Wednesday, 29 March 2017

The Spirit Level: A Rebuttal - Part I


This is a rebuttal of the 2009 book ‘The Spirit Level’ (subtitle: ‘Why Equality is Better for Everyone’) by Richard Wilkinson and Kate Pickett.

I have split it into four parts and intend to publish one part per week on the blog. This is part I and covers the front matter, and part I of the book. Parts II and III will cover those respective parts of the book and part IV will be a collection of some other thoughts and conclusions of sorts.

‘The Spirit Level’ was originally strongly recommended to me by a close friend whose thoughts and opinions I very much value and respect and I thought that I’d note down my general thoughts and impressions as I read through it.  I originally set out simply to write a review of ‘The Spirit Level’.  Through the course of reading the book and ‘The Spirit Level Delusion’ by Christopher Snowdon, a direct response to and critique of ‘The Spirit Level’, this has ended up becoming more of a rebuttal.

Before we get into the rebuttal/review, I want to note that although many of my comments may seem rather negative and overly critical, it was not my intention to ‘disprove’ Wilkinson and Pickett’s thesis.  My intention is to improve my own understanding, and if possible, to help improve others understanding too, and to attempt to arrive at a position that is as close to the truth as is possible.  If many of my comments seem particularly critical it is only because I have felt the need to elaborate more on points of significant doubt or major disagreement with the authors’, than on points of agreement, particularly points of general or widespread agreement.

On a similar note, I am not necessarily opposed to redistribution of income or wealth on ideological grounds.  There are good theoretical arguments to be made in favour of redistribution of wealth and/or income; I will not get into the details of these here, suffice to say that I find some of these arguments at least somewhat compelling.  Wilkinson and Pickett’s arguments I generally find far less compelling for a variety of reasons, which we will get into over the next 66 pages (yes 66 pages – this is pretty comprehensive).

For reference I am using the UK paperback version of ‘The Spirit Level’, published by Penguin, with a new postscript, in 2010.  Similarly, I am using the UK paperback version of ‘The Spirit Level Delusion’ published by Little Dice, also in 2010.  Any page, section or figure/table numbers I give refer to these versions of the books, page references or figure/table numbers may not exactly correspond in other versions or formats of the books.

Without further ado, on with the review.


The subtitle of ‘The Spirit Level’ is ‘Why Equality is Better for Everyone’, which seems to presuppose the conclusion, although the charitable reading is that the authors are simply writing with the advantage of hindsight after having done the research.


Reading the preface, the following statement caught my eye:

“At an intuitive level people have always recognized that inequality is socially corrosive.”

I really shouldn’t need to point out, but I will, that:
  1. People’s intuitions are often misguided and lead to incorrect conclusions.  Not always, but often – often enough that we should always stop to think things through more carefully than just accepting the direction our intuition or ‘gut feel’ guides us towards.  We should not necessarily be swayed simply by what has an intuitive appeal (or indeed by what we intuitively disapprove of), we need to endeavour to consider the evidence impartially (in so far as that is possible), without presupposing our conclusions.
  2. Describing inequality as “socially corrosive” is loaded, emotive language, not to mention vague and ambiguous.  It appears to have been used to evoke a specific emotional response from the reader.

I also note that the authors do not define at the outset precisely what they mean by ‘inequality’.  There are all sorts of different inequalities that could be being referred to: inequalities of wealth, health, leisure (time), income, education, political power, legal status, physical inequalities, intellectual inequalities, emotional inequalities, and inequality of opportunity (which may come about for a great many different reasons).  Not until more than halfway through chapter 2, do Wilkinson and Pickett state explicitly that in ‘The Spirit Level’ they are here discussing income inequality.

Part One: Material Success, Social Failure

Chapter 1: The end of an era

The first chapter of the book opens with the claim that it is:

“...a remarkable paradox that, at the pinnacle of human material and technical achievement,we find ourselves anxiety-ridden, prone to depression, worried about how others see us, unsure of our friendships, driven to consume and with little or no community life.”

I do not see the paradox, far less find this remarkable.  People have always been, and always will be, understandably anxious and worried about their friendships and social interactions and about how others see them.  Well, most people at least.  This is a perfectly natural result of our evolution as social animals and is to be fully expected.  There is no obvious link between our level of technological advancement and/or material prosperity on the one hand and our social and interpersonal anxieties on the other.  Until and unless the authors can put forward a compelling case otherwise, this need not be considered further.

The authors go on to state:

“Economic growth, for so long the great engine of progress, has, in the rich countries, largely finished its work.  Not only have measures of wellbeing and happiness ceased to rise with economic growth but, as affluent societies have grown richer, there have been long-term rises in rates of anxiety, depression and numerous other social problems.  The populations of rich countries have got to the end of a long historical journey.”

This is a remarkable claim.  It is also deeply unscientific, on par with claims that humans are the ‘most evolved’ species or that we’ve come to the end of our evolution, or claims that we have already invented everything worthwhile inventing.  Economic growth continues (albeit slowly in much of the wealthier countries considered in ‘The Spirit Level’ and with various other caveats) whether Wilkinson and Pickett believe it or not and whether they believe it is or is not beneficial.  It has certainly not “largely finished it’s work” in the same sense that evolution has not “largely finished it’s work”.  For examples, see here, here, here, here, here, here, here and here.

Also in chapter one, Wilkinson and Pickett consider differences in average life expectancy and self-reported happiness between a range of countries and how both vary with gross national income (GNI) per capita.  The data show large, rapid gains in both life expectancy and happiness for increasing GNI per capita, for those countries in the lower half of the income distribution (below roughly $20,000 per person), but then the trends more-or-less level off, showing little further improvement in either lifespan or happiness with increasing GNI per capita for those countries in the upper half of the income distribution (above around $20,000 per person).  As the authors state, this result probably shouldn’t surprise us and can largely be explained by the picking of low-hanging fruit in terms of healthcare, sanitation, nutrition etc. and by the law of diminishing returns (diminishing marginal utility): each extra dollar you earn is worth less to you than the one before it.  This is neither new ground, nor controversial and shouldn’t surprise anyone.  It is widely acknowledged that, all else being equal, happiness increases with income, but only up to a point.  That point being an income level at which you can reasonably comfortably pay for a roof over your head, clothes on your back, food on your table, schooling for your kids and a pint in the pub.

Once you’re earning enough to pay for the essentials and a little bit of leisure/entertainment, and you’re not constantly worrying about paying your rent/bills, any extra income is not necessarily going to have a substantial effect on your overall wellbeing or happiness.  For example, if you give a pay rise of £5,000 to someone currently earning only £15,000 that will make a massive difference to their life and wellbeing.  However, if this person is already earning a six-figure salary, the same £5,000 pay rise probably is not going to affect them all that much.  File this under: ‘so obvious it shouldn’t need to be said’.

Yes, I also know it’s “not possible” to make interpersonal utility comparisons.  Scott Sumner attempts to deal with this problem here.  As with Professor Sumner’s example, I can’t prove that a £5,000 pay rise would make less of a difference to someone earning £150,000 than to someone earning £15,000.  It is just an educated guess based on 30 or so years of observation.

This undermines the general thesis of the book.  By pointing out that the same additional income is far less significant to the already wealthy, it illustrates that everyone is already more equal than one would suppose by narrowly considering only income (or wealth).

What’s the actual material difference between someone who earns £15,000 per year and someone who earns £150,000?  Narrowly focussing exclusively on their incomes, the latter is ten times better off.  But are they really ten times better off, in terms of actual wellbeing or quality of life?  This is very difficult to answer because ‘quality of life’ is so subjective and because of the aforementioned difficulties of making interpersonal utility comparisons.  However, it is telling that Wilkinson and Pickett don’t even ask this question.

Here’s my amateurish attempt at an answer, this argument is not original to me – I have come across is several times in different forms, but can’t remember where I first encountered it:

Take the following hypothetical example:
Rich works as an investment banker, a job in which he earns what would be considered by most a very generous salary: £1,000,000 per year.  Joe works as a bus driver and earns £20,000 per year.

There’s no doubt that Rich is much better off financially than Joe.  Joe will be lucky to earn in an entire lifetime what Rich can earn in just one year.

How much better off is Rich, compared to Joe?

One possible answer is that Rich’s salary is 50 times Joe’s salary, therefore Rich is 50 times better off than Joe.

The above answer has an appealing simplicity to it, but I don’t think it paints the whole picture.

In strictly financial terms, yes Rich is 50 times better off than Joe.  But surely we care about more than just how many digits appear on people’s paychecks?  We also care about people’s wellbeing or quality of life.  However, these are far more abstract and difficult to measure or observe than people’s salaries or bank accounts.

Is Rich’s wellbeing or quality of life really 50 times that of Joe’s?  I’m not sure it’s possible to quantify such things with any amount of precision or confidence, but here is one (enlightening) way to look at it: the millionaire may have multiple houses, but they can still only stay in one at a time; they may have more pairs of trousers, but they can still only wear one at a time; they may eat higher quality food, but they can still only eat one meal at a time; they may have a newer, shinier, faster car, or even a fleet of cars, but they can still only drive, or be driven in, one at a time – a 7 year old Hyundai may not be as cool or as fast around a track as a new Bugatti, but it still performs the same essential function.  Having 7 cars to choose from – a different one for every day of the week – doesn’t necessarily make your life 7 times better than if you had just one.

See also here, here, here and here.

In Chapter 3 of ‘The Spirit Level Delusion’, entitled ‘The pursuit of happiness’, Christopher Snowdon points to an even more fundamental point regarding the Income vs. Happiness and Inequality vs. Happiness relationships looked at by Wilkinson and Pickett.  Snowdon points out that the ‘levelling off’ that is observed in Figure 1.2 of ‘The Spirit Level’ can likely be largely explained by the fact that in most wealthy countries reported happiness is close to 100% and, as a matter of arithmetic, it is simply not possible for it to exceed 100%.

Here’s Snowdon:

“...happiness cannot rise above 100% and no one argues that money can guarantee happiness for everyone.”

He asks rhetorically:

“What socialist miracle do Wilkinson and Pickett have in mind that would propel happiness beyond the 100% that would be required for the curve to keep rising?”

He continues:

“Having dug out the World Values Survey to show the levels of happiness for dozens of countries against income, you might expect Wilkinson and Pickett to draw up another graph showing happiness against inequality.  Considering the overall ‘equal societies are happier’ hypothesis, it’s hard to believe such a graph does not appear.  It should show a downward trend, with the most equal countries having the highest scores and the least equal countries having the lowest.”


“...this is not what happens at all.”  “...there is no correlation with inequality.”

“There is, however, a clear correlation with income.”

This can be seen by comparing Figures 3.2 and 3.3 on pages 54 and 55 of ‘The Spirit Level Delusion’.  As Snowdon points out, there is no correlation between income inequality and rates of self reported happiness for the set of rich countries considered (R² = 0.0024), but there is a moderate positive correlation between GNI per capita and self-reported happiness (R² = 0.387).

Chapter 2: Poverty or inequality?

In chapter 2 of ‘The Spirit Level’, the authors compare the level of income inequality between a group of 21 of the World’s richest, most developed countries – and between each US state – and consider whether there are any correlations between income inequality and a variety of measures of health and social wellbeing, which they list as:
  • level of trust
  • prevalence of mental illness, including drug and alcohol addiction
  • life expectancy and infant mortality
  • prevalence of obesity
  • children’s educational performance
  • number of teenage births
  • number of homicides
  • imprisonment rates
  • social mobility (not available for US states)

They combine all of the above (with each of the above items carrying equal weight) into what they refer to as an ‘Index of Health and Social Problems’.  They go on to show a strong correlation between income inequality and this index (particularly so between the group of rich countries considered), but only a weak correlation between gross national income (GNI) per capita and this index.  They do likewise for another index, the UNICEF ‘index of child wellbeing’.

Figure 2.2 (p. 20) shows a very clear correlation between income inequality and Wilkinson and Pickett’s ‘Index of Health and Social Problems’ for the 21 developed countries considered.

Figure 2.3 (p. 21) shows a weaker trend between national income per capita and the ‘Index of Health and Social Problems’.  However, with the notable exception of the USA – which is an extreme outlier on this measure – there is still a trend here.  For example, the three richest countries considered: Norway, Switzerland and Denmark all score better than the three poorest: Portugal, Greece and New Zealand.

Figure 2.4 (p. 22) shows a correlation between income inequality and the ‘Index of Health and Social Problems’ for the US states.

Figure 2.5 (p. 22) was the one that really interested me though.  Wilkinson and Pickett claim that it shows only a weak correlation between national income per capita and the ‘Index of Health and Social Problems’ for the US states.  However, what I thought notable was the large empty space in the top right of the figure – indeed, almost the entire top right half of the graph is empty!

This shows that whilst there are both rich and poor states which do well on the ‘Index of Health and Social Problems’, all of the states which do badly are poorer states.  For example, all of the states with an average ‘National income’ per person of over $25,000 are in the better half of performers when it comes to the ‘Index of health and social problems’.  Sure, there are only 4 states with an average income per person in this range, but this is indicative of the overall trend.  Conversely, the 7 worst performing states on the ‘Index of health and social problems’ are all in the lower half in terms of average incomes.

Georgia and Nevada are the only two states that look particularly bad by this measure; both do pretty poorly on the ‘Index of health and social problems’ despite not being particularly poor states.  However, inequality does not look like a promising candidate for an explanation here.  Whilst Georgia is one of the more unequal states (it ranks about 13th most unequal overall), Nevada is not (it appears to rank about joint 19th most equal, along with Montana and Washington).

It is notable that there are no rich states whatsoever with poor ratings on the index, regardless of the level of inequality in those states.  This one chart devastatingly undermines the overall thesis of the book, as it suggests that absolute wealth is important in avoiding bad outcomes.

Overlooking this problem entirely, Wilkinson and Pickett posit that the relationship between income inequality and these various health and social problems is a causal one.  They confidently assert:

“This evidence cannot be dismissed as some statistical trick done with smoke and mirrors.”

However, this claim is never substantiated.  Anyone with a cursory knowledge of statistics can see how the sub-setting of countries (after all, only 21 countries are considered by the authors) may result in a type 1 error (a false positive); this risk is expertly explained by Fallacy Man on The Logic of Science blog.  I’m not saying that Wilkinson and Pickett necessarily have made a type 1 error due to the way they’ve sub-setted the data, only that it’s a possibility and that it’s incumbent on them to demonstrate that they’ve taken reasonable steps to preclude this possibility.  This is something which they have not done.

We also don’t know how many different relationships the authors’ tested in order to find the correlations that their thesis rests on.  If you look at enough permutations of enough variables, you’re bound to find some that appear to be correlated as a result of pure chance.  Findings that have only a 5% probability of being the result of chance come up 1 time in every 20 after all.  Randall Munro nicely captures this idea in this xkcd comic.

Also in chapter 2, Wilkinson and Pickett repeat their remarkable claim that we have:

“ to the end of what economic growth can do for the quality of life...”

and that:

“Having come to the end of what higher material living standards can offer us, we are the first generation to have to find other ways of improving the real quality of life.”

As already observed, this is a remarkable and deeply unscientific claim which has no supporting evidence. Indeed, such a claim is at odds with history, from which it can be observed that economic growth has been by far and away the primary driver of improvements in quality of life for ordinary people for all of recorded human civilisation.  Extraordinary claims require extraordinary evidence; the authors have provided precisely zero evidence for these assertions.

The whole chapter seems to have been written begging the question that income inequality causes all sorts of health and social problems.  But by the end of chapter 2, the authors have yet to backup this claim.

I’d like to give the authors the benefit of the doubt, I really would, but I just can’t suppress the feeling that they’ve simply set out to confirm their priors.  I hope that this is just the result of careless writing and/or sloppy editing and that the research Wilkinson and Pickett have carried out has been intellectually honest and scientifically rigorous, but this is far from apparent from my reading of their book.

Chapter 3: How inequality gets under the skin

In chapter 3, the authors describe why they believe humans are so vulnerable to inequality, and discuss the rise in anxiety in recent decades and the link between income inequality and social status.  Significantly, the authors point out, that they do not attribute the rise in anxiety to income inequality; they state, on page 35:

“That possibility can be discounted because the rises in anxiety and depression seem to start well before the increases in [income] inequality...”

However, this is somewhat at odds with the chapter’s title and with the general overarching narrative of the book.  There is a subsection of this very chapter titled ‘Inequality Increases Evaluation Anxieties’ which consists of nothing but conjecture and runs contrary to the statement quoted above.  This strikes me as disingenuous at best, and even pernicious.  Wilkinson and Pickett openly acknowledge that there is no evidence that income inequality causes anxiety, and then – within ten pages of this admission – assert that income inequality nevertheless causes anxiety.  The mind boggles at the level of doublethink required here.