Monday 17 May 2010

Case study: Life expectancy


One of the central claims of The Spirit Level is that 'less equal' countries have lower life expectancies than more egalitarian countries. Its authors claim that the psychological stress of living in a less egalitarian society affects the health of all—rich and poor—and that this manifests itself in lower life expectancies. 

To demonstrate this they show a graph that looks very much like this:




Inequality is shown on the horizontal axis and is the difference between the richest 20% and the poorest 20% (eg. in Sweden, the richest 20% are 4 times wealthier than the poorest 20%). As you can see from the graph, there seems to be a downward trend in life expectancy from the more equal countries, especially Sweden and Japan, to the less equal countries. The low life expectancies of Denmark and Finland should make us wonder whether this graph really proves that egalitarianism results in good health, but a broad correlation remains nonetheless.

But there are two major problems. Firstly, a number of wealthy societies are missing from this graph. In particular, where are Hong Kong, South Korea, the Czech Republic, Slovenia? All of them are wealthier than Portugal and should be shown. 

Secondly, there are questions over the data used to ascertain life expectancy. Wilkinson and Pickett use figures from the United Nations Human Development Report, a respected and accurate source, but they use data from the 2004 edition. This is an odd choice because elsewhere in The Spirit Level, the authors rely on the 2006 edition. They even use life expectancy figures from the 2006 report elsewhere (the graph on page 7) so we know they were aware of them. 

This is their reference for a graph they use to show life expectancy earlier in the book...



But this is their reference for their graph showing life expectancy against inequality...


So why the use of old statistics and a carefully selected sample group? If we look at the data from the UN's 2006 report—and include all relevant countries—a very different picture emerges.



The association between inequality and life expectancy has disappeared, replaced by a modest trend in the opposite direction. And to prove that the 2006 report is not an anomaly, here are the results from the 2009 report:



Again, there is no correlation. Indeed the three worst performing countries have a very equal distribution of wealth (Czech Republic, Slovenia and Denmark). 

It's worth noting that the inequality/life expectancy hypothesis is not new. It first came to prominence in a 1992 British Medical Journal article, written by none other than Richard Wilkinson. That article inspired a flurry of research and, in The Spirit Level, the authors refer to a “vast literature” on the subject. There is, however, no mention of how much of this vast literature was written by Wilkinson himself, nor that much of the rest was critical of his theory. 

His BMJ study was debunked at length in the same journal in 1995 by Ken Judge. Judge pointed out numerous errors in Wilkinson’s research, including the use of “inappropriate” data. He criticised Wilkinson for using the lowest 70% of families as a measure of inequality when a more conventional measure is the bottom 10% or 20% of individuals. “The suspicion,” wrote Judge, “must be that the choice is derived from the data” (ie. Wilkinson was cherry-picking).

When Judge recalculated the data based on the more usual measure of income per head, the association between life expectancy and inequality disappeared. Judge concluded:

In retrospect, it seems extraordinary that a predominantly monocausal explanation of international variations in life expectancy should ever have been regarded as plausible. It is much more likely that they are the product of many influences, which probably interact over long periods of time.

This was only common sense. Further studies conducted in Denmark and Japan failed to support Wilkinson’s hypothesis and although some studies showed an association between income inequality and life expectancy in the USA, other evidence showed that this was more likely to be due to education, underinvestment and other confounding factors. In 2002, a large study of wealthy European countries showed no association between inequality and life expectancy.

In The Spirit Level, Wilkinson and Pickett cite a 1996 editorial from the BMJ which discussed the “big idea” that “the more equally wealth is distributed the better the health of that society.” At that time, the BMJ was broadly supportive of the theory but research into it was still in its infancy. Wilkinson and Pickett do not mention the editorial that appeared in the same journal six years later, which concluded: 

Now that good data on income inequality have become available for 16 western industrialised countries, the association between income inequality and life expectancy has disappeared.

By the time The Spirit Level was published, the slender circumstantial evidence upon which Wilkinson had made his case had been obliterated with the passing of time. Rather than change the theory to fit the facts, he and Pickett ignored the facts and persisted with the theory. In practice, that meant using old data when they were quite aware that more recent data was available.

2 comments:

DaveA said...

Hi Chris,

Looking at your graphs it confirmed a theory of mine/others. Have you noticed that equality of income is most equal, the colder the country?

I'll explain later, I am going out.

Anonymous said...

Slovenia and the Czech Rep have low levels of inequality and low life expectancy for very specific historical reasons related to their transitional status, so may not be fair to judge the overall trend by them. But Hong Kong certainly does throw a curve ball.