Healthy, wealthy, and wise?

Will Wilkinson linked to this beautiful graph:

Please scroll waaay down for a bigger and more readable image.

This is what Will has to say about the graph:

Gapminder has made available a lovely map/chart illustrating the relationship between GDP per capita and life expectancy. It shows that one thing economic growth is good for is keeping people alive longer.

I’m not going to disagree with that, I think it does suggest a positive relationship.  But I wonder if the graph might be a bit misleading.  Perhaps what it is really showing is that there is a positive relationship between the cultural factors that produce economic growth, and life expectancy.

Why do I make that distinction?  Because it seems to me that part of what we are observing is the relationship between region and life expectancy.  Suppose I were right.  Then you’d expect to see little correlation between life expectancy and income within a region of similar cultural characteristics.  (By the way, I mention “culture” as a catch-all term.  It could be climate, or diet, or genetics, or any other factor correlated with location.)

Note the pattern in Africa.  Most of the countries are poor, but a few are not (look at the big graph because they are often small.)  Do you notice that even the richer African countries, like South Africa, Gabon and Equatorial Guinea also have fairly low life expectancies?

Now look at the Middle East.  Relatively rich Saudi Arabia has a similar life expectancy to poorer neighboring countries.  I don’t want to deny that income has some effect, Kuwait and UAE are higher than most of the others. Bu the effect may be less than you’d think.  Israel seems like a counterexample to my argument; it is much richer and has a much high life expectancy than average.  But it is also culturally different from its neighbors (i.e. European migrants) so that actually fits my cultural explanation.  On the other end is Yemen.  But maybe the regions are not fine-grained enough.  Note that Yemen is closer to Ethiopia than to other Middle Eastern countries like Syria.

Another interesting outlier is Cuba.  Why is the life expectancy so high?  Naive leftists will tell you it’s their “national health care.”  They need to read Robin Hanson.  Wasn’t Cuba one of the richest countries in Latin America in 1958?  If I am right, it has exactly the sort of life expectancy you’d expect if income did not affect life expectancy.  Castro’s nutty economic policies have made Cuban’s poor, but they can enjoy their poverty for just as many years as the more affluent Chileans.

I see no relationship in South Asia, although the sample size is too small.  There are positive relationships in Europe and the Americas, but not necessarily at more fine-grained geographical levels.  For instance, many of the former Soviet republics are in the 65-70 year band, at widely different incomes.  Canadians live longer than their richer North American neighbors.

Now I would never argue that economics doesn’t matter at all.  If you starve your people, then life expectancy will fall (see North Korea/South Korea).  And I concede that China is almost certainly depressed by its relative poverty (when compared to richer countries with similar East Asian cultures.)  I don’t know how much of that is related to the fact that many Chinese went through North Korean-style trauma when they were younger.  Say what you will about Castro, but he avoided the worst excesses of communism.

But I also would guess that a glance at this chart overplays the influence of money, in and of itself.  It may be that much of the relationship is due to some third factor that causes both wealth and health, say education.  That could explain the relatively low longevity levels in some oil-rich countries.



25 Responses to “Healthy, wealthy, and wise?”

  1. Gravatar of Lorenzo from Oz Lorenzo from Oz
    5. June 2010 at 19:41

    Your first link does not work.

    Israel is a little bit more of a complicated case than you imply. Israeli Jews are not even close to being just “European migrants”. Many are of Middle Eastern origin. (Israel accepts Middle Eastern Jews as citizens, while the Arabs prefer to leave Palestinians as stateless sticks to beat Israel with: hence third, fourth, even fifth generation Palestinian “refugees”.)

    Since culture affects things like cuisine and other life-style habits as well as level of investment in female education, there may well be something to what you say. For example, Mormons in the US live longer (on average) than Swedes in Sweden. Despite “suffering” under the oh-so-awful US health care policy mess.

  2. Gravatar of Norman Norman
    5. June 2010 at 19:46

    Or you could simply argue that people who expect to live longer stand to benefit more from investments in physical and human capital, and are thus more likely to invest and have higher income levels. If life expectancy is heavily affected by climate, that could easily account for the pattern.

  3. Gravatar of Contemplationist Contemplationist
    5. June 2010 at 22:19


    I know this is controversial, but since you’re iconoclastic you may be willing to address it: What about the IQ and wealth of nations? IQ seems to have a very strong correlation with national income, and its one of the strongest (or at least, least mushy) social science variables we can measure.

  4. Gravatar of rob rob
    6. June 2010 at 01:15

    Cuba: I think this proves my theory that it isn’t wealth but smoking cigars that increases longevity. 🙂

  5. Gravatar of scott sumner scott sumner
    6. June 2010 at 05:05

    Lorenzo, Thanks, I fixed the link. I agree with your point about Israel. I doubt that cultural differences explain all the difference in their life expectancy, but perhaps a part of it.

    Norman, Maybe, but my hunch is that it is related to education and other cultural differences. That’s not inconsistent with what you say, but my take is that a better educated population is healthier, and also able to produce a higher GDP. Of course income also matters, but not as much as the graph would suggest.

    Contemplationist. I’d be surprised if that finding is controversial. I am pretty sure IQ and income are positively correlated. I think the controversy is over whether IQ differences between ethnic groups are due to genes or environment. Or more specifically, whether genes play any role, as almost everyone agrees environment plays some role.

    rob, Or perhaps makes the longevity more enjoyable. There’s also a theory that every person has the same amount of bliss, but some people live in the edge and squeeze that bliss into a shorter time period. The Mormons (mentioned above) spread it out over a longer period, by avoiding drugs, etc.

  6. Gravatar of DanC DanC
    6. June 2010 at 05:12

    It would appear that you could match the data to toilets per household and longevity.

    Toilets being a proxy for sanitation, clean water, and waste removal

  7. Gravatar of DanC DanC
    6. June 2010 at 05:21

    Sanitation issues

  8. Gravatar of StatsGuy StatsGuy
    6. June 2010 at 05:58

    Nice graph.

    Whenever I see scatterplots, my first inclination is always to look at the residuals. Here, there are some pretty obvious outliers, and in general if you compare the countries above the (implied) regression line to those below, you can see some patterns:


    Vietnam, Cuba, Costa Rica, Nepal, Chile… Call these relatively high equality countries.


    Nigera, South Africa, Russia, Saudi Arabia, and indeed USA. High inequality countries.

    Obviously, I’m cherry picking larger countries with larger populations to make the point, but I think it holds even with smaller countries. And, indeed, within ethnic groups – compare Thailand to Vietnam. Do we think the gene pool there is really that different? Compare the US to Spain…

    I wonder if I can get ahold of the raw data, and color the cirles by gini coefficient or some other income dispersion function instead of region.

  9. Gravatar of Chris Chris
    6. June 2010 at 06:13

    Life expectancy is a lagging measure since we do not know how long current people are going to live and it also depends highly on health in one’s younger years.

    It would be interesting to see current life expectancy charted against GDP 30, 40 or even 50 years ago as many of these countries made some dramatic changes over the last half century while most of Europe, the US and Canada has been relatively affluent for more than a century.

  10. Gravatar of StatsGuy StatsGuy
    6. June 2010 at 06:26

    Yeah, they have the data – world bank through 2007, really spotty. Also, when the world bank data is absent in some years, they don’t impute or interpolate from adjacent years. Blech… Even then, fewer than half of the countries report data between the 2000-2007 time frame.

    Cool time series animations… Traveling today, will play around with it later tonight. Click on, say, belgium and US, or norway and US, and observe the paths the countries take from 1945 to 2008.

  11. Gravatar of Indy Indy
    6. June 2010 at 06:30

    I’m with StatsGuy – GDP per capita doesn’t quite capture “average quality of life / health / medical care”, when there is large inequality – especially in those countries that depend on natural resource extraction for a large fraction of their GDP – the spoils of which don’t improve or effect the lives of most of the population. I think “median or modal household income” would be a much better measure.

    I also prefer “life expectancy at age 60” to get away from other factors like deaths from war or idiosyncratic prevalence of risky occupations (like in a country that depends heavily on a dangerous kind of fishing).

    But see what else the chart tells us – it looks like there are enormously diminishing returns to wealth – but this is in a chart that is already log-wealth vs non-log-life expectancy.

    Early on, every doubling of wealth seems to get around 5 extra years of life-expectancy – but after about $10,000 you have to nearly quintuple wealth to squeeze another 5 years out of it.

    And if you hold “Sumner-Culture” more or less constant it’s even more striking how little effect wealth has.

    Compare Nepal, Bhutan, Bangladesh, Pakistan, and India – wealth makes practically zero difference.

    The non-recently-war-ravaged (Yemen, Iraq) Middle-Eastern countries have a 10-fold span of wealth with practically no life-expectancy effect.

    Query: Based on this, what additional resource allocation (or share of growth) should Bangladesh allocate towards longevity-extension / end-of-life medical care? And the US?

  12. Gravatar of Wealth, Health, and Trust «  Modeled Behavior Wealth, Health, and Trust «  Modeled Behavior
    6. June 2010 at 08:09

    […] Scott Sumner has some comments relating to a vague allusion to “culture”…but I want to be more specific: […]

  13. Gravatar of Doc Merlin Doc Merlin
    6. June 2010 at 08:35

    Another data point in support of your theory is that hispanics in the US have a very high life expectancy despite being on average poorer than average.

  14. Gravatar of scott sumner scott sumner
    6. June 2010 at 12:37

    DanC, Yes, I agree.

    Statsguy, I agree that inequality explains part of the picture. But I also think there are many other cultural factors. Vietnam’s culture is very different from Thailand’s culture (or at least I have read that, I am no expert.) I read that Vietnam’s culture was more similar to China and Korea, than Thailand. Does anyone know anything about that?

    I’ve been hounded by someone named “Chris” over at my post about America’s amazing success. He insist Chile is a example of the horrors of neoliberalism. Glad to know it is relatively equal. 🙂

    Chris, Very good point. I predict that the average Chinese baby born today will live longer than the average Danish baby born today. (In HK life expectancy is already about 4 or 5 years higher than Denmark.)

    Indy, You said;

    “I’m with StatsGuy – GDP per capita doesn’t quite capture “average quality of life / health / medical care”, when there is large inequality – especially in those countries that depend on natural resource extraction for a large fraction of their GDP – the spoils of which don’t improve or effect the lives of most of the population. I think “median or modal household income” would be a much better measure.”

    I agree, indeed the point of my post is that GDP per capita may have much less influence on health then this graph suggests. But I’m not even sure median income would solve the problem. Median income in Cuba is far below Chile, for instance. But I do agree it would improve the fit slightly.

    You said;

    “Early on, every doubling of wealth seems to get around 5 extra years of life-expectancy – but after about $10,000 you have to nearly quintuple wealth to squeeze another 5 years out of it.”

    I just don’t see this at all. I see virtually no relationship between income and life expectancy within North America, or within Western Europe, or within the former Soviet block, or within Africa, or within South Asia. I just don’t see it.

    You said;

    “And if you hold “Sumner-Culture” more or less constant it’s even more striking how little effect wealth has.
    Compare Nepal, Bhutan, Bangladesh, Pakistan, and India – wealth makes practically zero difference.
    The non-recently-war-ravaged (Yemen, Iraq) Middle-Eastern countries have a 10-fold span of wealth with practically no life-expectancy effect.”

    Yes, I strongly agree.

    Bangladesh needs growth. It is unlikely they can boost life expectancy much at that income level. Education might help more than health care at this point.

    Doc Merlin, I discussed that in another post, but it appears that information is false. No one has been able to confirm it. I am very doubtful that American Hispanics live a long as the Japanese (which is what the data shows.) I believe the study that showed that is faulty.

  15. Gravatar of Jason Jason
    6. June 2010 at 16:32

    There is a very strong correlation between this graph and one that has caloric intake as the abscissa. (It is somewhat more compressed at the high end as per capita GDP has no physical limit, while caloric intake does.)

    This would definitely fall under culture — the native cuisine (plus food availability). And it “explains” some of the oddities mentioned above. For reference, I am looking at this data (all in kcals/day/capita):

    South Africa (2998 kcals) and Gabon (2754) may be rich, but their diets are more in line with Nigeria (2740) and Ghana (2907). It is possible high inequality in South Africa, and the mixed European and African cultures mean average caloric intake is distorted by high intake Europeans and lower intake Africans.

    Saudi Arabia and Kuwait are near each other (3143 and 3064, respectively), while Israel has a more European level of 3527.

    And indeed Yemen (2067) is closer to Ethiopia (1979) than to Syria (3034).

    Comparing Cuba (3273) with the rest of the Caribbean: Haiti (1869), Dom. Rep. (2295) and Jamaica (2851) — it has more calories. And more than Chile (2920) as well.

    And in a comment, Scott says that it is striking how little effect wealth has among Nepal (2359), Bhutan* (2124), Bangladesh (2281), Pakistan (2292) and India (2351).

    One thing that is interesting about using this caloric intake data is that while high inequality can distort GDP per capita by very large amounts (heavy tail, Pareto distributed income), people can only eat so much (physical, more Gaussian-like distribution), so things like India’s wealth only skew it slightly above Bangladesh in intake.

    (Talking about in a single culture here, given that there is the possibility of distortions as mentioned in the South African case, above.)

    There are some studies that show caloric restriction can lead to longer life in animals, possibly explaining some part of why Japan (2812) beats the US (3748), but I don’t think we’ve reached the level of world development where we could see this signal strongly (if it exists at all). It probably also matters where calories come from.

    I also imagine there is a certain amount of caloric variation by latitude: Russia (3375) may eat nearly as much as we do, but being a much colder country on average could mean a lot of calories go into maintaining body temperature, and so Russia falls lower on the life expectancy relative to its caloric intake.

    * Bhutan’s caloric intake was not readily available and took some digging on the internet. This may not be correct.

  16. Gravatar of Lorenzo from Oz Lorenzo from Oz
    6. June 2010 at 20:57

    Statsguy: the positive outliers are also fairly monocultural societies, hence communication levels are higher and patterns of different behaviour less. The negative outliers are generally more culturally diverse (even Saudi Arabia, given its enormous guest worker population), hence communication levels are lower (Saudi Arabia may also have a female education issue: both levels and effectiveness) and patterns of different behaviour more pronounced. Both of which have effects in income diversity too.

  17. Gravatar of John Dewey John Dewey
    6. June 2010 at 21:36

    I’m surprised anyone would doubt that higher GDP per capita raises life expectancy. As incomes rise, consumers must spend smaller shares for basic essentials – food, clothing, and basic shelter. That leaves more disposable income available for health care and personal safety.

    Another way to look at it: higher GDP per capita implies higher productivity, so that fewer labor resources must be utilized for food, clothing, and basic shelter. That frees up more labor resources for health care, for education, for safety and security, for sanitation, and for disaster relief.

  18. Gravatar of scott sumner scott sumner
    7. June 2010 at 06:51

    Jason, Thanks for that interesting data. I wonder if there is a peak, and then some point where more calories reduce longevity.

    Lorenzo, That’s a good point.

    John, I mostly agree. My point was that the effects are smaller than what one might infer from the graph. But I do believe there is probably some effect, at least at lower incomes. (I am less sure about rich countries, as more calories and more luxury may cause obesity.

  19. Gravatar of Master of None Master of None
    7. June 2010 at 10:15

    Here is the graph with literacy rate (i.e. “education”) on the X axis:

    Education shows much less correlation with health, or with GDP/capita.

    However, there appears to be a step-change when societies reach ~90% literacy. Above ~90%, the range of life expectancy is 65-81 years; below 90% literacy, the range is 45-76 years (and much greater dispersion).

    There are still strong regional biases with “education” as the explanatory variable.

  20. Gravatar of John Dewey John Dewey
    7. June 2010 at 10:54

    The question being discussed is whether GDP per capita strongly influences life expectancy, right? No one is arguing, as far as I can tell, that GDP per capita is the only factor influencing life expectancy. Previous analyses I’ve seen show that genetics, culture, and political environment (examples: war, genocide) also determine life expectancy.

    Is there any doubt that global life expectancy will continue to increase if global incomes increase? Does anyone doubt that increasingly greater resources will be employed in mitigating or eliminating factors which limit life extension?

  21. Gravatar of John Dewey John Dewey
    7. June 2010 at 11:05

    Scott: “I am less sure about rich countries, as more calories and more luxury may cause obesity.”

    Would you consider wording that a different way? Here’s my view:

    Higher GDP per capita enables more choices for the individuals in those nations. Some will choose more calories and less physical activity, resulting in obesity and earlier death. Others will choose more medical treatment and healthful activities, resulting in longer lives. Certainly culture can play a large role in shaping the choices of a nation’s population.

    I do not believe individuals or even national populations are predestined to make poor choices which have been enabled by increased wealth. So I’m reluctant to agree that increased wealth would ever cause obesity.

  22. Gravatar of John Dewey John Dewey
    7. June 2010 at 12:12

    Just to be clear about my last comment, I also believe that humans will, by nature, seek longer lives – will seek to survive. So I believe that the enabling power of higher GDP per capita is more likely to be used to increase longevity rather than reduce it. As evidence, we can look at the longitudinal analysis of per capita GDP and longevity – either for the U.S. or for the entire global population.

  23. Gravatar of scott sumner scott sumner
    8. June 2010 at 14:05

    Master of None, Thanks. Your observations sound plausible. They really should have used a different scale–more of a log scale or something.

    John, Those are two questions. Does income directly influence life expectancy, or do the underlying factors that affect income drive life expectancy. Thus what if you dropped money out of a helicopter onto a poor country each year, to bring average income up to a higher level. How much would life expectancy rise if all other factors stayed constant? That’s a little bit like the Saudi Arabia example.

    That’s a different question from whether we can expect life expectancies to rise as the world develops. The factors that are causing development (like education) might be causing the higher life expectancy, not income. Again, I think income does have a modest effect, but perhaps smaller than the graph would suggest.

    John#2, Causality is a difficult issue. If wealth causes people to get soft, even though they could choose to stay slim, is wealth causing obesity? I think it could be argued either way. But in the end I agree, I expect life expectancies to keep rising.

  24. Gravatar of John Dewey John Dewey
    8. June 2010 at 22:53

    Thanks for the reply, Scott.

    As I see it, the primary driver of GDP per capita has been for centuries the increases in productivity. So much so that I tend to ignore the exceptions such as Saudi Arabia and Norway and a few others which possess extraordinary natural resource assets. Because I ignore such outliers, I have assumed that increases in GDP per capita always arise from productivity increases. I agree that the outliers are different.

    If we do not agree that GDP per capita is a measure of a nation’s productivity, then I can see how we would come to different conclusions about whether the former drives life expectancy.

    Also, I agree that such factors as political environment, education, and culture can directly influence both productivity and life expectancy. I’m not sure how to determine which of these has a stronger influence on life expectancy:

    polit/educ/cult/other => +life expectancy

    polit/educ/cult/other => +GDP/capita => +life expectancy

    The increases in life expectancy in the U.S. over the past century lead me to believe it is the latter. But I’ll try to be open-minded about the issue.

  25. Gravatar of ssumner ssumner
    9. June 2010 at 04:23

    John, I think that sums up the issues very well. My reading of the graph was that it looked to me liek there was some of to first chain of causation going on, although I would never deny that there was also some of the second chain you illustrate.

Leave a Reply