måndag 15 februari 2010

Wilkinson ger Norberg, Nordling med flera svar på tal

Richard Wilkinson och Kate Picketts bok The Spirit level (på svenska Jämlikhetsanden) blir, inte helt oväntat, hårt kritiserad på allsköns (ny)liberala bloggar. Tex. Johan Norberg kritiserar bokens vetenskapliga metoder och slutsatser här och Danne Nordling påstår sig ha funnit ett underligt ''fel'' i boken. Jag vidarebefordrade den del av kritiken jag tyckte var berättigad till Wilkinson själv via mail. Här publicerar jag, med Richards tillåtelse, hans svar:

Dear August, there are lots of different ways of measuring income distribution and lots of difference income concepts over which you can measure inequality. As we said in the book, we averaged three years of figures published by the UN which were basically the same as the ones published by the World Bank. They came originally from the Luxembourg Income Study (LIS) which was set up to help produce internationally comparable figures. In our use of data we have an absolute rule that we take the data as published in our source. If we started to pick and choose data points that would be an important potential source of bias. As an example, look at figure 6.4 showing infant mortality. We included Singapore even though we cannot believe it has the lowest infant mortality in the world and in strongly detracts from the relationship with inequality.

I have since become aware that OECD quote very different figures for Japan which come from a different Japanese income survey from the other one, but I have not found out what the difference is in what the two surveys measure. However, my impression is that the OECD figures have not been through the same careful processing to maximise international comparability as those on LIS. But this is not a major problem for our thesis. If you take Japan out of our index graph (the one where we put al the health and social problems together), or put Japan almost anywhere else, there is still a strong relation between inequality and our index of health and social problems.

Australian research on trends in income inequality: As I have often said, there is a background rate of improvement in life expectancy - we gain two or three years with every decade that passes and in most developed countries that has been going on for over 100 years and there are no adequate explanations of it - though I am sure it used to be driven substantially be the long term (not short term) effects of economic growth. The forces behind this are not only unknown, but have changed. Since about 1975 death rates have been falling at older ages where little progress had previously been made. Time trends will be powerfully effected by the background rate of improvement, but the fact that there are now about 200 studies looking at health in relation to inequality shows that inequality must be a factor yet it cannot be anything like the most important factor. The life expectancy and inequality relationships in our book are one of the weaker ones we discuss, and what inequality seems to do is to make small differences in the background rate of improvement in life expectancy enjoyed by almost all rich countries. To correlate century long changes over time would mean controlling for the unknown background rate of change. And what the residual would then be correlated with are basically two periods of change. Inequality was high in most of the English speaking countries until about 1930, it then fell almost continuously till sometime in the 1970s and started rising again. In doing time series analysis, no one knows what lag periods to use: they should probably be different for every cause of death and every age group: we do know for example that health in later life is strongly influenced by early childhood experience. The fact that most older people dying now were probably born in a period of increasing equality yet die in a period of increasing inequality points to some of the difficulties in doing time series.

I don't have any doubt that Angus Deaton is wrong about the issue being poverty. Many studies have controlled not only for poverty, but also the effect of individual incomes across society. And absolute poverty could never explain why the effects of inequality go all the way across society. Nor could they exlpain the total lack of relation between most of our outcomes and Gross National Income per head among the richaest countries - see for instance Figures 1.1 and 1.3.

We will have to spend too much time in the future defending our thesis from fairly simple criticism even though we only shows what many people have believed for hundreds of years is true.

I am attaching a paper - a commentary - on the difficulty of understanding what is driving the long term improvement in health. You will understand that I cannot spare enough time to answer too many individual questions.

With best wishes, Richard

Och angående Tino
Sanandajis påstående att inequality is positively correlated with life expectancy då man använder sig av OECDs Gini-statistik, skriver Wilkinson följande:

Dear August, - just to add to what I said earlier, this morning we looked at our Index of Health and Social Problems in relation to the OECD Ginis you mentioned. The correlation is only a little weaker - close to 0.7 - and almost entirely because Japan is an outlier on these figures. You might like to look at the Powerpoint slide of that relationship which I have attached. We will add something on this to our FAQs. Add it to your blog etc if you want, but the most important thing is that the OECD figs do not make much difference to the correlation - except for Japan.
best, Richard

Wilkinsons nya Gini-index (Index of health and social problems in relation to inequality):

Wilkinson svarar kort på Tinos specifika kommentarer kring life expectancy och inequality (se tex. kommentarsfältet till detta blogginlägg):

Sorry - far to busy to follow up this stuff. I know the right will be doing everything to get rid of our material but we don't have time for blogs etc now. I notice country names are not given and assume he has included countries which are not among the richest in the world and so should control for GNP before looking at inequality. Should also try looking at mortality among infants and working age populations.
best, Richard


Wilkinson ger ett mer utförligt svar på Tinos gini-index:

Dear August,

These problems look serious only to the uninitiated, unaware of the vast literature. There are now about 200 peer reviewed analyses of the relation between various measures of health and inequality in different settings (see attached review). There have been times in the past where, because of rapid changes in income distribution, cross-sectional relationships have temporarily disappeared only to reappear when the new levels of inequality have had time to work their way through to affect culture and then health. As I said, death rates at older ages which now dominate life expectancy, are likely to be influenced by inequality throughout life. Although I started off working on health in relation to inequality, we now know that health is more weakly related to inequality than many of the other outcomes we look at.

Kate has quickly run through our data to see how the outcomes relate to the OECD inequality figures (see attached file). There are a few changes, some stronger, some weaker, but the vast majority much as before - see attached table. Inevitably things change, there are lag periods and there are different measures of income and income distribution. She has also included correlations for some of the relationships in question using the new inequality measures in the UN Human Dev Rpt 2009.

Richard Wilkinson

Nya siffror:


Wilkinsons svarar på Tinos senaste inlägg där det hävdas att Wilkinson nu ''erkänt'' att ''there is no statistically significant relationship between life expectancy and inequality'':

You are clutching at straws. You forget the 200 papers on inequality and health covering rich and poor countries, regions of Russia and China, S. America, the analyses of the 50 US states, plus the reasons why you might sometimes loose a simply cross-sectional relationship (to the naive delight of critics) and our knowledge of life-long causal mechanisms.

You might also like to read a recent meta-analysis of multilevel models of the relation between income distribution and health which covered individual data on 60 million people. With it I attach our editorial comment - Both from the British Medical Journal.

I cannot spare the time to continue with this.

Här har jag, för den intresserade, laddat upp alla filer (index etc.) som Wilkinson bifogade i sina mail.

28 kommentarer:

Tino sa...

Thank you.

I have responed in a updated post


The relationship between life expectancy and gini is not even close to statistically significant, regardless if the OECD or the UN data is used. He needs to account for this.

see also:


August Torngren Wartin sa...

I think the link you posted now is the wrong one(?), but I have read your new blog post with the after-tax ginis, is that the post your'e talking about?

Wilkinsons Gini-index showed in my post is, as you can see, an index of health AND social problems in relation to inequality. It may be true that the relationship is different (not statistically significant) if you're measuring only life-expectancy, and it would be very interesting to hear Wilkinsons comment on that. Maybe I'll send him another mail when I have time.


Björn Brändewall sa...

I'm no expert, but the long paragraph starting with "Australian…" sounds pretty much like Wilkinson is confirming that his conclusions are merely guesswork. He's basically saying "Sure, compared to economic growth it's hard to find any support for my thesis about life expectancy, but that is just because no one knows how these things should be measured. But a lot of people are LOOKING for a relation between health and inequality, and that must mean that such a relation exists!"

Doesn't make sense to me.

Tino sa...

Thank you again August.

* I do include country names. My first regression is EXACTLY the same 21 countries Wilkinson uses, the only difference is that I use Life Expectancy, instead of the "index" they have built. The relationship is not statistically significant.

* I have run a simple regression of life expectancy on Gini and Per capita income (all data from OECD for mid 2000s and 2005) for the 28 OECD countries:

p value for Gini: 0.783
p value for per capita GDP: 0.55

This means neither value is even close to being statistically significant (why is GDP not statistically significant? One reasons is that the OECD countries already come from a selected sample of high income, per capita income is a restricted variable).
Now Wilkinson's: 21 countries

p value for Gini: 0.451
p value for per capita GDP: 0.986

No statistically significant. Not even close.

* I have already run infant mortality, and have written that infant mortality (and to a lesser extent obesity) are the only health variables out of (now) 13 investigated that are linked to inequality in a statistically significant way.

* I traced down adult mortality for people aged 15 to 60 from the WHO. It not even close to being related in a statistically significant way to inequality.

For the full 28 sample the p value is 0.63, for his selected sample of 21 countries the p value is 0.67.

Please don't waste my time more with wild goose chases, and answer the direct questions: Why is life expectancy not related in a statistically significant way to inequality, if inequality is a major killer?

By the way, I notice he had time for a 750 word response as recent as yesterday. How convenient that he doesn't have time now...

August sa...

I've now sent this further to Wilkinson. It will be interesting to hear his comment.

The reason why he gave me a long answer on sunday was probably because it was during a weekend.

''Please don't waste my time more with wild goose chases, and answer the direct questions: Why is life expectancy not related in a statistically significant way to inequality, if inequality is a major killer?''

Good question. As I wrote before: Wilkinsons Gini-index showed in my post is, as you can see, an index of health AND social problems in relation to inequality. It may be totally true that the relationship is different (not statistically significant) if you're measuring only life-expectancy, and it would be very interesting to hear Wilkinsons comment on that.

August sa...

Anyway, saying that inequality kills can still be ''correct'' if he measures the relation between all social- and health problems and inequality, and not only life expectancy, right? As you already know Wilkinson wrote the following in his answer: ''The life expectancy and inequality relationships in our book are one of the weaker ones we discuss, and what inequality seems to do is to make small differences in the background rate of improvement in life expectancy enjoyed by almost all rich countries.''

Tino sa...


Wait a second. Social problems are one thing, health is another. If inequality is correlated with social problems (such as say illiteracy), would you call that killing?

If inequality really kills, and if the effects of inequality on health are very strong, it should effect life expectancy.

It is possible that inequality kills, but that the effect is very small, and we cannot detect it. But in that case you don't go around claiming you have evidence. If fact one of the slides in the presentation is a claim that inequality reduces life expectancy. The slogan of the presentation is "Inequality Kills".

This if my first objection.

My second objection is that Wilkinston is very dishonest in this debate. He pretends that he responded to my objection about Gini with his index. I have no idea what he put in the index. Furthermore, I did not question the correlation between inequality and social problems, I questions the claim that inequality in income kills you. Why use some mixed index, arbitrarily weighted, instead of the straightforward measure of health?

He is trying to trick the readers.
I believe strongly that inequality and social problems are connected, although I think mostly social problems cause inequality, not vice versa. Lack of integration among immigrants in Swedish and low skilled immigrants in the U.S is a social problem that causes inequality (because they do worse as earners and in other part of life). But everyone already knows this.

The Spirit Level is making a new and extraordinary claim, that they have *proven* a causal link between inequality and mortality.
Extraordinary claims require extraordinary evidence. But forger about extraordinary evidence, such as a direct casual link between inequality and shorter life that accounts for endogeneity, they don't even have a statistically significant correlation between their main variables!!!

Did they then report this problem out in their presentations, as objective scientists? No. Instead I, some random graduate student, had to waste my time reproducing their claims and finding it doesn't hold.

I believe that inequality is a very bad signal, equal societies are more healthy than unequal one, but inequality in income is the symptom, a manifestation of social ills, not the deep problem. If you give $10.000 to each homeless person in Stockholm very little will be achieved, you have to somehow get to the root, such as substance abuse. They believe inequality is what causes problems, crudely if you give the homeless enough money they will no longer drink, their mental illness will be cured etc.

But what we believe is immaterial. My subjective opinion is not more important than anyone else's. Neither is Wilkinson's. What matter is *scientific evidence*.

Instead we are getting their opinion, packaged as empirical science.

Tino sa...


Let me clarify my personal beliefs for the record: While the main problem of inequality is as a bad signal of deep problems, it is a problem of it's own. As long as you have access to health care it doesn't kill you that others are earning more than you, but it creates discomfort, weakens social ties, and makes democracy work less efficiently.

That is why I think we should have redistributive taxes, and be very careful with inequality creating social experiments, such as mass immigration. Homogeneous countries such as Japan have fewer social problems than diverse countries like the U.S (and increasingly Sweden).

But again: what I believe about inequality is not the issue, the issue is what evidence we have to make claims. I do not go around claiming that I have scientifically proven that Swedish immigration policies kill people, just because it raises inequality.

August TW sa...

Ok! I generally agree with what you write, don't have much to object.

I've sent your objections to Wilkinson and hope he will write some answer.

I haven't had time to read all your updated blog posts yet, but I just wanted to ask you if your AFTER-TAX gini index still showed that ''inequality is positively correlated with life expectancy'' or was that only when you used before-tax ginis?


August sa...

One other thing: I also wonder if you think gini is a better measure of inequality than 20/20?

August sa...

I've now updated with new gini-correlations etc.

Tino sa...

I encourage everyone to go ahead and klick on the table. You will see with your own eyes that inequality is not related in a statistically significant way with life expectancy. P value 0.24 means that it is not even statistically significant at a 10% level (usually you require 5%).

Wilkinston is through this table ADMITING that his main variable, inequality, does not have a robust tie to the main measure of health, life expectancy.

Mission accomplished.

He refuses to admit this in words however. Instead we get to know the number of studies does about the issue. What he does not tell you is that many of the of those studies find no effect (I link to a few as examples). The number of studies tell us nothing. There are hundreds of studies made on taxes and growth, but there is no accepted robust relationship.

Would you believe me if I told you: Taxes are proven to be bad for growth. I have no robust evidence for this, but here are references to 200 studies, so it must be true!

Wilkinston want to remove Japan from the sample. Even if he does this it is not statistically significant at a 5% level. And what justification does he have to arbitrarily remove one country from the sample, a country with 127 million inhabitants? Can I remove the U.S from the sample when doing obesity? There is a name for this: data mining. There is another name, but let us not get nasty.

I can tell you that this is not how you do science. If you are an undergraduate student, and remove one country with no justification from your small sample in order to get closer to statistical significant, your professor will likely fail you.

Does Wilkinston admit in his lectures and in his book that life expectancy is not linked to inequality?

If he doesn't, and gives the audience the opposite impression ("Inequality Kills") he is committing academic fraud.

August sa...

´´Wilkinston want to remove Japan from the sample. Even if he does this it is not statistically significant at a 5% level. And what justification does he have to arbitrarily remove one country from the sample, a country with 127 million inhabitants? Can I remove the U.S from the sample when doing obesity?´´

Because the inequality numbers for Japan are so different between OECD and UN-numbers etc. That´s why he also included some correlations where Japan is not included.

And you can see that Wilkinson also wrote:

´´Although I started off working on health in relation to inequality, we now know that health is more weakly related to inequality than many of the other outcomes we look at.´´

August sa...

But I agree that he should admit more openly that the relationship betnween inequality and life expectancy is not statistically significant.

Tino sa...

August: The relationship is not statistically significant even if you use his 21 selected countries and use the UN-HDI data (which, by the way, get their Gini figures from the World Bank).

The problem is not the data. The problem is the dishonest presentation of the data.

Anyway great journalism August.

August TW sa...

''The problem is the dishonest presentation of the data.''

I have to agree regarding life expectancy.

But the other relationships (between for example different social problems and inequality) seems to remain pretty strong regardless of which measure of inequality you use.

Tino sa...

I see that Wilkinson is in full retreat.

Let me just note his dishonesty. While he keeps mentioning 200 papers, he does not tell you that far from all 200 papers actually found any evidence. According to Wilkinson's own (I assume very biased) " reading 87 studies showed wholly supportive analysis".

The meta study in the British medical journal which was not written by him and therefore can be taken more seriously, only has 28 studies, not 200, a large part of which do not find statistically significant results.

More importantly, even in the meta-analysis they only find a "modest adverse effect", find publication bias for the cross-sectional studies, point out that they are "prone to biases in the original studies."

They are unlike Wilkinson aware of the causality problem I discuss to in my blog, and point out that the correlation between of inequality and health is reduced or goes away when you control for background variables.

They explicitly caution the readers "The findings need to be interpreted with caution given the heterogeneity between studies, as well as the attenuation of the risk estimates in analysis that attempted to control for the unmeasured characteristics of areas with high levels of income inequality".

Let me cite the conclusions in full:

"Conclusions: The results suggest a modest adverse effect of income inequality on health, although the population impact might be larger if the association is truly causal. The results also support the threshold effect hypothesis, which posits the existence of a threshold of income inequality beyond which adverse impacts on health begin to emerge. The findings need to be interpreted with caution given the heterogeneity between studies, as well as the attenuation of the risk estimates in analyses that attempted to control for the unmeasured characteristics of areas with high levels of income inequality."

Please also look at page 8 of the British medical journal study, and look for the Gini for Japan. It is 0.314, as I point out! Was Wilkinson unaware of this?

This man is abusing the trust people have in academia.

Nor does he give us any valid reasons why we would "loose a simply cross-sectional relationship".

And even if he is now retreating and blaming in on some mysterious but unmentioned valid reason why the correlation does not hold, he has no explanation why he claimed that cross country life expectancy was related to inequality in his presentation:


go to page 17.

Lastly, if there is a "life-long causal mechanisms", that should be reflected in life expectancy, genius.

I find it laughable that someone who has built his career on the correlation between health and inequality without once establishing an exogenous effect calls me naive.

I see he again suddenly understood that he has no time. Good excuse. Oddly, people who are winning a debate rarely run out of time.

Tino sa...


I am sure if you actually did a thorough analyses, some social problems are caused by inequality, such as crime.

But it is impossible to trust the Spirit Level claims, if they are cheating so blatantly about inequality and health. These people are not acting as scientists, they are salesmen pushing a nice story to sell their book.

Remember a couple of their other relationships also fell apart when I tried to replicate them, such as mental illness and innovativeness.

We need careful studies done by objective people, who also control for causation. Preferably the study should use some form of natural experiment.

You still have the problem that the causality goes the opposite direction: Social problems CAUSE inequality, or are both caused by deeper social problems (such as lack of integration).

In Sweden children in the Rosengård have higher dropout rate than Children in Småland. Malmö also has more inequality in income, because the immigrants earn less.

But it is not the lower incomes of Rosengård that is CAUSING the problem, both are caused by a deeper problem of lack of integration.

You have this problem for a lot of social variables they claim are caused by inequality, such as trust. Of course trust is lower in Malmö than homogenous Tyresö. But it's not caused by income. If the government raised taxes on the rich in Malmö trust would not go up one bit, as their native theory suggests.

We just had an economic crisis that hit the stock market wealth of the rich a lot. Do people in Rosengård trust the upper middle class Swedish people in Malmö more? Of course not.

Inequality is mainly a negative symptom of deeper social illness, not the main cause.

I believe strongly in targeting social problem, and funding this by taxing those who can afford it, namely the rich. But I don't think taxing the rich and giving the money to the middle class will solve the social problems of the underclass, which The Spirit Level childishly claims.

Martin Villner sa...

Tack för att du följer upp debatten kring jämlikhetsanden, kul att följa den.

Några kommentarer på tidigare inlägg:

Detta med att ”ojämlikhet dödar”; även om förväntad medelivslängd inte har så stark korrelation till ojämlikhet (som de också skriver i boken) så finns det ju andra sociala/hälsoproblem som de tar upp i boken och som korrelerar med ojämlikhet, tex psykisk sjukdom, droganvändning, fetma, mord, våld. Dessa påverkar också livslängd och hälsa. Så ”ojämlikhet dödar” är ju inte helt missvisande ändå även om det är tillspetsat.

Fattigdom eller ojämlikhet?; I boken disskuterar han om det är fattigdom och/eller ojämlikhet som förklarar skillnaderna mellan olika länder/delstater för de olika variablerna. Som de själva skriver om detta:

”Nor can an increase in health and social problems be the cause of widening income differences. As we have seen, a wide range of health and social problems tend to move together – countries which do badly (or well) on one outcome tend to do badly (or well) on others. If they were not all results of inequality but were instead separate causes of inequality, that would not explain why they move together. Indeed it is not plausible to think that problems such as homicide, obesity and low standards of child wellbeing – which are all associated with inequality – could be a cause of it.”

Det verkar orimligt att tex mord, barns välbefinnande eller antal interner skulle orsaka ökade inkomstskillnader. Så inkomstskillnader verkar därför vara en trolig förklaring till skillnaderna mellan länderna/delstaterna.

/Martin Villner

Anonym sa...

Fantatiskt bra initiativ. Se till att skicka den till alla ovan nämnda kritiker, som Johan Norberg!

atw sa...

redan gjort. /atw

Dr. Jordan B Peterson sa...

Why are you obsessed with the life expectancy measure? Wilkinson's index is preferably, statistically, because uses multiple measures -- and he demonstrates a very powerful relationship. He also duplicates many of his analyses within the United States, using state data, which apparently invalidates your claim of inappropriate data selection. As far as I can tell, you are trying to invalidate his broad demonstration by showing that a single regression does not manifest itself significantly if additional countries are included. Even if that is true -- and it could well be -- you are reducing a complex argument to a single point. Why?

Term Papers sa...

Many institutions limit access to their online information. Making this information available will be an asset to all.

Jeff sa...


It's because he's doing what every staunch status quo defender does when their weak position is attacked. They find something small to hook on to, they then distort that hook by adding more stuff in (in this case countries, despite Wilkinson fully explaining why he used the countries he did) they then use different criteria altogether and go "Aha look, their whole thesis is wrong."

It's weak, pathetic and intellectually dishonest. What's more is this is from some spotty economics student and the spirit level is written by a professor emiritus with 30 years experience in research (i.e, more experience in one field than the clown who wrote this blog has of living).

This is standard right wing tact, and it's getting more desperate now because the evidence that unfettered capitalism just isn't working now (did it ever?) is mounting up and they're worried.

It always comes down to the same thing as well, it's the "I'm alright, Jack" argument. In other words, i've got mine and screw everyone else. Selfish, weak and fact free idiots, every single one of you.

Jens N. Roved sa...


I agree with you. The funny thing about it is that they have so much time to sit and fight their little personal flamewars on the internet. With so much time to spend, one should thing that they had better do some quality research of their own and get it published in social sciences peer reviewed journals, which is the opposite claim of what they make towards Wilkinson and Pickett.

They are blatantly ignorant of their own stupidity in putting up arguments where they makes themselves guilty of their own projections towards highly esteemed professors and scientists.

Jens N. Roved sa...

Correction of typos: Thing = Think

Makes = make.

Getting late here.

RogerGLewis sa...


Saliet cut and paste here no commentary. The whole paper deserves a proper reading draw your own conclusions.


Best Wishes. Personally I saw the Wiljkinson lecture on the Spirit Level on Kunstkapkanalen and found the arguments very compelling Economics is a belief system its clinging to GDP as a sensible measure of broad economic performance is very strong evidence of that.

Florian Burmeister sa...


"Här har jag, för den intresserade, laddat upp alla filer (index etc.) som Wilkinson bifogade i sina mail."

Länken behöver uppdateras.