The success of Richard Wilkinson and Kate Pickett’s The Spirit Level (reviewed here in August) in setting the terms for much political discussion unsurprisingly triggered a burst of publications taking a sceptical look at their case. Prime amongst these is Policy Exchange’s publication Beware False Prophets, by Peter Saunders, whose title gives you a fair clue as to its line.
As the book says on its back cover:
In The Spirit Level, Richard Wilkinson and Kate Pickett claimed that egalitarian societies benefit rich and poor alike. Crime rates are lower, infant mortality is reduced, obesity is less prevalent, education standards are higher, average life expectancy is longer, social mobility is more extensive, and so on. They concluded that we would all benefit from a more egalitarian distribution of income…
[However] their evidence is weak, their analysis is superficial and most of the correlations in their book do not stand up.
So how does the Policy Exchange publication come to this conclusion and is its argument well founded? The critique is based on three core points: how strong the data is for the 23 countries that Wilkinson and Pickett looked at, what happens if you look at more countries and what the evidence is when comparing individual US states with each other.
On the 23 countries some individual criticisms are made of specific graphs that carry weight. However, the criticisms are not consistent with each other across the different graphs – sometimes Saunders suggests the data is overly dependent on one country, sometimes on others. Picking and choosing which countries to exclude from graph to graph does not make for a coherent case that there is no underlying pattern. Moreover, the basic point Wilkinson and Pickett are arguing is buttressed by numerous other studies beyond their own (see a partial round-up here) and so picking apart some of their individual data is a bit like picking apart the methodology of one individual pollster in some polls; it is a long way short of undermining the polling industry overall.
A more general point made is that there is a consistent difference between Scandinavia countries and Anglo countries. The Scandinavian countries are generally more equal and perform better on education standards, life expectancy and so on, whilst the Anglo countries are less equal and perform less well. That raises the question as to what extent it is the greater level of equality that makes Scandinavia different compared with other factors.
Peter Saunders also looks at what happens to the Wilkinson / Pickett data if you add in more countries, finding that in many cases the relationship between inequality and outcomes goes. However, he does this in part by adding in many small countries – those with populations of 1-3 million. So his critique really amounts to saying that the relationships may be different in small countries from large countries. The existence of such a difference is neither inherently implausible nor does it alter the political implications for the UK (population: a lot more than 3 million).
Saunders also adds in some large countries to the data set, though as he acknowledges these are generally poorer than the Wilkinson / Pickett set. As their theory is deliberately and explicitly one for affluent countries only, saying that something different happens in less affluent countries compared to more affluent ones does not undermine thesis. Their own response to Saunders’s work does not, though, directly address the question of how sensitive their results are to quite where you draw the cut off amongst countries nor why some of the affluent countries Saunders looks at were excluded from their study.
This mixed picture is common across all of Saunders’s areas of criticism – a few points that stick, some that simply sharpen the definition of the Spirit Level‘s thesis as applying to medium to large rich countries, a few that Wilkinson and Pickett rebut and some where, even if you read their different accounts side by side (or look at what they said when appearing together, such as at an RSA meeting), they do not directly address each other’s points.
The ‘trump card’ that Wilkinson and Pickett have to play is the amount of other research that points in the same direction as their research. The two main weaknesses left are the question of whether there is really something else going on in Scandinavia and also (though not mentioned by Saunders) the point I made when reviewing the Spirit Level:
Not only does it rely overwhelmingly on comparisons across countries at the same point in time, rather than in tracking ailments varying over the years, the limited amount of such evidence deployed is almost all of the ‘inequality increased and then things got worse’ form. There is no automatic reason why, even if increasing inequality makes things worse, then decreasing it will make things better. The world is not always symmetrical. Moreover, even if the effect works strongly ‘in reverse’, is it the most cost-effective route to take? If inequality causes stress which causes social ills, is targeting stress going to be more successful?
My verdict then? Where the Spirit Level to have been original in the data patterns it found, these criticisms would add up to reasons to be cautious about its findings. However, because of the extensive other supporting data available, the findings still look solid even though – as I said in my review of it – the political policy implications are far more nuanced.
Amazon sell both The Spirit Level and Beware False Prophets and Policy Exchange both sell their pamphlet and provide it as a free pdf.
Follow @libdemvoice.org on Bluesky
Like us on Facebook
Subscribe to our feed
Sign-up for our daily email digest
Thought for the Day: "We know that at night-time someone goes by amongst the trees, but we never speak of it"
Order of Service
Neil Alexander selected for Inverness and Nairn
Welcome immigrants (2)
Mark Valladares
David Evans
Luke Magalhaes
Zachary Adam Barker
Thelma Davies
10 Comments
Do you mean a critique?
To repeat my main objection to this literature, we could all get Japanese/Norwegian levels of income inequality and Japanese/Norwegian levels of social solidarity among citizens if we adopt Japanese/Norwegian levels of ethnic homogeneity.
The case is tantamount to saying that the United States of America should not accept non-white immigration, and thus inflate its own numbers on almost all levels of social loveliness, regardless of what it means for those who would have otherwise immigrated.
I should point out that it is the critics of The Spirit Level who make the case for ethnic homogeneity as an alternative hypothesis, not the Spirit Level itself which ought to be read first before it is criticised. It is not a difficult book to read.
I’ve read The Spirit Level and the pamphlets attacking it. I’m also ideologically inclined towards a more equal society. However, I don’t agree with you that the book stands up to its critics.
Firstly, I don’t agree with you about the point about excluding outliers – that is the extreme examples. The examples given by Saunders and Evans are the USA for murder rate and Japan for longevity. If these extreme examples are excluded the fact is that the graphs do look very different. I don’t understand why you think it matters that the extreme is provided by different countries on different graphs. I don’t see the relevance of that. I think most people would agree that the USA’s lack of gun control, for example, has a significant effect on their murder rate, making the USA exceptional. This is a point about how you deal with statistical evidence and how you do social science. Excluding outliers is something I was taught to do when I did my social science degree. It is why we talk about median incomes, not the mean (average) income, for example. So this looks like bad social science to me.
Secondly, the fact that a bad piece of science reaches the same conclusion as a good piece of science doesn’t make it good science, so I don’t agree with you that if there are other studies that reach the same conclusions, this makes The Spirit Level good. However, the studies referred to in the Left Foot Forward article you link to do not actually reach the same conclusion as The Spirit Level. The Spirit Level claims very specifically that we are all better off in a more equal society, including the people who were at the top of the scale to start off with. The Marmot Review, just to cite one of the studies referred to in Left Foot Forward, looked at health inequalities and suggested that everyone in society should have the life chances which are now only enjoyed by the few. The Marmot Review did not say that the few would be better off as a result. So, the Marmot review does not back up The Spirit Level at all.
Finally, you are ignoring the fact that The Spirit Level has also been criticised, even if more mutedly, by social science academics on the left, who object that the bad science in the book, which is very easy to critique, will result their own more rigorous research on inequality being ignored.
Sorry, there was one other point, I forgot to make, about small countries. You are right that Saunders adds in very small countries, but you are ignoring the fact that the argument in the Spirit Level already relies on data from small countries. Sweden’s population is 9.3 million. Norway’s is just under 5 million and Denmark’s is 5.5 million. Several of the countries excluded by Wilkinson and Pickett are much larger than this, South Korea and The Czech Republic for example, and their claim that data was not available for these countries does not stand up to scrutiny. Their cherry picking of data could not be more obvious.
Any book that questions the status quo and challenges vested interests is bound to be attacked, crticised and smeared. This book is no different.
It’s a great read and is certainly thought-provoking. Read it yourself and draw your own conclusions.
Jane Leaper – you suggest that the USA is exceptional because of its lack of gun control, and that this makes conclusions about murder rates/unequal socities unreliable. However: the authors of The Spirit Level did not only analyse international data – they applied the same analysis to the 50 states of the USA, and found the same result.
Jane Leaper – one other point: ‘Excluding outliers is something I was taught to do when I did my social science degree.’ Is that not another term for cherry-picking?
Sid
The Kingdom of Erehwhon has 100 households.
10 of them have no income (the first centile – the palace slaves).
10 have 1 pence per annum (the serfs – the second centile).
10 of them have an income of 2 pence (the peasants – the third centile).
10 of them have an income of 3 pence a year (the yeomen – the fourth centile),
10 of them have 4 pence a year (the craftsmen- the fifth centile).
10 of them have 5 pence a year (the traders – the 6th centile).
10 of them have 6 pence a year (the poets – the 7th centile).
10 of them have 7 pence a year (the philosophers – the 8th centile)
10 of them have 8 pence a year (the soothsayers – the 9th centile).
10 of them have 1,000,000 pence a year (the Royal Family – the tenth centile).
Gulliver visits Erehwhon and looks around. He says to the King “Your majesty, your people are poor.” “How can you say that, replies the King. “The average income is 100003 pence per household. That’s a fortune”. “But don’t you think you should ignore the income of yourself and your relatives?” asks Gulliver. “But that would be cherry picking data” replies the King.
Sid again.
Saunders points out that not only is the USA an extreme outlier on the murder graph (it’s murder rate is far far higher than for any other advanced industrialised country), but if it is removed from the chart there is actually no correlation left between murder rate and inequality: Finland, one of the most equal societies having a higher murder rate than Singapore which is one of the most unequal.
In relation to the USA States graph, other critics have pointed out that there is an equal correlation between race and violence. I understand they have been accused of racism by Wilkinson and Pickett, but in fact all they are doing is pointing out the problem with monocausal explanations, and of interpreting a statistical correlation as a cause. The murder rate in the USA has multiple causes. Wilkinson and Pickett also ignore the fact that the homicide rate in the USA has been dropping at the same time that inequality has been increasing.
I suggest listening to the Radio 4 analysis podcast, which is even handed and includes academic criticism which is not from the right. It is several decades since I studies statistics, and I then went on to become a family lawyer, in which, thank God, statistics do not have a role to play. So, I’m not the best person to be explaining this stuff. However, if you look at the Wikipedia article for outliers, I think it explains it well enough. You might like to look up the difference between a mean and a median as well.