was falling primarily as the result of the China miracle, but even as global inequality was falling, inequality within China was rising and continues to rise - it's higher than inequality in America.]

" inequality within China was rising and continues to rise"

The World Bank says inequality hasn't risen in China for 15 years.

Here's a piece from the somewhat reliable source where our host writes a column:

Notice, however, that the graphs aren't historical and the article says nothing other than there has been a 15 point rise from 1990. I'll stick with the World Bank until I see other historical data.

Life expectancy at birth is defined by the UN/WHO as "Average number of years that a newborn is expected to live if current mortality rates continue to apply."

The big implicit assumption in the plot is that current inequality (Gini index) is the major contributing factor to the current mortality rates. Can inequality/poverty of 20 or 40 years before be killing people today and therefore controlling the current mortality rates?

The USSR collapse was a great natural experiment. 25 years ago, 10% of the Israel population were immigrants from the USSR. The immigrants increased Israel's mortality rates. Cancer and cardiovascular diseases killed people anyway, even if there is a better health system in Israel. I think it's safe to assume current mortality rates are affected by past events. Perhaps the past poverty and inequality count.

A second and more important problem is the calculation of the Gini coefficient. Mr. Snowdown corrected the income for taxes and social transfers. It's not clear if the 2009 plot under scrutiny did the same correction or not. In any case, the 3rd plot correcting for taxes and transfers provides a better assessment of income and therefore inequality.

"The big implicit assumption in the plot is that current inequality (Gini index) is the major contributing factor to the current mortality rates."

There is no need to make that assumption. Plenty of alternative explanations are available. Perhaps the United States is just uniquely bad at Public Health.

or perhaps the US with cheaper abundant food, labor-saving devices, more years in the classroom and sedentary jobs, automobiles, general lack of exercise, contributes to bad lifestyle choices. Hard to get better outcomes from public health expenditures when people refuse to take personal responsibility. Given the rate of obesity in this country, the outcomes are more positive than should be expected.

The last few years the problems have been an increase in drug overdoses and suicide. Perhaps a link in the legalization of drugs and these deaths should be explored.

There are things we could do at the margin on all of those. Lower speed limits. Higher alcohol taxes. More publicly funded fitness facilities, in parks, etc.

And of course discouraging gun ownership.

How will that affect heart disease, mental illness, diabetes, and drug use?

To your point, if “freedom” is now irrelevant to the discussion and a four letter word among liberals, it would be far easier to just pay people to be Mormon.

Yes at the margin a higher tax on alcohol would reduce heart disease.

You said something factual, it’s an MR miracle.

If you’re going to go that route, then just tax food directly. You could probably reduce heart disease a fair amount by eliminating food stamps as a first step.

"Yes at the margin a higher tax on alcohol would reduce heart disease."

False. Alcohol consumption reduces heart disease.

No, it doesn't. Moderate alcohol consumption is correlated with healthy lifestyles because, surprise, people who drink alcohol moderately are healthier and wealthier than people who drink alcohol immoderately, and because non-drinkers often abstain for reasons correlated to ill health, such as preexisting conditions or exceedingly overbearing ideological dogma.

often abstain for reasons correlated to ill health, such as preexisting conditions or exceedingly overbearing ideological dogma.

You evidently fancy Mormons and old-school Methodists are in poor health.

Or South Asian muslims, whose rates of diabetes and heart disease are off the charts. I, too, can cherry-pick data to suit my narrative...

Yes, it does. Cardiovascular health is better even for those who abuse alcohol and have 6 or more drinks PER DAY

I don't think it is too controversial to say that alcohol lowers heart disease. The question is whether this lowered risk offsets the increased due to other factors, like increased cancer rates and accidental deaths due to alcohol. My view is that as long you are not prone to alcoholism then drinking moderate amounts of alcohol is at worst neutral to your health. So if you enjoy it (which I do) imbibe. In terms of public health policy, as a libertarian I would say that you have to have a very high barrier before you take an active approach to reducing people's use by taxing it. You would have to show very significant negative external effects (not just effects on the individual consumer) that couldn't be handled any other way. Alcohol is clearly not there in my view.

And of course discouraging gun ownership.

Gun ownership is quite unremarkable in rural areas and small towns in this country, where violent crime is seldom a problem.

Inanimate objects are not your problem. Hoodlums are your problem. To deter, punish, and incapacitate hoodlums is something progtrash just aren't interested in.

You do understand that private guns kill massively more innocents (through accident and suicide) than "hoodlums?"

No they don't.

Suicide by firearm are 2 and 1/2 rimes greater than firearm homicide in the U.S.

It seems slightly disconcerting to say those that kill themselves are not innocent, but opinions apparently differ.

Don’t be absurd, there are under 500 fatal accidental firearm deaths per year in the United States. Under 500 out of 2,700,000 deaths per year. That’s in comparison to over 17,000 homicides, of which about half are by firearm.

So yes, 17,000 murders compared to < 500 accidental deaths means your statement reeks of innumeracy.

The “suicide gotcha” is nonsense. It’s intentionally misleading.

I don't think anyone here has really opposed what I've said. More have committed to it. "Suicide is ok, because freedom" is an example of that.

Dude how innumerate can a person be, really?

Let me frame it this way:

For every gun (yearly) in the United States, there’s a 0.0056% chance it will be used in a successful suicide attempt. There’s a 0.002% chance it will be used in a murder.

You want to reduce deaths? Eliminate Food stamps, institute a food tax, and allow health insurers (to include Medicare and Medicaid) to discriminate by behavior: exercise, tobacco use, drug use, alcohol consumption, church attendance, sexual partners, marital status, etc.

Talk about restricting freedom...that sounds far more dystopian than greater restrictions on guns

It’s dystopian for insurance companies to charge more for drug addicts, smokers, etc?

Seems fair to me. I’d assume a home insurance company should have the right to charge a family more if they’re self admitted arsonists.

Doing around church attendance seems pretty horrible in terms of balkanizing society on religious grounds. Doing so around tobacco seems like you'd get a lot of dead lung cancer patients, then nothing, as smoking rates are trending down towards nothing under all healthcare systems.

In general it seems pretty crazy to think that people are engaging in any of these behaviors because they know that their insurer will "pick up the tab", and that they'd be incentivized not to if they were uninsurable. Unmarried men are unmarried men not because insurance makes them heedless to the risks to their life span but because low status men have a hard time attracting and keeping a wife. Drug addicts are drug addicts for complex psychological reasons that have no thing to do with thinking that they'll be alright because they're insured. And so on.

Major effect of allowing those forms of discrimination probably you might get a political rush to single payer universal coverage. It'd be much harder to make the argument that insurers exist to spread risk, rather than minimize it for themselves and then collect premiums for the privilege of doing so.

What kind of self-righteous idiot wants to lower speed limits?

Here in California we have 25 MPH in school zones "when children are present." Not everyone slows down, or all the way to 25. I was surprised in a neighboring state(*) to see 15 MPH in school zones, and people carefully driving it.

Surely that must save some kids, at the margin.

* - Arizona or New Mexico?

Great, make traffic worse in California. Everyone will love you for it. Run for office on a platform of lowering speed limits. You'll be unstoppable. An electoral freight train, even.

Not interested. Some accidental deaths are a small price to pay for my freedoms and rights.

My understanding is that it's our mix of population. Our Swedes live longer than those in Swede, and Japanese live longer than in Japan. Also African-Americans live longer than in Africa. Confirming most evidence that we have very good, though expensive, public health. It's just the mix of population that hits the overall numbers.

It kind of can't be like that though. White American life expectancy is 79 years. Still lower than Western Europe, and Britain, Germany, Ireland which are the main sources of population.

The sort of thing Sure is talking about is probably what hits your population life expectancy though, not problems with American healthcare (which probably works fairly well but cannot stem all the diseases of consumption and the like). US life expectancy has become uncoupled from Western European in 1990s, so consumption differences beginning in the early age of great American wealth advantage seem plausible enough.

I only want to point out that Hong Kong is *not* a country; neither the people/government of Hong Kong nor China think of HK as a country. This of course doesn't detract from the point of the post....

Real income levels in Hong Kong, Macao, and Taiwan are > 3x those of mainland China. Macao was a dependency of Portugal for 300 years, Hong Kong a dependency of Britain for 150 years, and Taiwan has been ruled by China's central government for just 4 of the last 120 years. The development of their economic institutions and public life has been altogether different. It's reasonable to treat them separately in analysis.

While the people of mainland China and the Hong Kong government (under the direction of mainland Chinese government) do not think of Hong Kong as a separate country, I suspect that a substantial majority of Hong Kong, in a fair poll, would consider it a separate country. Perhaps more pertinently, Hong Kong is a vastly different than mainland China legally, culturally, socially, etc. The differences between mainland China and Hong Kong are certainly larger than the differences between U.S. and Canada, or between France, Spain and Italy.

Although I'm a U.S. citizen, I've lived in Hong Kong for the past 11 years.

How about running a correlation between healthcare expenditures and life expectancy. We're really good at spending money in our healthcare system relative to others. Or not.

Yes. We spend a million to cure a pair of anti-vax nutters kid of tetanus rather than give him a $150 vaccine.

Couldn't you just change the graph in your picture to "Problems" and "Underclass % of Pop". Then apply that to every single social science issue ever...

You'd have to contrive a definition of 'underclass' that was precise enough for statistical analysis and applicable across countries.

He means blackity-black-black-black.

He always means that. 'Black' however, isn't a satisfactory proxy for lumpenproletarian. The employment-to-population ratio for blacks is about 10% lower than that of the rest of the population. The vast majority of blacks are from wage-earning households and a modest bloc are salaried or are proprietors.

This is the true insidious aspect of significance testing: it allows authors to skip rigorous theorizing (or, more accurately, to substitute in their preferred theory after they achieve significance). How do you blame billionaires for the violence of America's underclass? You certainly can't lead a paper with the argument that gang violence is secretly motivated by resentment of Jeff Bezos. So you do some correlational p-tests, find your magic significance relationship, then, since You Have Found a Real Effect, you get to smuggle whatever half-baked theorizing you want into your conclusions. You can be as vague as you want; there's no requirement that you generate plausible mechanisms or check effect sizes for the causal effect.

In conclusion, these graphs prove the "pernicious effects that inequality has on societies: eroding trust, increasing anxiety and illness, (and) encouraging excessive consumption." It's science.

How can you do a chart like that without Russia, China and Brazil? Not to mention various Dumpster-fire countries that would make the US look a lot healthier.

The vertical axis also makes a big deal out of making it to 83 rather than 79. I suppose if you're 78 it really IS a big deal, but the real-life value of those four years could be subject to both statistical and philosophical debate.

So that is it Americans want. A Final Solution for old people. Such is life in Drumpf's America.

If you are 78 in the U.S., the odds are about the same that you will live to 82 as those in countries where the life expectancy is 82.

Man this is a lot of energy spilled over +/- 2 years of life expectancy. If this is their best metric, the whole endeavor is probably BS.

p.s. inequality is obviously endogenous with nation size. Idk why they thought it was okay to weigh a dozen tiny ethnically homogeneous countries the same as the USA.

Because that's what this is about, right? A variant on the theme that America should be more like idealized European countries?

Some evidence, based on the results of one UK general election, correlating voting habits and life expectancy.

bonjour paulyprintmoremoneykrugman,
we are out of money!
how long does it take to print 3billion euros?
you will see this truck tomorrow
the whale1 wants to talk about that paper
where you explained how good we control health care costs!
also we think you americans need to explain to the whale
very carefully what
you are doing with the hormones & the children!

Too few data points.

Consider also the lack of relationship between mortality inequality ( and income inequality (

The US is structurally higher in both, and European countries structurally lower in both, in general, since the 1960s at least...

But changes in income inequality in Spain, UK and Sweden (where it's a step change), do not impact the mortality inequality. See - and Spain and Sweden remains the lowest in mortality inequality in the panel, despite shifting to substantially higher income inequality, while France remains structurally more mortality unequal than others despite moving to lower income inequality.

Baby Edward Tufte: "Holy truncated Y axis, Batman".

The entire best-worst delta in the original dataset is five years or so.

(Kill some outliers and compare US to Sweden and it's more like three.)

Inequality is not inequity. The US government manufactures poor, ignorant and violent people with its welfare state policies.

Yes, it does.

My heart is not hard. I have come to consider the Great Society the most pernicious technology of the 20th century, to make scientists's misgivings about nuclear energy quaint in their facility.

Even aside from the fragility of the claim, when a chart is meant to show that two variables are correlated, you should be suspicious if there is a line of best fit superimposed over the data. A line of best fit will almost always have a slope, and correlated variables should be visually obvious in a scatter plot.

Japan’s life expectancy ranks high on all three charts above. In 2011, Japan ranked #1 and the U.S. came in at #34 on the life expectancy list by country. Of those 34 countries, Japan had the highest percentage of men smoking…close to half. It had been near 80%, but they were cutting back.

High Japanese longevity perhaps stems from one of America’s more under rated economic practitioners, Douglas MacArthur. In 1948 MacArthur implemented policy requiring the breakup of Japanese commerce and industry controlled by a minority and exploited to their exclusive benefit.

I was told I had mild emphysema. I was shocked, I had only had minor breathing problems at times. However I had smoked for 17 years when I was very young and had quit over 38 years ago, when I developed asthma. I always heard your lungs were cleared 5 years after you quit smoking, but they don't tell you the damage is already done! Mild is not mild, I am on oxygen all the son purchased herbal remedy for emphysema from solution health herbal clinic ,which i used for 15 weeks and am totally Emphysema free ,all thanks to solution health herbal clinic, solution health herbal clinic also cure all type of disease in humans life. Contact and visit their website at , details E-mail: (

Comments for this post are closed