Wednesday, 27 November 2013


This website exists as a forum for additional notes and discussion related to the book The Spirit Level Delusion. Some of this information takes the form of extended footnotes, which can be accessed at the right-hand side of this page.

The meat of the argument is, of course, in the book itself—which is available here (UK), here (USA) and here (Sweden). If you haven't read it, my articles in the Wall Street Journal and Spiked Review of Books give a brief overview.

Please scroll down for more recent entries, including a fact-checking of the response given by The Spirit Level's authors to my 20 Questions, a rebuttal to their article in Prospect magazine and some highlights from Kate Pickett's interview on BBC Radio 4's More or Less.

You can also read (for free) the new chapter included in the second edition of the book, which deals with Wilkinson and Pickett's response to criticism. Download Chapter 10 as a PDF.

"A devastating critique"
The Economist

"If you haven’t read a book that made you laugh out loud on the bus or the Tube in a while, try Christopher Snowdon’s superb release, The Spirit Level Delusion. But the book’s subtle humour is not the reason I am recommending it. The Spirit Level Delusion is, above all, a book that delivers and goes well beyond the promise of its subtitle – 'fact-checking the left’s new theory of everything'... It may well be that the next big battle for a free society will be fought against the new anti-wealth egalitarianism. Christopher Snowdon has provided defenders of freedom with powerful ammunition."
— Kristian Niemietz, Institute of Economic Affairs

"Snowdon picks so many holes in the theory that were it a building it wouldn’t be passed as structurally sound by the most crooked of third world local government surveyors... Next time someone starts spouting off about “equality” – a goal that has dug more graves than all the gods in history combined – send them a copy of Snowdon’s excellent book and make sure they read it from cover to cover." 
Ed West, The Telegraph

"This year’s most important publication"
— James Delingpole, The Telegraph

"The myth of inequality as the root cause of just about all social ills is dismantled... Snowdon’s thorough appraisal of available data and literature, and examination of alternative causes – all underpinned by acerbic wit – sees to that." 
Sam Hamilton, Medical Writing

"The Spirit Level Delusion not only successfully and dramatically undermines much of the evidence in The Spirit Level, but also takes on the other fashionable opponents of economic growth... His engaging discussion unpicks the evidence of the anti-growth brigade and demonstrates that it is selective and partial. This book is excellent “tube reading”.
Philip Booth, City AM 

Homicide and teen births

Part of The Spirit Level hypothesis is that teen births and homicide are somehow caused by inequality (chapters 9 and 10). However, Wilkinson and Pickett face a problem insofar as inequality has been rising in most countries for many years while rates of teenage births and murder have been falling. They attempt to square this circle on page 142 when talking about their worst performer, the USA.

Homicide rates in America, after rising for decades, peaked in the early 1990s, then fell to their lowest level in the early 2000s. In 2005, they started to rise again. Similarly, after peaking in the early 1990s, teenage pregnancy and birth rates began to fall in America, and the decline was particularly steep for African-Americans. But in 2006, the teenage birth rate also started to rise again, and the biggest reversal was for African-American women.

The standard Gini measurement of inequality (see below) isn't very helpful to Wilkinson and Pickett in this instance since it shows inequality to have been rising pretty much continuously since the early 1990s which is exactly when the homicide and teen birth rates started to fall.

Faced with this obstacle, they resort to an obscure discussion paper which paints a quite different picture of the US trend, with...

...inequality rising through the 1980s to a peak in the early 1990s. The following decade saw an overall decline in inequality, with an upturn since 2000.

This finding is contrary to all other evidence and has been described as "the equivalent of the sun orbiting the earth", but it nevertheless allows Wilkinson and Pickett to triumphantly conclude...

So there is a reasonable match between recent trends in homicides, teenage births and inequality—rising through the early 1990s and declining for a decade or so, with a very recent upturn.

In other words, the 1990s saw falling inequality and therefore falling rates of teen births and homicide, whereas the Noughties saw rising inequality and therefore rising rates of teen births and homicide.

This is patently at odds with the facts. As I mentioned in The Spirit Level Delusion, it is a stretch to say that the homicide rate "started to rise again" in 2005. In fact, there was a tiny blip in 2005-06 when the murder rate went from 5.5 per 100,000 to 5.7 (see below). After that, the downward trend returned. By 2011, it was 4.7 per 100,000.

Moreover—and the reason for this little blog post—I recently had cause to look up the US's teen birth rates which last year "reached historic lows for all age and ethnic groups". Here too we see a little blip in the middle of the Noughties followed by a continued decline.

The Spirit Level was published in 2009 and so the data from the most recent years were naturally not included. Nevertheless, it was, at best, rash of its authors to present a slight upturn in the figures as the start of an inequality-fuelled rising trend. No matter which set of inequality figures one uses—and the Gini figures are vastly more credible—Wilkinson and Pickett's argument does not stand up.

We now know for certain that the small increase in the homicide and teen birth figures in 2005-06 was just a blip. The facts are quite clear. Inequality has been rising in the USA while the homicide and teen birth rates have been falling. There is simply no correlation between these variables. We need to look elsewhere for an explanation.

Tuesday, 13 November 2012

The Spirit Level Delusion: Chapter 10

As an addendum to The Spirit Level Delusion I have also written an additional chapter discussing Wilkinson and Pickett's response to the criticisms The Spirit Level has received since it was published. As a courtesy to readers, this is available here as a free download (PDF).



Prof. Colin Mills on The Spirit Level

Colin Mills, professor of sociology at Oxford University, makes some very worthwhile observations about The Spirit Level on his blog. Perhaps mischievously, Mills describes himself as "merely a sociologist" (Richard Wilkinson's qualifications are in history and sociology) and he confines himself to the chapter about inequality and health—Wilkinson's speciality. He notes that Wilkinson and Pickett misrepresent the evidence base, ignore studies which do not support their hypothesis and fail to adequately reply to their critics.

On a personal level, I was interested to see that he spotted that Figure 2.5 of The Spirit Level indicates that there is a relationship between health and income in the US. This was one of the graphs that first got me interested in testing The Spirit Level hypothesis back in 2009. Wilkinson and Pickett do not comment on the obvious fact that all the states which score poorly on their 'index of health and social problems' have a lower-than-average per capita income. This is absolutely crucial in understanding why they find relationships, albeit often slight, with inequality elsewhere—the least equal states are also the poorest.

But wait a minute. My first reaction when I looked at Figure 2.5 was that actually there is quite a clear relationship between aggregate income levels and health/social problem outcomes in US states. Richer states have better outcomes even without special pleading for the influential peculiarities of Idaho, Wyoming, Montana....

At this point I began to lose a little confidence as you appear not to be playing it straight with your readers, not all of whom, I assume, will look too closely at the figures or be alert to what could be considered a bit of textual leger de main.

Indeed. I have briefly discussed this issue here.

Mills has done what 99 per cent of The Spirit Level's readers have not done (and should not have to do) and looked at the scientific literature which Wilkinson and Pickett so often misrepresent.

But I do want to read the systematic reviews myself and not just rely on being told what they say. Just as well really, for if I had to rely solely on the 456 citations in The Spirit Level I would never have come across the two lengthy articles written by Lynch et al. in (2004) for the Milbank Quarterly entitled 'Is income inequality a determinant of population health?'. Luckily I have well informed colleagues who could point me in the right direction. Of course you know what Lynch et al. conclude from their careful scrutiny of "98 aggregate and multi-level studies examining the association between income inequality and health". For the benefit of those readers who can't penetrate the pay-wall let me quote from their conclusions (pp 81):

"Among affluent countries does income inequality help explain international differences in population health? The evidence suggests that income inequality is not associated with population health differences - at least not as a general phenomenon - among wealthy nations. Do levels of income inequality explain regional health differences within countries? In aggregate-level US studies, the extent of income inequality across states and metropolitan areas seems reasonably robustly associated with a variety of health outcomes, especially when measured at the state level. In multilevel US studies, using both individual and aggregate data, the evidence is more mixed, with state-level associations again being the most consistent. For other countries, the aggregate and multi-level evidence generally suggests little or nor effect of income inequality on health indicators in rich countries...but there may be some effects in the United Kingdom." [my emphasis].

Strange that you don't mention Lynch et al.'s papers (I know you cite them in your own review article but without, as far as I can see, any serious effort to explain why they get such different results to your own). Odd in several respects: firstly it was probably the most comprehensive independent (ie not counting papers written by yourselves) systematic review of the evidence on health then published when you were drafting The Spirit Level. Secondly, because it is not completely unfavourable to your position. After all it concludes that there is some evidence of an income inequality effect in US state level data (and possibly in the UK) - though their second paper which examines time-series data casts more doubt on the US case. I simply cannot understand why you fail to mention it, or if it is flawed in some way, rebut it, refer to your own rebuttal published elsewhere (if there is one) or the rebuttals of others (if there are any). Over and over again you tell us that the weight of the evidence is on your side and that there is a broad consensus amongst experts working in the field. But this simply isn't true, is it? At the very least your now perplexed readers could be forgiven if they find your omission, well, a little shifty.

He also comments on Wilkinson and Pickett's misleading response to their critics, including John Goldthorpe (also a professor of sociology at Oxford University who, like Mills, identifies himself as politically left-wing).

My colleague wrote a somewhat critical review of the book which was published in a well known peer reviewed journal. After reading your "reply" I was again deeply puzzled. What you say has, at most, only tangential relevance to the substance of my colleague's criticisms. The casual reader of what you write would come away with the impression that some sociologist had made a rather footling objection to the effect that you hadn't paid enough attention "to the vast amount of careful work now available on social class classifications" [your words] - surely a case of the cobbler only having eyes for leather. But wait a minute, that is not at all the substance of the critique.

He also makes the point, which Peter Saunders makes in his critique Beware False Prophets, that Wilkinson and Pickett conflate 'more equal' societies with classless societies, and yet their most equal country—Japan—is anything but classless.

In Japan income inequality is less marked than in many developed nations and health outcomes are comparatively good. This would seem to conform to the Wilkinson-Pickett party line. But it is also the case that Japan is a highly status conscious (in the sociological sense) society. It is obligatory to acknowledge inferiority and superiority both in terms of behaviour and in terms of the use of honorifics. In any unfamiliar social situation the initial process of figuring out who is relatively inferior to whom is a source of considerable anxiety.

To put it simply, in Japan systematic inequality is strongly structured by considerations of social status, superiority and inferiority yet health outcomes are relatively favourable. If the relevant psycho-social mechanism is to do with social status (in any sociologically meaningful use of that term) then measuring status inequality by means of income-inequality puts Japan at the wrong end of the spectrum! This is much more than a petty point about occupational coding, but you wouldn't guess that if all you had to go on was The Spirit Level.

Mills' whole post is worth reading in full here.

Well, by now if you haven't already lost patience and dismissed me as yet another enemy of equality, you are probably muttering that you have dealt with all this before if only I would care to read more of your own work. The thing is, I have read it, and I'm not the only one to notice in it a recurrent pattern. Time after time you tell critics that you have dealt with their objection in one or another of your publications but when I turn to them what I find is indeed a reference to your critics, but not an actual response to the exact criticism they make and often a discussion of some quite unrelated issue. Why you do this is, to me, quite baffling...

So, in the end Professors Wilkinson and Pickett, you face a credibility gap. People like myself who want, broadly speaking, the same things as yourselves can find the time to ferret out, read and consider the evidence you don't tell us about. Joe Public, which I take it The Spirit Level is aimed at, has neither the time nor the access to the primary sources, let alone the training to make an informed judgement. They have to take what you say on trust. That is why university professors speaking with all the lustrous institutional prestige that implies have, in my opinion, a duty to be scrupulously honest, especially when writing for a popular audience. And when the brickbats come they should not be able to get away with emphasizing the popular nature of their writing whilst ducking behind the protective shield of peer review. What we all need are better reasons to believe. I'm not the only social scientist or social democrat who thinks you haven't yet given us nearly enough.

Wednesday, 21 December 2011

Inequality and anxiety

Richard Smith has written an interesting article in Educational Theory which looks at the causal mechanism Wilkinson and Pickett propose in chapter 3 of The Spirit Level to explain the correlations they show in the rest of the book. As Smith writes...

The argument runs as follows. Unequal income leads to unequal status, and in a world where people are alert to and anxious about where they are positioned on the social ladder, this anxiety affects both mental and physical health. Psychological insecurity and distress rise; self-esteem falls.

There is, however, a problem tallying this theory with the empirical evidence.

A highly inconvenient fact for their thesis, and one that they fully acknowledge, is that over the time-scale under consideration self-esteem as well as anxiety ‘‘showed a very clear long-term upward trend. It looked as if, despite the rising anxiety levels, people were also taking a more positive view of themselves over time’’ (SL, 36). Surely, it would seem, anxiety about status should be reflected in lower, not higher, self-esteem.

Wilkinson and Pickett’s solution is to distinguish ‘‘healthy’’ self-esteem from the defensive kind found in those prone to violence, racism, and insensitivity to others. This latter kind is fragile and more akin to ‘‘whistling in the dark’’ (SL, 37); we might compare this analysis with Ruth Cigman’s discussion of ‘psychological fraudsters.’’

In the context this looks like a rather desperate strategy on Wilkinson and Pickett’s part to save the explanation in terms of concern for status and self-esteem.

I agree with Smith's analysis, but I think there is more to be said. Chapter 3 is crucial to everything that follows in The Spirit Level, but, despite being superficially plausible if read casually, what little evidence they present does not support their argument. Indeed, the evidence goes some way to refuting it.

They start the chapter with two graphs showing the rise in anxiety amongst US college students between 1952 and 1993 (based on research by June M. Twenge) (p. 34). These are the only graphs in the chapter and they show anxiety rising continuously since the early 1950s. Whatever else might have been responsible for this trend, it was not income inequality, as Wilkinson and Pickett acknowledge:

We are not suggesting that [these rises in anxiety] were triggered by increased inequality ... the rises in anxiety and depression seem to start well before the increases in inequality which in many countries took place during the last quarter of the twentieth century. (p. 35)

Not only did inequality not rise until much later, but inequality actually fell in the first half of the period. What bearing do these graphs have on their inequality hypothesis? It is not at all clear, but Wilkinson and Pickett promise to explain, saying: "It is important to understand what these rises in anxiety about before their relevance to inequality becomes clear."

Their explanation begins with the discussion moving from anxiety to self-esteem (p. 36). Self-esteem appears to have risen over the period in much the same way as anxiety, but this anxiety, as Smith says, seems incongruous with rising self-esteem. Wilkinson and Pickett square this circle by arguing that this is really quasi-self-esteem, which reflects the way school-children are taught to have excessive faith in themselves, leading to narcissism.  There is, they say, good self-esteem and bad self-esteem, and this is the latter.

Smith has doubts about this argument, but even if we take it at face value, the connection with inequality eludes us. Neither anxiety nor self-reported self-esteem are in any way correlated with changes in the Gini coefficient.

Wilkinson and Pickett then introduce some causes of psychological stress which include "low social status, lack of friends, and stress in early life." They offer some rather banal observations such as "friends make you feel appreciated" and "how people see you matters." (p. 39) This is all perfectly plausible, but so what? Do people have fewer friends in less equal societies? We are not told, but there is no particular reason to think so.

The next study Wilkinson and Pickett discuss leaves us none the wiser because that, too, does not address income inequality at all (p. 41). They then return to the issue of anxiety and seem to be on the brink of explaining its relevance to inequality.

Why have these social anxieties increased so dramatically over the last half century—as Twenge's studies showing rising levels of anxiety and fragile, narcissistic egos suggest they have? Why does the social evaluative threat seem so great? A plausible explanation is the break-up of the settled communities of the past. (p. 42)

Again, this is more than plausible. The rapid rise in geographical mobility over the last half-century may well offer part of the explanation for the decline in social cohesion that has been documented by Putman and others. People who leave home are less likely to benefit from being around old friends and family. "Familiar faces," they write, "have been replaced by a constant flux of strangers. As a result, who we are, identify with, is endlessly open to question."(p. 42)

A lot of this appeals to common sense and is often based on sound sociological evidence, but still the link with income inequality is nowhere to be seen. It only arrives in the final pages of the chapter, and even then, speculatively. They repeat their admission that there is no correlation between rates of anxiety and rates of inequality...

Although the rises in anxiety that seem to centre on social evaluation pre-date the rise in inequality...

But they make the association all the same... is not difficult to see how rising inequality and social status differences may impact on them. (p. 43)

This is no more than a hunch and no evidence is presented in its favour. In fact, is is difficult to see how inequality impacts upon "the rise in anxiety", because none of the variables they associate with anxiety seem to be linked with inequality, and "the rise in anxiety" itself is demonstrably not linked with inequality in the time-series graphs that kicked off the chapter. Nothing they say in the intervening pages explains why anxiety rose when inequality was falling and continued to rise (at the same rate) when inequality went up.

Further hunches follow:

Greater inequality seems to heighten people's social evaluation anxieties by increasing the importance of social status. (p. 43)

No evidence is presented for this assertion (the only references given are a quote from a nineteenth century philosopher and a study which shows that people are judged on first impressions—neither mentions inequality).

Greater inequality is likely to be accompanied by increased status competition. It is not simply that where the stakes are higher each of us worries more about where he or she comes. It is also that we are likely to pay more attention to social status in how we assess each other. (p. 44) (my italics)

They may believe so, but what they think "likely" can only be viewed as conjecture unless it is backed up with evidence. This, they fail to provide. Instead, they discuss the "stark contrast between the way people see and present themselves" in Japan and the USA. The Japanese, we are told, are more modest while Americans are "more likely to attribute individual successes to their own abilities." This leads Wilkinson and Pickett to jump to the heroic conclusion that:

As greater inequality increases status competition and social evaluative threat, egos have to be propped up by self-promoting and self-enhancing strategies.

... Not only do large inequalities produce all the problems associated with social differences and the divisive class prejudices which go with them, but, as later chapters show, it also weakens community life, reduces trust, and increases violence. (p. 45)

If the comparison between Japan and the USA is supposed to seal the deal with the reader, it falls flat. As Saunders and others have noted, it would be hard to find a more hierarchical and status-conscious society than Japan. The modesty Wilkinson and Pickett attribute to the Japanese applies to many Asian societies. Furthermore, Japan's high suicide rate is not indicative of a low-anxiety nation.

By contrasting Japan and America, Wilkinson and Pickett are comparing apples with oranges. They are very different countries, to be sure, but the difference is rooted in culture, not income inequality. Wilkinson and Pickett say that by focusing on these two countries, they can contrast "the most equal with almost the most unequal of the rich market democracies". "Almost" is the key word here. The most unequal country in their book is actually Singapore, but if they had compared Singapore with Japan, they would not have achieved the desired effect.

So, to recap, Wilkinson and Pickett's theory goes as follows: Those of us who live in rich societies are "much more anxious than we used to be" (p. 33) and this rise in anxiety has been mirrored by a rise in self-esteem. This self-esteem, however, is not real self-esteem but narcissism driven by stress. This stress is driven by feelings of inferiority and the "social evaluative threat" which are "likely" to be driven by inequality.

There are many problems with this torturous chain of reasoning, not least Wilkinson and Pickett's tendency to oversimplify complex ideas and treat contentious theories as fact, but leaving them aside, let us take the proposition that stress and anxiety are the root causes of many modern social and health problems. If inequality is, in turn, the root cause of this anxiety, there must be a correlation between inequality and anxiety over time, but no amount of sophistry can disguise the fact that no such relationship exists.

Chapter 3 involves so many digressions and non-sequiturs that the unwary reader can be forgiven for having forgotten this little fact by the time they reach the end, and yet those two graphs indicating no correlation between anxiety and inequality seriously undermine their hypothesis.

But those graphs tell us something more. Wilkinson and Pickett are forced to accept that the rise in anxiety between 1950 and 1980 cannot be blamed on inequality because inequality was flat or falling during this period, but they never ask the obvious question—if inequality was not the cause, what was?

If the anxiety data is to be believed—and Wilkinson and Pickett cast no doubt on it—something important clearly happened after 1952 (at the latest) which led to greater anxiety, but Wilkinson and Pickett display not a hint of curiosity as to what this might be. Had they not been so committed to their a priori hypothesis, they might have paid more attention to this mysterious third variable. As it is, they ignore it for the remainder of the book in favour of their hunch that inequality is somehow the culprit despite their own evidence.

Sunday, 4 December 2011

Wilkinson in The Guardian

Last week, Richard Wilkinson wrote an article for The Guardian in which he claimed that a large reduction in inequality took place in the 1930s which led to a large rise in life expectancy, despite economic hardship.

Rather surprisingly, health – and probably other indicators of wellbeing – continued to improve in the great depression of the 1930s. This is likely to have been partly because that period saw the most rapid sustained increase in equality on record.

There are two problems with this assertion. The first is that the Great Depression of the 1930s did not see "the most sustained increase in equality on record." The big fall in inequality began in 1939-40 as a result of the Second World War and continued through the austerity years of the 1940s. Rationing, conscription and full employment are the most plausible explanations for this decline. 

As the graph below shows, this was an international phenomenon known as the Great Compression. A similar graph appears in the revised edition of The Spirit Level (p. 296) clearly showing little change in British inequality in the 1930s. (Both graphs show the share of income held by the top 1% of earners. Other measures tell the same story.) It is puzzling that Wilkinson, whose bachelor's degree was in economic history, can confuse the Great Depression with the Great Compression.

The second problem is that Wilkinson's only piece of empirical evidence linking reduced inequality with better health during this period comes from life expectancy data (he offers no evidence at all to support his hunch that "probably other indicators of wellbeing" also improved). Life expectancy is indeed a good proxy for health, but there was nothing remarkable about the improvements seen in the 1930s. As the graph below shows, life expectancy increased throughout the century at a steady rate. There is no historical correlation between inequality and health (nor with any of the other criteria studied in The Spirit Level)

Life expectancy neither rose more sharply when inequality was low, nor declined when inequality was high. Because of the sustained rise in life expectancy, it is possible to point to any event in the 20th century (except for the two world wars) and say that it coincided with a rise in life expectancy. It would be fatuous, but it would be factually correct. But if you were to say that x caused a rise in life expectancy—as Wilkinson is doing here—you would rightly be accused of mistaking correlation for causation and committing a basic post hoc ergo propter hoc error.

Wilkinson is so committed to the theory that inequality is the main driver of social outcomes that he is compelled to view any period of history which saw a decline in inequality as a time of national revival, regardless of what the history books say. In The Spirit Level, he and Pickett display nostalgia for wartime Britain, austerity Britain and the 1970s, despite these being notoriously miserable times to be alive. It is appropriate that the grim 1930s be added to that list.

The main aim of Wilkinson's article is to dismiss the academic field of happiness studies. More of that in January when the Institute of Economic Affairs publishes a monograph on that subject...

Wednesday, 25 May 2011

Does the Better Life Index support The Spirit Level?

Yesterday, the Organisation for Economic Co-operation and Development (OECD) published the Better Life Index. This project aims to measure the quality-of-life in countries using eleven criteria.

Since it was founded in 1961, the OECD has helped governments design better policies for better lives for their citizens. More recently, the OECD has been keenly involved in the debate on measuring well-being. Based on this experience, these 11 topics reflect what the OECD has identified as essential to well-being in terms of material living conditions (housing, income, jobs) and quality of life (community, education, environment, governance, health, life satisfaction, safety and work-life balance).

Having gone to all this trouble to create a reliable index of well-being, it's worth asking whether the OECD data supports The Spirit Level's hypothesis that "more equal societies almost always do better"?

No, it doesn't. The poorest country—Portugal—does worst, but the data points appear to be scattered randomly.

In fact, if we drill down into the data we can see that not only is well-being not better in "more equal" countries, but the OECD's figures do not support The Spirit Level's key argument—that "more equal" countries have stronger social support networks which lead to a healthier and happier population.

Quality of social support network:

Life expectancy:

Self-reported health:

Life satisfaction:

There is no association between inequality and any of these variables. Indeed, with the exception of self-reported health—which seems to show the opposite of what The Spirit Level says—the regression lines are about as straight as you could expect from a randomly assorted set of data.

Taken together, the OECD's Better Life Index appears to support the view of a previous effort by The Economist to devise a quality-of-life index, which concluded:

There is no evidence for an explanation sometimes proffered for the apparent paradox of increasing incomes and stagnant life-satisfaction scores: the idea that an increase in someone’s income causes envy and reduces the welfare and satisfaction of others. In our estimates, the level of income inequality had no impact on levels of life satisfaction. Life satisfaction is primarily determined by absolute, rather than relative, status (related to states of mind and aspirations).

(Note on data: The first graph shows an index of all the OECD's criteria with each given equal weight. Singapore and Hong Kong are not OECD members so they do not feature in the list. Their absence can only benefit The Spirit Level's case here because these unequal countries tend to perform well under most criteria. As per The Spirit Level, tax havens and countries without equality data are excluded. All other counties with wealth greater than Portugal are included. The OECD produces well-respected inequality figures but these are not used in The Spirit Level (perhaps because they show Japan to be quite unequal). I have used The Spirit Level's preferred inequality figures here and throughout The Spirit Level Delusion.)

Tuesday, 1 February 2011

20 questions for Richard Wilkinson & Kate Pickett

[Wilkinson and Pickett (W & P) have now responded to these questions. This post has been updated to show their answers, with my replies below. These are a set of straight-forward questions on simple points of fact that were raised on this blog in April 2010. Some websites have portrayed W & P's replies as being a response to the evidence and arguments in the book The Spirit Level Delusion. They're not. They have never responded to those, but their replies here are of interest all the same.]

1. Why do you exclude the Czech Republic, South Korea and Hong Kong from your analysis when all these societies are wealthier than Portugal?

Wilkinson & Pickett: There are different ways of measuring average income in different countries; the choice of measure makes small differences in precise ranking of countries by wealth. We chose countries ordered according to the Atlas Method, because this is used by the World Bank to classify countries into Low, Medium and High Income categories. Our source is the World Development Indicators Database, World Bank, April 2004*. From this list we selected the 50 richest countries, excluded those with populations less than 3 million and those without income inequality data from the United Nations. Our aim was to examine the impact of inequality on health and social problems among rich countries, where average levels of income are not related to health, happiness or well-being. Our selection criteria also mean that we only consider the older, rich, developed, market economies, and so allows us to compare like with like. The countries which our critics suggest we should fail to meet the criteria.

This is a key question and unfortunately W & P only reiterate what they said in the Appendix of The Spirit Level. I have never challenged their assertion that only wealthy Western industrialised nations should be included in the analysis. The question is which countries are "rich". Any cut-off point will be open to debate, but picking the top 50 for no other reason than that it is a round number is unnecessarily arbitrary.

W & P say that they selected the countries by picking those which appeared "on the flat part of the curve at the top right in Figure 1.1 on p. 7, where life expectancy is no longer related to differences in Gross National Income" (The Spirit Level, p. 280). This is simply not true. The image below is a close up of the countries in Figure 1.1. Hungary, the Czech Republic, Slovenia and South Korea (Hong Kong is not shown) are all on "the flat part" and are all shown as being richer than Portugal and yet they are not included.

All I do in The Spirit Level Delusion is include countries of similar or greater wealth that should—by W & P's own criteria—have been included in the first place. I have not added poorer countries in, rather I have included rich developed societies that W & P excluded without good reason.

The World Bank gives figures for the Atlas method (in US dollars) and Purchasing Power Parity (in international dollars). The latter is more appropriate for international comparisons—the whole point of international dollars is to facilitate international comparisons. Even using W & P's arbitrary top 50 cut-off point, the World Bank's figures for 2005 show:

(9) Hong Kong: $34,900
(45) Slovenia: $22,140
(46) South Korea: $22,010
(48) Portugal: $22,070
(50) Czech Republic: $19,560

And the figures for 2008 show:

(16) Hong Kong: $43,960
(44) South Korea: $28,120
(46) Slovenia: $26,910
(55) Czech Republic: $22,790
(59) Portugal: $22,080

Whichever data set is used, there is no justification for including relatively poor countries like Greece and Portugal while excluding several nations which are manifestly comparable.

If we wish to "compare like with like", it is particularly important to include Asian countries other than Japan, because they demonstrate that the supposed benefits of Japanese equality are actually the result of social and cultural factors unique to Asian societies. Although very unequal, Hong Kong and Singapore perform well under almost every criteria although the W & P hypothesis predicts that they should perform worst. This reveals a puzzling lack of a dose-response relationship.

[*Although published in 2004, the figures refer to 2002 - CJS]

2. Why do you exclude Singapore from your graph of mental illness when you included it in the same graph when it was published in Olivers James' Affluenza?

Wilkinson & Pickett: Comparing the prevalence of mental illness in different societies has long been thought to be problematic because of cultural differences in labelling mental illness or in help-seeking behaviours. To overcome these limitations, the World Health Organization established a consortium to provide international comparisons of the prevalence of mental illness. As referenced in The Spirit Level, we use these WHO estimates for Belgium, France, Germany, Italy, the Netherlands, Spain and the USA. We added in estimates from Canada, the UK and Australia because they used almost exactly comparable methods (diagnostic interviews of random samples of the population) to the WHO studies. We did not include a survey of mental illness from Singapore in either of our peer-reviewed publications on this topic, or in The Spirit Level, because the WHO surveys included questions on illegal drug abuse and, in 1988, the death sentence became mandatory in Singapore for manufacturing, importing, exporting or trafficking drugs in small quantities. Possession of small quantities was taken as prima facie evidence of trafficking. We therefore consider that self-reported estimates of mental illness in Singapore survey will be under-estimates. However, even if Singapore is included, there is still a statistically significant association between income inequality and mental illness (r=0.58, p=0.04).

This doesn't answer the question of why they included Singapore in the graph they created for Affluenza and The Selfish Capitalist. Drugs are illegal in all the countries studied in their graph and admitting to having used a substance in a confidential mental health questionnaire is not grounds for prosecution for possession, regardless of the severity of the punishment.

Rates of anxiety and mood disorders found in the Singapore study indicate that the position of Singapore in the Affluenza graph is about right (ie. about the same as France).

We will come back to the reliability of their mental illness graph in question 14.

3. Why do you say that the USA’s decline in homicide ended in 2005 when 2008 saw the lowest number of homicides since 1965? As you must know, America's murder rate has halved in the last two decades despite rising inequality.

Wilkinson & Pickett: We started writing The Spirit Level in January 2007 and delivered it to our publisher in February 2008, so clearly we could not have accessed homicide data from 2008 – typically official statistics are published 2-3 years post-collection. At time of writing (mid-2010), the most up-to-date data are for 2008.

The homicide rate in the USA has indeed declined, on average over the past two decades, whilst income inequality has been rising. But, as we discuss in The Spirit Level, and show here, there is a match over time between bottom-sensitive measures of income inequality and changes in homicide rates.

The Spirit Level contains references to several articles published after February 2008, but we shall let that pass. On this occasion, I accept that my criticism was too hasty. I accept that W & P would not have had the homicide data for 2008 before The Spirit Level was published in hardback. Presumably they will correct their claim that the murder rate started to rise again in 2005 in the future editions?*

The substantive point remains, however. America's murder rate is at its lowest level since 1965 and it was very premature for W & P to suggest that the 2005-06 blip represented the start of an upward trend. This seems a small point but it important to their overall argument. The reason W & P are so keen to claim that the murder rate began rising in the 2000s is that the huge decline in American homicide coincided with growing equality. The sharp but inconvenient decline in the murder rate (and of crime generally) from the early 1990s does not sit easily with their theory that inequality is the root cause of homicide. As shown in The Spirit Level Delusion and elsewhere, W & P's evidence for linking the two is extremely tenuous.

* (UPDATE 2011: The second edition has appeared and they have not.)

4. Why did you use older data for your life expectancy/inequality graph than you used elsewhere in The Spirit Level? Is it because more recent data show no correlation with inequality?

Wilkinson & Pickett: To avoid the effects of random fluctuations in inequality measures in each country, we took the average of inequality measures published in four consecutive years of the UN Human Development Report. We then matched outcome data (including life expectancy) as nearly as possible to the same time frame as the measures of inequality. When looking at life expectancy against National Income per head we again took the most up to date measures of those covering the same time frame. There are also many recent studies that demonstrate a relationship between income inequality and health, see for example the study of more than 60 million individuals by Kondo and colleagues.

This simply doesn't make any sense. W & P's national income data come from 2002, therefore it is logical to use life expectancy data from 2002. Instead they use data from 2004 (as published in the UN Human Development Report (UN HDR) 2006).

But when they look at inequality, they use life expectancy figures from 2002 (as published in the UN HDR 2004), even though their inequality figures are averaged out from several different years. A better explanation is that the older edition of the report was preferred because it provided slightly better evidence of a correlation between inequality and longevity, albeit only when W & P's selection of countries were used.

I am by no means the first person to have spotted Richard Wilkinson's tendency to be choosy about which sets of data he uses to make his case:

"He [Wilkinson] gives no satisfactory explanation about why "the poorest 70%" should be chosen, and the suspicion must be that the choise is derived from the data."

Ken Judge, 'Income distribution and life expectancy: a critical appraisal', British Medical Journal, 1995

"The strength of association between absolute income and life expectancy seems quite sensitive to which countries are included. What constitutes an appropriate set of countries of this sort is certainly open to debate. Nevertheless, we have shown that compared with Wilkinson's selection of 23 countries, the addition of the other 10 equally wealthy nations that constitute the full sample, significantly changes the results."

J. Lynch, P. Due, C. Muntaner & G. Davey Smith, 'Social capital - Is it a good investment strategy for public health?', Journal of Epidemiological and Community Health, 2000

"Although many aspects of this debate are still unresolved, it has recently become clear that the findings of that paper [Wilkinson, BMJ, 1992] were an artifact of the selection of countries... the evidence for a correlation between income inequality and the health of the population is slowly dissipating’

Johan Mackenbach, 'Income inequality and population health', British Medical Journal, 2002

W & P also mention the study by Kondo et al., published in the BMJ in 2009. They have a particular fondness for this study, which measured self-reported health. They cited it during the Royal Society of Arts debate and referenced it in their response to Peter Saunders. They also mentioned it in their response to an article I co-authored in the Wall Street Journal. In the latter, they insisted that it "shows unequivocally that inequality is related to significantly higher mortality rates."

In fact, the study concludes that:

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 findings need to be interpreted with caution given the heterogeneity between studies

This paper certainly gives more support to W & P's hypothesis than most, but it is anything but unequivocal.

5. You use the high rate of teen births in Portugal (in 2002) as proof that inequality is related to teen births. Why do you not mention that abortion was illegal in Portugal until 2007?

Wilkinson & Pickett: We do indeed show that teenage births are related to income inequality in rich countries, as have UNICEF. This is not dependent on Portugal; indeed if we exclude Portugal the relationship with inequality is slightly stronger, not weaker. In the USA, unlike internationally, data on teenage conceptions were available so we use those, rather than births, as they are unaffected by state differences in access to abortion, and we show the same robust relationship with inequality.

The relationship is not stronger without Portugal. Nor could it be, since Portugal is an unequal country with a high rate of teen birth (primarily the result of its high rate of teen marriages and low abortion rate). With Portugal dropped from the analysis, the r-squared drops from an already feeble 0.10 to 0.089.

I do not suggest that the correlation is dependent on Portugal, only that a ban on abortion is going to have a major effect on the birth rate and is an example of the kind of real-world differences that are routinely ignored in The Spirit Level. Throughout The Spirit Level, W & P refuse to acknowledge social, demographic, historical and cultural differences between nations unless those differences can be used to plug gaps in the theory.

For example, when Finland is seen to have a high rate of homicide, they are quick to point out that it has a high level of gun ownership, but, absurdly, they do not give the same benefit of the doubt to (unequal) places like Israel and the United States.

In the case of Portugal, two major, common sense explanations for the high level of teen births are staring them in the face (the other being Portugal's high rate of teen marriages) but since the high Portuguese figure superficially supports their hypothesis, these confounding factors are never acknowledged.

As I acknowledge in The Spirit Level Delusion, the UNICEF report mentions inequality as one possible explanation for teen births but it makes it clear that it is not the primary cause. As Peter Saunders has shown, English speaking countries tend to have higher teen births. The UNICEF document makes it clear that cultural differences provide a better explanation for teen birth rates than inequality. The low rates of teen births in Singapore and Hong Kong (not shown in The Spirit Level) give further support for this, as does the fact that the only non-Anglo nation with a higher rate of teen births is Portugal for the reasons given above.

6. Why do you not include the crime rate in your index of health and social problems? Is it because the crime rate tends to be higher in 'more equal' countries?

Wilkinson & Pickett: It has often been pointed out that homicides are one of the few crimes which can be compared reliably between countries. Comparisons of other kinds of crime are affected by differences in the law, in reporting, and by other extraneous influences. Car crime, for instance, is affected by the number of cars  and rape is dramatically affected by reporting (see our answer to Q 19 below. While there are some research papers showing relationships between inequality and property crime, there are no sources of data (including those used by Snowdon) which deal adequately with these problems. Hence, we confined our attention to adult and juvenile homicide rates. There are more than 50 studies showing that inequality is related to violence, see for example the review by Hsieh and Pugh 25 and the recent study by Elgar and Aitken.

I mention the potential pitfalls of comparing recorded crime data in The Spirit Level Delusion, which is why I also show the results of the EU Crime Survey. The Crime Survey uses the same methodology for all countries and is based on people's experience of crime, whether recorded or not. Both data sets indicate higher levels of crime—particularly property crime—in more equal societies.

7. Why do you say that homicide is inversely related to suicide when there is no evidence for this?

Wilkinson & Pickett: In fact, there are several pieces of research which show that homicide rates are inversely related to suicide, see for example (44) (45)

Having spent so long fact-checking The Spirit Level, it came as no great surprise when I turned to the first study referenced by W & P and found that its conclusion read:

The overall correlation between homicide and suicide rates was weak and statistically insignificant

It is certainly true that homicide rates are higher in Africa and the Americas (this is confirmed in the second study they reference—see below), while suicide rates are higher in Europe.

Perhaps W & P would argue that this graph shows that suicide is inversely associated with homicide. I think most people would say that it shows that there is simply a much higher homicide rate in Africa and the Americas. There is reason to believe that rising prosperity leads to lower homicide rates and higher suicide rates. But when we compare the wealthy countries that are the focus of The Spirit Level, there is no correlation between suicide and homicide (see The Spirit Level Delusion, p. 82).

8. Why do you suggest that people in more equal countries give more to charity when the reverse is true?

Wilkinson & Pickett: We do not say that people in more equal countries give more to charity - instead we show that more equal countries donate more in development aid to foreign countries. We do cite Eric Uslaner’s work which shows that people who have high levels of trust are more charitable. Snowdon presents data from the Charities Aid Foundation, which suggests that more unequal countries (especially the USA) have higher levels of individual charitable giving. However, as the Charities Aid Foundation points out, charitable giving in the USA is heavily influenced by tax policy, and may also be a response to the exceptional need created by the US lack of social security systems. Only 3% of US charitable giving goes overseas, so total US donations to overseas development are substantially lower than other rich countries.

Low levels of US government aid are partly a reflection of low trust in government (strongly related to inequality) and also of a lack of social security and welfare provision. Together these shift the onus of support to wholly inadequate private charitable giving.

On page 60 of The Spirit Level, Wilkinson and Pickett write that:

"Not surprisingly, just as individuals who trust other people are more likely to give to charity, more equal countries are also more generous to poorer countries."

Bearing in mind that, at this point in the book, W & P have just 'proved' that people in egalitarian societies are more trusting of other people, how else is the reader supposed to interpret this sentence, other than as a claim that people in more equal countries give more to charity?

But they don't. They give less to charity, for reasons that are well-understood. The Charities Aid Foundation (CAF) certainly does point out that the USA's tax system encourages donations to charity. Are Wilkinson and Pickett suggesting that this is a bad thing? It seems that way. (nb. Only Sweden, Finland and Austria do not incentivise charitable giving).

The CAF also notes that people in high-tax countries are less likely to give to charity. This is hardly surprising, since people are less likely to give to charity if (a) they have less disposable income, and (b) they live in a society which expects the state to take responsibility for such matters. I address both issues in my book and quote the CAF.

Wilkinson and Pickett reveal their political bias with their dismissive attitude towards voluntary, private charity and their preference for publicly funded, compulsory aid. They are fundamentally wrong about all their assumptions given above. Low levels of government aid are not a response to lack of trust in government—the government decides the level, after all. Instead, they are the result of lower levels of taxation and a historical emphasis on private philanthropy. W & P are also wrong to say that total US aid (private + public) is "substantially lower" than that of other countries. The USA is consistently in the top 10 in both per capita terms and in absolute terms.

And they are wildly wrong when they make the unreferenced claim that only 3% of US donations goes overseas. This is just the kind of causal disregard for the truth that makes fact-checking The Spirit Level such an exhausting affair. Of the $307 billion the US gave to charity privately in 2008, $116 billion went overseas. That is 38%, not 3%. And while Americans prefer private donations to state-funded aid, the USA gives far more than 0.7% of its GDP to the developing world. (0.7% being the United Nations target.)

9. Why did Kate make a video called ‘Why Cubans live longer than Americans?’ when all the sources show that life expectancy in Cuba is lower than in the USA?

Wilkinson & Pickett: Kate was not consulted about the title for this online clip from a short interview. What she actually said was that countries such as Cuba, Costa Rica and some poorer European countries have life expectancy as high, or higher than, the USA. In fact, in the 2006 revision of the United Nations World Population Prospects report, for 2005-2010, infant mortality rates in Cuba were 5.1 per 1000 live births, compared to 6.3 for the United States, and life expectancy was in Cuba was 78.3 years, compared to 78.2 years in the USA.

It is difficult to believe that Pickett would have no say about the title of a video she presented, or would have no power to rename or withdraw the video once it was given a misleading title. But whoever named it can hardly be blamed, since she does indeed claim that Cuba and Costa Rica have life expectancies that are "as high or higher than the United States."

W & P's reference to the World Population Prospects report is another example of their habit of data-mining to find the 'right' figures. As they must know, that particular report gives a prediction, it is not a real figure. More reliable figures for life expectancy come from the UN HDR—the preferred source used in The Spirit Level. All of these reports show life expectancy being higher in the USA than in Cuba and Costa Rica. The most recent edition shows that life expectancy is 79.1 in the USA and 78.5 in Cuba. Costa Rica's life expectancy is 78.7.

10. Why do you write about "increased family break-down and family stress in less equal countries" when divorce and single-parent households tend to be more common in more equal countries?

Wilkinson & Pickett: Although lone parent families are not more common in more unequal countries, changes in income inequality are correlated with rising divorce rates in US counties.

This evades the question. The facts show that divorce is more common in more equal countries, and the Scandinavian countries have amongst the highest rates of lone parent families (as do some Anglo-Saxon countries). It is simply untrue to say that there is "increased family break-down and family stress in less equal countries".

11. Why do you say that community life is weaker in less equal countries when these nations have more people involved in community organisations (charities, sports clubs, environmental groups etc.)?

Wilkinson & Pickett: Robert Putnam’s measures of ‘Social Capital’ are based on membership of voluntary and community associations of the kind you mention. Both in his earlier study of the Italian regions and in his study of the American states he shows there is a very strong tendency for the more equal regions and states to have stronger community ties measured in this way. Looking at changes over time in the US as a whole he also says:

"Community and equality are mutually reinforcing... Social capital and economic inequality moved in tandem through most of the twentieth century. In terms of the distribution of wealth and income, America in the 1950s and 1960s was more egalitarian than it had been in more than a century. ...those same decades were also the high point of social connectedness and civic engagement. Record highs in equality and social capital coincided. Conversely, the last third of the twentieth century was a time of growing inequality and eroding social capital. By the end of the twentieth century the gap between rich and poor in the US had been increasing for nearly three decades, the longest sustained increase in inequality for at least a century. The timing of the two trends is striking: somewhere around 1965-70 America reversed course and started becoming both less just economically and less well connected socially and politically." p.359

Sociologists distinguish between generalized trust (trust of people with whom we do not have an intimate relationships) and particularized trust (trust of people like ourselves). Generalized trust is related to social capital, and many researchers, including Putnam, have linked these measures of social capital to greater equality. Indeed, they have shown that it is inequality that affects trust, rather than the other way round.

Again, this does not answer the question. The international evidence shows that people in less equal countries are more likely to be members of clubs, societies, sports associations, religious groups etc. The only exception is trade union membership, which is often compulsory or strongly 'encouraged'. The graph below shows data from the World Values Survey.

What Putnam shows is that membership of these kinds of groups has declined in the United States over the last fifty years (as it has elsewhere). But Putnam never endorses the theory that inequality is the cause of the decline in club membership or social capital.

Indeed, W & P's selective quoting from Putnam's Bowling Alone is a good example of how they provide the illusion of an academic consensus where none exists. In The Spirit Level, they quote the same passage as the one quoted above, with the same sections deleted. The unwary reader might interpret this quote as evidence that Putnam believes that inequality leads to less social capital.

But, as ever with W & P, it is what they do not show that is most instructive. The unabridged first line of this quote reads: "Community and equality are mutually reinforcing, not mutually incompatible." After the last line W & P quote, Putnam concludes: "This pair of trends illustrates that fraternity and equality are complementary, not warring values."

As these words suggest, Putnam is not arguing that income equality necessarily increases social capital, he is merely arguing that the former does not damage the latter. This is obvious from the section heading which asks the question: 'Is social capital at war with equality?'

Why would he ask that question? Because, as Putnam points out: "the abundant social capital of the 1950s was often exclusionary along racial and gender and class lines. Generally speaking the haves engage in much more civic activity than the have-nots." Because of this, it is sometimes argued that left-wing attacks on elitist social institutions weaken some forms of community life. Putnam argues that these fears are ill-founded, and he says the positive correlation between equality and social capital in the USA since the Second World War "powerfully contradicts the view that community engagement must necessarily amplify inequality."

But Putnam is far more cautious in claiming a causal link between income equality and social capital than W & P. In fact, this brief passage is the only part of his 500 page book that evens mentions income equality (R. Putnam, Bowling Alone, 2000; pp. 358-60). When he sums up the evidence to explain why community life has broken down in the USA since the 1950s, he identifies four factors ('generational change', 'television', 'work' and 'sprawl' - p. 284). He does not include inequality.

12. Do you accept that the World Values Survey data show no correlation between 'happiness' and inequality, but a strong correlation between 'happiness' and income?

[There are so many points to make about W & P's answer here that I have split it up]

Wilkinson & Pickett: We accept that there is no relation between inequality and WVS measures of happiness, but among the rich countries neither is there a relation between happiness and Gross National Income per head (see our figure 1.2 in The Spirit Level).

W & P's claim that there is no relationship between happiness (as measured by the World Values Survey) and Gross National Income is incorrect. As can be seen from the graph below, there is a clear correlation, even at a very high level of economic development.

W & P never show this graph. Instead, they use a graph that shows rich and poor countries together (Spirit Level, p. 9), which obscures what is shown above. Consequently, they are able to begin their book by claiming that "happiness levels fail to rise further as rich countries get still richer". This is a crucial part of their argument, since it suggests that economic growth has "done its work"and can therefore be curtailed. It is, however, factually incorrect.

The other important thing about the happiness surveys is that they show no relationship with inequality. This is a crucial observation, since it is powerful evidence that more equal societies are not happier.

During the Royal Society of Arts debate, Richard Wilkinson responded to this evidence by claiming that happiness does not have a social gradient and, therefore, would not be affected by levels of inequality. This is not true. Any number of studies have shown that the rich are happier than the poor eg. Easterlin (1974). They have, however, now acknowledged this...

In our debate at the RSA, Richard meant to say that happiness and income have a reverse social gradient, rather than no social gradient. The correlation between income and happiness among individuals within countries has been shown to be a relationship with relative income and social status. 

The correction is welcome* but the subsequent argument is weak. There are many reasons why the rich are happier than the poor. In part, their absolute income allows them to afford a better standard of living and, in part, their relative income allows them to compare themselves favourably to their peers. While both are important, the material benefits of absolute income receive short shrift in The Spirit Level. W & P focus only on relative income and claim that richer people are happier only because money buys status and status symbols. This is opinion, not fact.

It has also been shown that additional income makes much more difference to the happiness of the poor than the rich. This would suggest that redistribution would improve over-all happiness. 

No. It suggests that additional income would improve the happiness of the poor. We should, therefore, pursue a system that will maximise increases in the incomes of the poor. There is no evidence that the zero-growth egalitarianism proposed by W & P will do that.

But the key argument in The Spirit Level is that the poor would become happier if the rich become poorer, even if the poor got no richer in the process. In other words, rounding up and deporting every millionaire in Britain would, in itself, be enough to improve social outcomes, even if no one got richer as a result. This is where W & P depart from the bulk of the scientific literature and, I would argue, from common sense.

Several economists who study happiness (e.g. Blanchflower and Oswald) show that, in sub-national analyses, more equal societies, for example more equal US states, are happier. 

The majority of happiness studies have shown that inequality has no effect on happiness, which is why the happiness studies are not mentioned in The Spirit Level. (See, for example, reviews by Clark and Senik (2010) and Hopkins (2008).) Nor do Blanchflower and Oswald say that more equal states or countries are happier. This is a complete misrepresentation of the literature.

International comparisons of subjective variables, such as happiness, are notoriously unreliable (for example, self-reported health appears better in countries with higher death rates) This is why in The Spirit Level we concentrated very largely on objective measures of health and wellbeing.

Since the happiness surveys show no correlation with inequality and a strong correlation with income, it is perhaps unsurprising that W & P now choose to distance themselves from them. It is a shame they felt the need to put this "notoriously unreliable" data at the heart of the first chapter of their book. Note also, that the trust survey—upon which they base so many of their arguments—is a wholly subjective and self-reported measure. They also put great faith in the Kondo study (see question 4) which studies self-reported health.

[*A "reverse social gradient" implies a reverse relationship ie. that the poor are happier. Since even W & P cannot believe this (surely?), I assume this to be a slip of the pen.]

13. On page 19 of The Spirit Level, you say you included alcohol addiction as a 'health and social problem', but you never discuss it in the rest of the book. Is this because the highest rates of alcoholism are in Scandinavia?

Wilkinson & Pickett: It is important to distinguish between alcohol use and alcohol abuse. Alcohol use is difficult to measure and often has no social gradient – consumption tends to be higher in higher social classes. This is in marked contrast to binge and problem drinking. We include alcohol abuse (as measured by surveys of mental illness that cover drug and alcohol addiction) in our Index of Health and Social Problems, and have previously demonstrated a significant relationship between deaths from alcohol-related liver disease and income inequality in US states.

I agree about making a distinction between per capita alcohol consumption and overall alcohol consumption, but several international surveys of mental illness surveys have shown that the more equal Scandinavian countries have the highest levels of alcohol abuse. It would be interesting to see what W & P's alcohol abuse graph looks like. It does not appear in The Spirit Level and the data is not included in their spreadsheet.

14. Why do you show no data about the (high) prevalence of mental illness in Scandinavia?

Wilkinson & Pickett: The World Health Organization has not yet produced internationally comparable data on mental illness for Scandinavian countries, but we eagerly await such data. In the absence of robust estimates from the WHO, we know of no high quality data to justify the suggestion that Scandinavian countries have a higher prevalence of mental illness.

W & P's mental illness graph provides one of the most egregious examples of data-mining in The Spirit Level. Much more evidence for mental illness prevalence exists than is shown in The Spirit Level and, although W & P are implying otherwise, they do not confine themselves to "robust estimates from the WHO". The figures they use for the UK, Australia, New Zealand and Canada (all less equal countries) come from other sources. When more than one source is available, W & P consistently pick the figure that best suits their hypothesis. As I explain in The Spirit Level Delusion, there are serious questions about the reliability of some of the figures from the WHO study, which the study's authors acknowledge.

I don't claim that rates of mental illness are necessarily "higher" in Scandinavia, but the available evidence suggests that they are as high as in the United States. For example, a large study of all four Nordic states concluded:

Overall, it can be concluded that the prevalence rate of generalized anxiety disorder and major depression in different European countries, including Denmark, Finland, Norway, and Sweden, are as high as those found in the United States. (‘Prevalence of generalized anxiety disorder in general practice in Denmark, Finland, Norway, and Sweden’, P. Munk-Jorgensen, Psychiatric Services, 57 (12), December 2006; p. 1738-44)

Similarly, a study of mental health in Norway compared rates of the most common mental disorders with those in the USA and found "almost identical rates for alcohol abuse, major depression, and social phobia in the two countries." (‘A Norwegian psychiatric epidemiological study’, E. Kringlen et al., American Journal of Psychiatry, 2001, 158; pp. 1091-98)

Incidentally, the largest systematic review of the prevalence of mental disorders in the EU (Wittchen & Jacobi, 2005 - summarised here (p. 47)), found the 12 month average to be 27%—ie. almost identical to estimates from the USA.

15. If equality creates good health, why does Denmark currently have the lowest life expectancy of any country in your list?

Wilkinson & Pickett: As with our other analyses, we (unlike our critics) do not pick and choose different countries to include or exclude according to whether or not their outcomes fit the inequality data. Denmark does indeed have much lower life expectancy than we would expect given its level of inequality. We have never claimed that income inequality is the only cause of worse health and social problems in a society. There will always be countries that do a bit better or worse on any outcome than we might predict given their level of inequality. Some researchers have attributed Denmark’s relatively poor health to its high levels of smoking.

Denmark doesn't have a particularly high level of smoking (see page 24 of SLD) and there is surprisingly little correlation between high smoking rates and low life expectancy when whole nations are compared. This only serves to highlight the problem of looking for the effect of one variable amongst aggregate data.

The simple fact is that the Scandinavian countries all have a very low level of inequality but vary wildly when it comes to life expectancy. Using data from the most recent UN HDR, out of 28 countries Sweden comes 8th, Norway comes 12th, Finland comes 19th and Denmark comes 27th.

Japan comes 1st, and this is the jewel in W & P's crown since Japan has an even lower level of inequality (by their measure, at least). But very unequal Hong Kong comes 2nd. Of course, readers of The Spirit Level will not know that because—although W & P "do not pick and choose different countries to include or exclude according to whether or not their outcomes fit the inequality data"—Hong Kong is never shown.

16. Why were Singapore and Hong Kong excluded from your graph on obesity?

Wilkinson & Pickett: The International Obesity Taskforce did not report data on obesity for Singapore in the 2002 report which was available when we were writing The Spirit Level. Hong Kong is not a nation state but even if it were it does not meet our inclusion criteria (see point 1).

The Singapore figure was published in 2004. The IOTF currently cites this figure so presumably we can expect it to appear in future editions of The Spirit Level.*

Hong Kong's somewhat unusual political status is no reason to exclude it. It is a country and both the World Bank and the United Nations treat it as such and collect data for it as a separate entity. It is certainly a separate and distinct society and The Spirit Level's subtitle is 'Why More Equal Societies Almost Always Do Better'. If US states fall under the category of separate societies, it is surely reasonable to include Hong Kong.

Why wouldn't Hong Kong meet the inclusion criteria? It has a population of over 7 million, it has sound data on inequality and came 16th in W & P's preferred list of rich countries.

* UPDATE 2011: The second edition is out and it is still missing.

17. Do you accept that the "correlation" between trust and equality rests entirely on figures from the four Nordic countries and that there is no pattern amongst the remaining 19 nations?

Wilkinson & Pickett: Absolutely not. These countries are NOT outliers, but lie on the trend line. However, even if they are excluded there is still a statistically significant correlation among the remaining countries (r=-0.46) as well as among US states where the correlation between trust and inequality is also highly significant (r=-0.7).

Again, this depends on which data are used. W & P's trust data come from the 1990s. If you use World Values Survey data from the 2000s, it is clear that the Scandinavian countries are outliers. The correlation disappears when they are excluded (see below). Even if you prefer W & P's older data set, their correlation is extremely weak.

18. Why do you say that young people "defer sexual activity" in more equal countries when there is no evidence for this?

Wilkinson & Pickett: We don’t say that people defer sexual activity in more equal countries – we simply discuss Professor Jay Belsky’s theory about quality versus quantity reproductive strategies which biologists have identified in many species.

W & P certainly do discuss Jay Belsky's theory that—as they put it—"people who grow up learning 'to perceive others as trustworthy, relationships as enduring and mutually rewarding and resources more or less constantly available' would mature later, defer sexual activity, be better at forming long-term relationships and invest more heavily in their children's development."

W & P precede this discussion by saying "there seem to be additional reasons why teenage motherhood is sensitive to degrees of inequality in society". They follow it by saying: "So there may be deep-seated adaptive processes which lead from more stressful and unequal societies." Apparently the reader should not infer from this that women in more equal societies defer sexual activity?!

19. If greater equality makes countries less violent and more law-abiding, why does Sweden have the highest rate of rape and theft of any country in your list? Why does Finland have the highest murder rate in Europe?

Wilkinson & Pickett: As we discuss in The Spirit Level, there are multiple influences on health and social problems, and income inequality is only one factor (albeit a strong and robust factor, demonstrated in more than 50 studies) affecting murder rates. Finland has a higher rate of homicides than we would predict, given its level of inequality, probably because of its high level of gun ownership. If we control for gun ownership in US states, the relationship between inequality and homicides actually gets stronger. For crimes other than homicides, comparing crime data among different countries is problematic, due to reporting differences. It seems sensible to assume that rape is more likely to be reported in societies where women’s status is higher.

It is not true that "more than 50 studies" have found a "strong and robust" association between inequality and homicide. Again, W & P mention a third variable (gun ownership) when they discuss Finland but do not give the same benefit of the doubt to Israel and the USA. In The Spirit Level, they do the same when trying to explain how the most equal state (Alaska) has one of the highest rates of homicide. They also make the questionable assumption that murders in America are carried out with legal firearms.

The assumption that Sweden has the highest rate of recorded rape because women are more likely to report sexual assaults is pure supposition.

20. Since when has the definition of a tax haven been a country with fewer than 3 million inhabitants? Isn't this just an excuse to leave out Slovenia?

Wilkinson & Pickett: The cut off for a small country has to be defined somehow – countries with populations around our 3 million cut-off point include Slovenia, Namibia, Lesotho, and Botswana. Slovenia is the only rich country with close to 3 million inhabitants excluded from our analyses. What happens if we add it in? Not much – the correlation between income inequality and homicides is r=0.42 (p=0.04) with Slovenia in, and r=0.43 (p=0.04) with Slovenia out. For imprisonment, the correlation with Slovenia in is r=0.66 (p<0.001), with Slovenia out, it is r=0.65 (p<0.001)….etc

No, the cut-off for a small country does not have to be defined somehow. A cut-off is completely unnecessary. All we need to know is which countries are tax havens (is Switzerland not a tax haven?). All the 3 million cut-off does is exclude Slovenia.

And, yes, including Slovenia often doesn't make a great deal of difference. That's isn't the point. We don't pick and choose the countries on the basis of whether they will make a difference. We include them so we have the full range of rich societies upon which to base our conclusions. Since The Spirit Level's conclusion is that "it falls to our generation to make one of the biggest transformations in human history" (p. 265), it would prudent to make sure the assumptions it is based on are not plucked out of the air.

It should also be noted that in one of their published studies, Wilkinson and Pickett "included only countries that had income inequality data and were among the richest 50 in the world and excluded those with populations of 50 in the world and excluded those with populations of less than two million to avoid possible tax havens." This allowed them to include Slovenia which, for the criterion studied in that paper, happened to strengthen the desired relationship. The switch from two million in this study to three million in The Spirit Level did not weed out any tax havens and served only to exclude Slovenia.

And finally...

A word about the scientific literature. W & P are fond of claiming that their work is backed up by 100s of peer-reviewed studies. In a recent letter to The Guardian, for example, they once again referred to "hundreds of other academic research papers which show similar patterns."

Do not be fooled. Very few of the studies referenced in The Spirit Level claim that health or social problems are caused by income inequality per se (as opposed to absolute income or other socio-economic factors). Of the few that do make such a case, many were written by Richard Wilkinson and/or Kate Pickett (they refer to no fewer than twelve of their own studies in The Spirit Level).

The bulk of the references are to newspaper articles, opinion pieces, other people's books, studies that discuss specific issues (eg. stress, violence, obesity) and the sources of the raw data (eg. UN, OECD). Often the studies referenced give equivocal support or contradict W & P (see Questions 4 & 7, for examples). The only area which has a significant body of scientific literature is health and inequality, and much of it disagrees with Wilkinson's hypothesis. As Wilkinson admitted in a recent interview with the magazine International Socialism, there is virtually no evidence from other academics to support the bulk of the claims made in The Spirit Level.

Wilkinson: "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."

These are, then, very much W & P's own ideas. Although The Spirit Level has launched their theories into the popular consciousness, they find little support in the scientific literature. It is no surprise, then, that their book has received a much cooler reception from academics (see below) than from ex-politicians and journalists.

"The bottom line is that this is a well-written, stimulating polemic. It nevertheless suffers from the same problems as one-trick ponies: if the one trick does not impress you, the show is a failure. Wilkinson and Pickett’s trick simply does not hold up to empirical scrutiny. When assessing this book as a contribution to the debate on the “right” level of income differences in modern society, it is a highly interesting, sympathetic attempt at addressing some of the important problems of Western societies. Yet, when assessing this book from a scientific point of view, one is forced to conclude that it is a failure."

— Christian Bjornskov, Professor of Economics, University of Aarhus, Population and Development Review, June 2010

"Wilkinson and Pickett have no time for nicely balanced judgements. They believe that the evidence they present shows beyond doubt that more equal societies ‘do better’, and they are also confident that they have the right explanation for why this is so... Their case is by no means so securely established as they try to make out... it has been called into question by other leading figures in the field – a fact that WP might have more fully acknowledged... WP’s inadequate, one-dimensional understanding of social stratification leads to major problems in their account of how the contextual effect is produced."

— John Goldthorpe, Emeritus Professor of Sociology, Nuffield College, Oxford; ‘Analysing social inequality’ European Sociological Review, 2009

"The evidence in The Spirit Level is weak, the analysis is superficial and the theory is unsupported."

— Peter Saunders, Emeritus Professor of Sociology, University of Sussex, Beware False Prophets

"The book will probably irritate most economists, including those like me who are sympathetic to its basic stance... source of irritation is the authors’ apparent belief that the application of regression methods to economic and social statistics is as novel to social science as it apparently is to medicine.  The evidence presented in the book is mostly a series of scatter diagrams with a regression line drawn through them.  If you remove the bold lines from the diagram, the pattern of points mostly looks random, and the data dominated by a few outliers... An obvious conclusion is that there are many societies which perform well in terms of their own criteria. America, Sweden and Japan are just different from each other. Their achievements are not really commensurable. But Wilkinson and Pickett are not content with this relativist position."

— John Kay, former Director of Institute of Fiscal Studies and Professor of Economics at London Business School, Financial Times, 2009


Sanandaji et al. have also written a detailed response to Wilkinson and Pickett's rebuttal of their criticisms. If you are interested in the debate about The Spirit Level (and, if you've got this far, you probably are!), I warmly recommend it. Click here.