In our book The Spirit Level, Kate Pickett and I demonstrated that, first, many problems which are more prevalent lower down the social ladder are worse in societies with bigger income differences, and second, that almost everyone would benefit from reduced inequality. To some, however, these seem impossible notions. Writing in the August 2010 edition of Prospect, Matthew Sinclair from the Taxpayers Alliance claimed our research was “simply untrue.”
Sinclair believes he has spotted statistical sleights of hand that hundreds of fellow academics who reviewed our research papers for numerous journals have failed to detect. Decades of peer-reviewed epidemiological research, funded by research councils have, he imagines, been torn to shreds by Christopher Snowdon—author of The Spirit Level Delusion. While Snowdon is described as a “public health researcher,” in actual fact he has no public health qualifications and appears never to have published research in a peer-reviewed journal. Instead, his main contribution to public health is a diatribe against tobacco control and a denial of the ill effects of second-hand smoke.
What The Spirit Level shows is that more equal societies enjoy better physical and mental health, lower homicide rates, fewer drug problems, fewer teenage births, higher maths and literacy scores, higher standards of child wellbeing, less bullying in schools, lower obesity rates, and fewer people in prison. Furthermore, that more equal societies also have a stronger community life and are more cohesive. Over the years a large number of research papers have shown that one or other of these problems are more common in more unequal societies. Of course we could simply have picked out the most dramatic examples from other people’s work to show in our book. However, in order to show the remarkably consistent tendency for problems associated with relative deprivation to be more common in more unequal societies, we wanted to show that this pattern occurs in one problem after another, even when we use exactly the same group of countries and the same measures of inequality.
To rule out the effects of differences in material living standards, we looked only at the very richest societies where—in contrast to poorer countries—there is no longer any correlation between national income per person and outcomes such as health, happiness or wellbeing. Identifying the 50 richest countries according to the World Bank’s preferred method for classifying countries into “high”, “medium” and “low” income categories, we also removed countries with fewer than 3m people (to avoid tax havens), and those lacking internationally comparable income distribution data.
This left us with 23 rich market societies. We took our data from the best sources, such as the World Health Organisation, the United Nations and the World Bank. To double-check our findings, we then repeated our analyses for the 50 US states, to see if more unequal states showed the same consistent tendency to have more of these health and social problems. In almost 30 different cross-national analyses, we show the same tendency for one problem after another to be significantly worse in societies with bigger income differences.
Inevitably the choice of problems we examined was partly decided by the availability of comparable data, but no one has suggested that we do not discuss the most important problems with steep social gradients. But to safeguard against the criticism that we picked problems to suit our argument, we showed (initially in a paper published by the British Medical Journal) that the Unicef Index of Child Wellbeing is also strongly related to inequality. It has 40 components covering every aspect of child wellbeing, which we had no hand in selecting.
We made an absolute rule never to pick and choose data points to suit our argument. This is in sharp contrast to our critics who, with the sole purpose of reducing the statistical significance of each association, want to exclude some countries or US states here, others there or, failing that, to alter the picture by adding in poorer countries. However, suppose that, despite not accepting their rationale, we allowed all our critics’ ad hoc exclusions together—the Scandinavian countries, US, Japan, or southern states of the US? Even then, our index combining ten key health and social problems remains strongly related to inequality. So even if the links between inequality and individual problems can be made to disappear by dubious practices, the overall picture is actually extremely robust.
Our critics also ignore the fact that these relationships have been widely demonstrated by other researchers. For example, as early as 1993 in the Criminal Justice Review, Hsieh and Pugh reviewed 34 studies of income inequality and violent crime and found a consistent correlation between the two—the authors estimated that it would need 58 new studies which found no effect in order to overturn this result. But studies since then have continued to confirm the link.
Similarly, our review of research papers published in peer-reviewed journals found that the tendency for health to be worse in more unequal societies has been demonstrated well over 100 times (see Social Science and Medicine, 2006). Faced with research papers showing that this pattern is repeated among the regions of Russia, the provinces of China, the counties of Chile, or rich and poor countries together, we wonder what regions, provinces, counties or developing countries our critics would find excuses to exclude to deny a relationship?
Again in contrast to our critics, we offer a coherent theory of why so many health and social problems are linked to greater inequality. Rather than being caused directly by material conditions or being simply a reflection of selective social mobility sorting the resilient from the vulnerable, the link with income inequality suggests that the problems associated with social status are responses to the stresses of social status differentiation itself.
We remain puzzled by the stance the Taxpayers Alliance has taken to our work. As we point out, greater equality need not depend on high taxation. Within the US the state of New Hampshire has amongst the lowest taxes. It has no income tax or state sales tax but, like other more equal states, it does well in terms of a host of social measures including rates of infant mortality, homicides, teenage pregnancies, imprisonment, levels of trust and children’s school performance. It stands as an example of the benefits of a fairer and more equal society.
Trew
I don’t doubt the quality of your data and believe that your methods of establishing correlations are sound. The question is not one of correlation, however. Here (and in your book) you do not even mention the endogeneity problems inherent in all these questions. Without doing so, one cannot say on the basis of simple cross-sectional correlations that lowering inequality will lead to better economic and social outcomes.
Your language is mostly careful in this regard and so I’m sure you understand this. The problem is that a policy maker might not. It is potentially a waste to funnel resources to reduce inequality when the benefits you only associate with greater equality will not necessarily follow as a result.
Woods&Trees
Though the technical term is not used, the enogeneity problem is dealt with in the discussion in chapter 13 of The Spirit Level.
James
I just want to commend the authors on their response to criticism. I am fascinated by their research but I am hardwired to be sceptical due to a persistent university professor who drilled into me the need to doubt and check everything. What we can say at this stage is that the authors have subjected their research to rigorous analysis by peers and have also scientifically refuted much of the criticism. If only those who criticise them were the same.
It is laughable, and yet also disturbing, that their main critic is described as a public health researcher when in fact they are simply a public health researcher doubter, with as far as one can tell, no expertise and no background in the subject.
I am all for scepticism and I am all for juvenile politicos playing games, but when the issues are this serious, could the critics not try and find someone with a bit of scientific kudos to mount the attack?
Christopher Snowdon
I’m pleased to be able to agree with Wilkinson and Pickett (W & P) on one point. I’m not a public health researcher and was surprised to be described as such. That, sadly, is where agreement ends.
Wilkinson and Pickett once again imply that they are merely the messengers of a scientific consensus and that there are 100s of peer-reviewed studies saying the same thing. Can we please put this one to bed? Even if quantity was a substitute for quality, the argument does not apply here. There is a large body of conflicting research about health and inequality and a smaller body of research studying violence and inequality. Both are hotly debated, not least because it is very difficult to isolate the effects of income inequality from the effects of low income.
Beyond this, Wilkinson and Pickett are out on their own, making claims that have virtually no support in the scientific literature. In contrast to what he says in the third paragraph of this rejoinder, Wilkinson recently told the magazine International Socialism:
“There are about 200 papers on health and inequality in lots of different settings, probably 40 or 50 looking at violence in relation to inequality, and very few looking at any of the other things in relation to inequality. In a way, the new work in the book is all these other variables—teenage births, mental illness, prison populations and so on—and the major contribution is bringing all of that into a picture that had previously been just health and violence.\
W & P confuse making assumptions based on other people’s research with having those people actually agree with them. For example, they cite studies that quite reasonably associate overeating with stress, but it does not follow that obesity rates vary internationally because the population is stressed about inequality. At best, this is speculation.
W & P continue to cite perfectly sound studies showing there to be social gradients to health and social problems as evidence that inequality affects a nation’s overall performance. It does not. These are completely different issues.
Which leaves us with W & P’s own evidence, which relies on comparing whole countries, a notoriously unreliable method which leaves unlimited scope for misinterpretation. The criticisms recently made of this evidence by myself and others closely echo criticisms made in peer-reviewed journals when Wilkinson used similar methods in the past. They also echo criticisms made by the few serious academics who have reviewed The Spirit Level.
Anyone who believes that W & P “never pick and choose data points to suit our argument” should compare references 2 and 6 in The Spirit Level (p. 271) and ask themselves why one year’s data were used for one graph and another year’s data used for the next. Anyone who believes that they use “the same measures of inequality” should turn to page 224 and ask why a dramatically different measure of inequality was preferred when working hours were studied (clue: see how it looks when we use W & P’s more usual measure of inequality). Anyone believing that they have not “picked problems to suit our argument” might ask why they show how much overseas aid is given by a country’s government, but do not show how much is given privately (there is no correlation with inequality when the two are combined).
As for always using the same group of countries, one of The Spirit Level’s most serious flaws is the baffling assumption that “rich market societies” come in batches of 50. If there is to be a cut-off point beyond which economic growth has “largely finished its work”, it should be based on something more than a round number. Without a convincing justification for why places like the Czech Republic and South Korea – let alone Hong Kong – cannot be considered rich market societies, we must ask the next question: why do these societies conspicuously fail to fit Wilkinson and Pickett’s theory? The United Nations classes these countries as being of “very high human development”, why doesn’t The Spirit Level?
I hope that readers will take the time to look at these issues themselves, but, if not, they should at least take a deep breath and ask themselves which is more plausible – a theory that seeks to explain the workings of whole societies by reference to a single factor, or one that says that a country’s performance is the result of countless historical, geographical, political, legal, demographic and economic factors, of which the public’s response to income inequality may or may not be one.
David Colquhoun
I found the political approach of the Spirit Level Delusion not at all to my taste, and I’m glad to see a response to it.
Nevertheless sometime I’d like to see responses to the specific criticisms that Snowdon raised about choice of countries and choice of times. I appreciate that Prospect may not be the place to do it, but it would be valuable for people who like detail.
Kathryn Busby
@ David
You can find Kate and Richard’s detailed response to the specific criticisms at:
http://www.equalitytrust.org.uk/resources/response-to-questions
Christopher Snowdon
To which…
Reply to Wilkinson and Pickett
Sanandaji et al. response to Equality Trust
minifesto
I find it interesting that New Hampshire is cited as an example of a more equal society having the lowest taxes. I left the UK in 1959 due to the system of punitive taxes inflicted by a left wing government. I should really thank them as apart from my decision to marry my wife it was the best decision I have ever made.
Richard Lawson
Snowdon objects that W&P’s method “relies on comparing whole countries, a notoriously unreliable method which leaves unlimited scope for misinterpretation”. However, he uses this method in his book, adding in three countries, the Czech Republic, Slovenia and Hungary, which has the desired effect of reducing or inverting the gradient in W&Ps results.
Clearly, there are many factors which go to generate the data that any one country will display. Overall wealth, culture, traditional social bonds, climate, ethnic mix, pollution levels and politics; these are a few that come to mind. Considering these many factors makes it all the more remarkable that W&P have produced so many clear correlations. They mention in TSL that the multifactorial nature generates the wide scatter on so many graphs. If we had a reliable means of weighting these factors, producing an index of income equality modified by other variables, the scatter might be reduced.
The effect of Snowdon’s introducing the trio of ex-Soviet countries is interesting. Although more equal, they tend to score higher with some social problems, which suggests that there are special factors associated with the rigid political system that they are emerging from. Communism may have produced income equality, but there was a huge inequality of political power between the people and the Party, and Marmot’s work suggests that this would produce health problems.
Having read W&P and Snowdon, I am confident that W&P’s thesis is robust, but capable of adapting to and incorporating some of the dimensions that Snowdon’s book suggests – once they have been supported by serious research.
R
I don’t have any intention of disputing statistical research (or expertise to do so), but the obvious problem here is that identified by Trew. In layman’s terms: is this correlation or causation?
For what it’s worth, I find it intuitively unsurprising that countries within this sample that are more equal also have fewer social problems etc. The context (culture etc) that leads to choices and preferences causing a country to be more equal may well also mitigate such problems.
In short, loads of statistical evidence demonstrating this correlation doesn’t imply that we should do anything specific from a public policy perspective.
TIMOTHY_ROBERTS
Good stuff! I look forward to the rebuttal to the rebuttal to the rebuttal. Meanwhile:
a) arguments based on differences in patenting propensity between different countries make me extremely sceptical (given widely differing laws and practices);
b) can any neutral Swede report on the checking or non-checking of travel tickets in Stockholm?
Falafulu Fisi
The first question that I would ask the authors of “Spirit Level”.
What makes you think that linear trend analysis is applicable or appropriate here? The data is so sparse and also too noisy (points are scattered everywhere) to even assume that there is a linear trend that exists. Did you just use linear trend simply because that’s what everyone is using when analyzing data?
Why not attempt to fit the data into a polynomial equation? Note that linear trend is simply a polynomial equation of order 1. The reason I am bringing this up, because you assume a linear relationship from the world go. I can fit your data into polynomial equation that will give a best fit with minimal error (goodness of fitness). I can guarantee you that if a higher order polynomial if fitted to the data, then the functional relationship is different (not linear – ie, order 1 polynomial). Now, such polynomial may be monotonically decreasing or increasing (I don’t know because I haven’t analyzed the data yet). This means that anything goes. It could be a decreasing relationship or it could be an increasing relationship depending on the polynomial coefficients found from the fitting.
I find the data analysis simply cartoonist because not only it is static (real society is dynamic) but the statistical analysis may be inappropriate.
PS : I specialize in scientific computing and developing numerical analysis algorithms for high-speed/high-end data analysis is my domain of expertise.
James
The equality act is a sham! It pretends to give with one hand and take away with the other. If the whole process is scuppered by leaving existing laws and legislations out of its scope then it will not go anywhere in resolving many of the problems in our society that allows discrimination on the basis of Age, Disability, Gender, Race, etc.
There are nationality laws for instance that discriminate on age and gender and yet they go un-addressed and ignored where entire groups of children of British citizens have been disenfranchised simply because of when they were born.
Sign up to the petition
Please sign the online petition for children of British fathers and mothers to be treated equally at http://www.gopetition.co.uk/petitions/british-nationality/signatures.html