Saturday, September 30, 2017

Sapolsky (3): The "Spirit Level" Debate over Inequality and Health

One of the weakest parts of Robert Sapolsky's Behave is his claim that economic inequality necessarily causes bad physical, mental, and social health, even in societies where poverty has been largely abolished.  The problem is that he relies on the research of Richard Wilkinson and a few others who argue for this position without acknowledging that many critics have pointed out flaws in this research, which one can see in the debate over the book coauthored by Wilkinson and Kate Pickett--The Spirit Level: Why Greater Equality Makes Societies Stronger.  This book was first published in 2009, and it became an international bestseller.  It provoked many critical responses, including two books: Christopher Snowdon's The Spirit Level Delusion (2010) and Peter Saunders' Beware False Prophets (2010).  (Saunders' book can be found here.  The Epilogue to the second edition of Snowdon's book can be found here.)  In the paperback edition of their book, Wilkinson and Pickett added a "Postscript," in which they attempted to answer their critics.  A longer version of their reply to critics can be found here.  Remarkably, Sapolsky is completely silent about this debate.

Sapolsky is also silent about the political implications of this debate.  The apparent failure of the Marxist economies and the apparent success of the capitalist economies at the end of the 20th century had had a dispiriting effect on the Left.  Marx's prediction that the impoverishment of the proletariat in capitalist societies must necessarily lead to revolution had proven false, because capitalism was raising the standard of living for all classes, while socialism was failing.  But then Wilkinson and Pickett seemed to show that capitalism was ruining human life by creating high levels of economic inequality, so that while absolute poverty was disappearing, relative poverty was rising: those people living low-status lives felt poor, because they saw that others with higher status had so much more, and the chronic stress from this feeling of relative poverty made people sick.  Moreover, this sickness from inequality created lots of social problems: not only higher rates of disease and reduced life expectancy but also higher rates of crime, mental illness, social distrust, obesity, poor educational performance, teenage births, and high rates of imprisonment.  Capitalist inequality was making everyone desperately unhappy, and the only solution was socialist programs for creating greater equality through redistribution of the wealth and welfare state policies.  Thomas Piketty and others have elaborated this argument about the corrosive effects of capitalist inequality.  This has given new life to the Left.

As I indicated in my previous post, Sapolsky's distinctive contribution to this lefty critique of capitalist inequality is his evolutionary explanation of inequality as the necessary consequence of the move from an egalitarian state of nature for hunter-gatherers to a hierarchical dominance structure in societies based on an agricultural mode of production; and from this he draws the conclusion that it is impossible to restore equality in modern societies that cannot go back to hunter-gatherer life, which means that no socialist policies can ever succeed in overturning inequality and its corrosive effects on human health.

Unlike Sapolsky, I do not see any clear scientific support for Wilkinson's theory of the inequality/health relationship--particularly, the idea that inequality is inherently harmful to human health because of the chronic stress that it creates for low-status people, even in the absence of real poverty.  The flimsiness of the empirical evidence for this theory is evident in the debate over Wilkinson and Pickett's Spirit Level.

Sapolsky's assertion that "numerous studies" support this theory, as if there were a general consensus among researchers about this, is not true (441).  Wilkinson and Pickett say that "there are around 200 papers in peer-reviewed academic journals testing the relationship between income inequality and health in many different settings" (279).  But if you check the footnote for this assertion, you will see that they are citing one of their own papers (Wilkinson and Pickett, "Income Inequality and Population Health: A Review and Explanation of the Evidence," Social Science & Medicine 62 [2006]: 1768-1784).  If you read their paper, you will see that they survey 169 results from 155 studies on inequality and wealth; and of these, they identify 88 as supportive of their theory and 81 as either unsupportive or "mixed" in their results.

One of the best surveys of this research is by Andrew Leigh, Christopher Jencks, and Timothy Smeeding ("Health and Economic Inequality," in W. Salverda, B. Nolan, and T. Smeeding, eds., The Oxford Handbook of Economic Inequality [Oxford: Oxford University Press, 2009], 384-405).  It's available online.  They conclude that "the empirical evidence for such a relationship [between inequality and health] in rich countries is weak.  A few high-quality studies find that inequality is negatively correlated with population health, but the preponderance of evidence suggests that the relationship between income inequality and health is either non-existent or too fragile to show up in a robustly estimated panel specification.  The best cross-national studies now uniformly fail to find a statistically reliable relationship between economic inequality and longevity.  Comparisons of American states yield more equivocal evidence."

The popular appeal of Wilkinson and Pickett's book comes from their graphs that apparently show a statistical correlation between economic inequality and bad health across 23 nations.  (A few of these graphs can be seen online).  For example, here is a graph that seems to show that life expectancy is longer in more equal rich nations.  This graph is easier to see at the online location, where it is figure #17.



This is typical for most of the graphs.  It's a simple linear regression model with the level of income inequality in the nations on the x axis for the explanatory variable and level of health (in this case, life expectancy in years) on the y axis for the dependent variable.  A best-fit line is drawn through the scatter points of data to indicate the trend.  In this case, the declining trend line shows life expectancy declining with rising income inequality.  This downward sloping line depends mostly on the relatively high life expectancy in Japan and Sweden and the relatively low life expectancy in the USA and Portugal.

The data for this graph is from the 2004 United Nations Human Development Report.  A reader who notices this might wonder why they used the 2004 report when the 2005 and 2006 reports were available to them, and actually they do use the 2006 report elsewhere in their book.  They even use the 2006 report for its life expectancy data in another graph (compare pages 7 and 82 in their book).

Christopher Snowdon has shown that if Wilkinson and Pickett had used the 2006 report for their data, the graph would have looked like this:


 
A larger picture of this graph can be found here.

Now the trend line is going up!  If they had used the 2009 report for their data, the trend line would again be going up.  So here increasing income inequality is slightly correlated with increasing life expectancy. 

Now you should notice that Snowdon has added Hong Kong, Slovenia, and the Czech Republic, which were excluded by Wilkinson and Pickett.  Hong Kong shows that a very wealthy but less equal society can have high life expectancy.  Slovenia and the Czech Republic show that more equal but less wealthy societies tend to have low life expectancy.

So the trend line here will slope down or up depending on one's selection of the data points.  Wilkinson and Pickett selected data points that would give them a downward sloping line, while Snowdon selected data points to give him an upward sloping line.  At least Snowdon points this out to his readers, while Wilkinson and Pickett hide this from their readers.

But let's say that we accept Wilkinson and Pickett's graph as showing us a correlation between less equal societies and lower life expectancy.  What exactly does this tell us?  If we remember the common saying that correlation is not causation, then we see that this graph by itself tells us nothing about causation, although Wilkinson and Pickett want their readers to assume that it does show that inequality causes low life expectancy.

Moreover, Wilkinson and Pickett never follow the common practice in the statistics of correlation of testing for alternative explanations that might be confounding variables.  For example, if we compare Japan (the most equal society) and Hong Kong (the least equal society), we would have to notice that despite their great differences in income inequality, they tend to perform about the same not only in life expectancy but also in many other respects.  Is this perhaps explained by their similarity in their Asian culture?  Wilkinson and Pickett never consider this possibility because they never consider any alternative explanation beyond income inequality.

In the Postscript to their book, Wilkinson and Pickett explain their failure to test for alternative explanations: "including factors that are unrelated to inequality, or to any particular problem, would simply create unnecessary 'noise' and be methodologically incorrect" (285).  Since they assume that inequality must be the only explanation, any other possible explanation is to them "unnecessary 'noise'"!

Actually, of course, just glancing at their data might suggest many alternative explanations.  Consider, for example, freedom as measured by the Human Freedom Index, which has been the subject of a post.  Most of the countries with high life expectancy rank in the top 15 of the Human Freedom Index: such as Switzerland (2), Canada (6), Australia (6), and Sweden (15).  The one exception is Japan (32).  Is it possible that greater freedom has something to do with higher life expectancy?

Wilkinson and Pickett point to the Scandinavian countries as setting the standard for how egalitarian societies can promote human health and happiness.  But they ignore the fact that these countries generally score high on both the Economic Freedom Index and the Human Freedom Index (combining economic freedom and personal freedom).  As I have argued in other posts here and here, the Nordic social democracies are not purely socialist, because they are actually capitalist welfare states.

Here's another graph from Wilkinson and Pickett:


This seems to show that homicide rates are higher in more unequal rich countries.  But notice that the upward slope of the line depends entirely on the USA as an outlier.  It is standard statistical practice to throw out outliers to avoid creating spurious correlations.  Wilkinson and Pickett do not do this, because taking out the USA here would create a graph with no correlation between inequality and homicide, which is contrary to the result they want to find.  They say nothing about the high rate of gun ownership in the US as a possible explanation for high homicide rates in the US.  But they do mention gun ownership in their attempt to explain away the high homicide rate for Finland, which is a more equal country, and the low homicide rate for Singapore, which is a less equal country.  "In the United Nations International Study on Firearm Regulation," they observe, "Finland had the highest proportion of households with guns, and Singapore had the lowest rate of gun ownership" (136).

One might notice another problem in these graphs from Wilkinson and Pickett's book.  Inequality is measured by inequality in income.  Is this the best standard?  Is it possible that in countries with extremely high income tax rates--like the Scandinavian countries--people will be motivated to hide their true income or accumulate wealth in forms other than income?  If so, then measuring inequality by income inequality will tend to make countries with high income taxes appear more equal than they really are.

Actually, as Snowdon points out, Wilkinson and Pickett use at least five different measures of inequality in their book, which allows them to change the measures to achieve whatever results they want to find.  For example, on page 239, they compare the incomes of the top 10% and the bottom 10% in the US and UK, which shows that "both countries experienced very dramatic rises in inequality which peaked in the early 1990s and have changed rather little since then."  But on page 296 they want to show that inequality peaked just before the financial crisis of 2008, and to achieve this result, they measure inequality through the share of wealth held by the top 1%.  This is the only place in the book where they use this as the measure of inequality.  They don't use this measure elsewhere in the book, because by this measure Norway and Denmark are less equal than the USA, and so using this as the measure of inequality would not give them the results they're looking for.

Wilkinson and Pickett admit that there is at least one social problem that is more common in more equal countries--suicide.  To explain this, they suggest: "suicide is often inversely related to homicide.  There seems to be something in the psychological cliché that anger sometimes goes in and sometimes goes out: do you blame yourself or others for things that go wrong?  In Chapter 3 we noted the rise in the tendency to blame the outside world--defensive narcissism--and the contrasts between the US and Japan" (175).  So, you see, in the less equal countries, when people are unhappy, they are inclined to kill other people; but in the more equal countries, unhappy people have the decency to kill themselves rather than others!

Wilkinson and Pickett recognize that "social integration" is important for human health: "It's not just social status and psychological wellbeing that affects our health.  The relationships we have with other people matter too. . . . Having friends, being married, belonging to a religious group or other association and having people who will provide support, are all protective of health" (76).  But then they are silent about the possibility that people  with low social status but high social integration might show good health and happiness.

As I have indicated in a previous post, Charles Murray (in Coming Apart) has shown that people in the white underclass in America can be very happy if they are married, if they find satisfaction in their work, if they live in places where neighbors help one another, and if they are active religious believers.  Contrary to what is suggested by Wilkinson, Pickett, and Sapolsky, low social status need not by itself make people unhealthy and unhappy. 

My defense of "good inequality" can be found here, here, and here.

Here is a video of a debate between Wilkinson, Pickett, Saunders, and Snowdon.



No comments: