Of course, from a humanitarian or utilitarian viewpoint, whether the increase in average life expectancy is a consequence of reducing inequality or reducing poverty may seem irrelevant. But the political importance is that without a general effect of inequality the rich of country A are better off than the rich of country B – their individual life expectancy is (albeit marginally) actually reduced as a consequence of any reduction in inequality that involves a reduction in their incomes. Were a politician in country A to point to country B’s income distribution and life expectancy to show the benefits of redistribution to the poor, the rich would be likely to reject such a policy on the basis that they would only see their incomes reduced with no benefit to them. A hard sell indeed. Note the other lesson from this diagram – that even though there is a clear relationship between individual income and individual life-expectancy, the average incomes of country A and country B are more or less identical despite a large difference in average life-expectancy. This is another feature of the relationships between incomes and health that make statistical interpretation a trap for the unwary.
Of course, there are many different distribution patterns that can exist – an even spread of incomes is most improbable – but Figure 2 demonstrates that a shift from the red line of Figure 1 to the purple line (also directly benefitting the wealthy) is not so easily distinguished from a shift from red to blue (not directly benefitting the wealthy) as Wilkinson and Pickett’s work might suggest (or at least as easy as those reading it might have thought).
The second potential problem for Wilkinson and Pickett is that even when individual income is statistically accounted for, and there is a remaining definite association of average welfare with income inequality over and above this – as is the case for most regional analyses of the United States – there may be some third factor which tends to cause both inequality and the poor outcome (such as low average life expectancy). Particularly for the US, racial divisions are a strong candidate here. In this case, income inequality (as opposed to income itself) and low average life expectancy will be correlated in that they tend to occur together, but there is no necessary direct causation from inequality to low average life expectancy. We would still not therefore expect to see an improvement in the life-expectancy of the better off as a direct consequence of a public policy of redistribution, but only as a result of tackling this third factor.
These complex statistical relationships, and the failure of some of the earlier work on inequality and health to address them, has rather muddied the water on an issue which is inevitably controversial. Wilkinson and Pickett themselves were probably a little naïve not to expect the barrage of (often, but not always, partisan and skewed) criticism that has come their way since ‘The Spirit Level’s’ publication. In particular, given the points made above, it was always going to be easy to attack the rather simplistic use of correlation diagrams in their book.