As Left Foot Forward reported on Monday, several recent publications have challenged the central conclusion of the book that more unequal societies do worse than more equal societies, on a variety of social outcomes.
The debate on the conclusions of the influential book The Spirit Level continued today as its authors, Kate Pickett and Richard Wilkinson, clashed with its main critics.
As Left Foot Forward reported on Monday, several recent publications have challenged the central conclusion of the book that more unequal societies do worse than more equal societies, on a variety of social outcomes. The critics focus on the statistical methods of the regression analyses, the selection of the data, and the neglect of explanations other than income inequality.
A central contention of Peter Saunders, co-author of Policy Exchange’s recent pamphlet Beware False Prophets is that excluding outliers renders many of the correlations presented in the book insignificant. A former Professor of Sociology at the University of Sussex, Saunders argued that it was a basic rule of regression analysis that outliers should be excluded.
Prof. Pickett responded by showing that many of the correlations that Saunders questioned regained their statistical significance once another outlier – Japan – was also excluded. She also showed that within the groups of Scandinavian and Anglo countries, which Saunders claims should not be compared because of historical and cultural differences, there are strong correlations between inequality and various measures of social problems.
The Policy Exchange publication also claims that other third variables better explain many findings than does income inequality. For instance, the proportion of Afro-Caribbean population within US States is a more powerful predictor of homicide rates than is income inequality.
Prof. Pickett’s response to this argument was that when running regressions other variables should not be included without a motivating theory, because in soaking up the variation in the dependent variable, they reduce the strength and significance of the income inequality variable.
Chris Snowden, author of The Spirit Level Delusion, questioned the choice of countries in the data sets, and why other rich countries such as Hong Kong and Slovenia had been excluded. In his book he widens the data set, showing that many of the correlations are weakened or disappear in this new context.
Prof. Wilkinson’s main response was that their aim in writing the book had not been to establish the existence of strong causal relationships, but to test a theory of social gradients in a very specific context: rich market democracies.
The definition of rich market democracies countries came from the World Bank, precisely to avoid accusations of picking data to suit their argument. Such an approach also rules out excluding outliers such as the USA and Japan, whether or not they support the contention that social gradients explain many social problems.
The strongest defence made by Prof. Wilkinson was that while there are hundreds of peer-reviewed articles confirming much stronger relationships between inequality and social outcomes than those presented in The Spirit Level, the book tested these relationships in the context of rich countries, and found strong evidence that more equal societies do better.
In defence of their statistical methods and conclusions, the authors presented a graph showing correlation between income inequality and social mobility which had been criticised for relying on just eight data points. Since publication of the book in 2009 more data has become available. When added to the graph these three new data points fall on or very close to the linear regression line.
Wilkinson argued that as predictive power is the strongest test of a scientific theory, this evidence was overwhelming in showing that steeper social gradients cause societies to be less healthy, more violent and criminal, and less successful across a number of other measures.
Leave a Reply