Wednesday, July 6, 2011

More on The Spirit Level

There are quite a few interesting comments on my earlier post on The Spirit Level on Economist's View. I can't respond in detail to all of them, but here are a few additional thoughts.

A few commentators seem to think I'm unsympathetic to the book, which isn't accurate. So let me be more direct in my assessment of The Spirit Level: Why Greater Equality Makes Societies Stronger. I find the empirical data presented to be highly suggestive and interesting. Pickett and Wilkinson demonstrate simple trend relationships between the income inequality score of a set of countries and states and a series of social problems. I myself do not have reason to doubt their statistical methods, though other critics have done so, including Lane Kenworthy in a balanced and reasonable critique in 2010. What troubles me about the central argument of the book is what I find to be a flatly unconvincing hypothesis, that the psychological states created in a country’s population by status conflict and income inequality suffice to explain the variation in health outcomes for individuals in these societies. There are many determinants of a person’s satisfaction with his/her life besides income inequality; and there are many social factors that influence health outcomes for individuals besides their perception of income inequalities and the level of stress that those inequalities create in them.

In my view, a good causal explanation requires an analysis of the mechanisms that bring about the relation between cause and effect. I find only one mechanism underlying their accounts of this relationship – essentially a psychological mechanism linking the individual’s perception of income inequalities to health and behavioral outcomes. They flesh this out with a description of the physiology of stress induced by rising income inequalities leading to rising mental illness; perceived status differences leading to decline in trust; stress based on perceived inequalities leading to increased “comfort eating” and obesity; status competition leads to increased violence; etc. This is what I mean by monocausal; there is just one causal process at the heart of their analysis.

The variety of causal mechanisms that seem pertinent to me in consideration of their data include several different kinds. First, as I mentioned in my post, there are serious and systemic sources of separation between individuals within societies that have nothing to do with income inequalities and that might be expected to have the same kinds of psychological and physiological effects on individuals: racism, ethnic hatred, sexism, economic conflicts that do not have to do with inequalities, and so on. For example, in the United States there are credible efforts to explain racial disparities in health outcomes on the basis of the effects on individuals living in circumstances of endemic racial discrimination and prejudice. (See, for example, Arline T. Geronimus, ScD, Margaret Hicken, MPH, Danya Keene, MAT, and John Bound, PhD, "“Weathering” and Age Patterns of Allostatic Load Scores Among Blacks and Whites in the United States"; link.) These effects are unrelated to income inequalities and derive instead from another aspect of social separation. So race rather than income appears to explain this set of health disparities. So perhaps this implies that if we had a way of measuring a “composite index of non-economic social separation” we might have a similar set of graphs, with societies with a lot of racism and ethnic hatred at the upper right quadrant of the graph. I don’t suppose that this index would be highly correlated with the quintile-quintile ratio that the authors use as their proxy for economic inequality.

Second, there are structural influences on a population’s health outcomes that do not derive from the degree of income inequalities that exist in a society but are nonetheless highly influential on health status of the population. For example, the quality and accessibility of rural healthcare has a large influence on the health status of rural people; the accessibility of health insurance to inner city residents has a large influence; the targeting of food marketing in favor of “comfort foods” has a large potential influence on the whole population; and differential treatments and discrimination within health systems for different populations of patients have large potential effects. None of these factors derive from the ratio of quintile incomes – the degree of income inequality – in the society; they are independent causes of differential health outcomes across sub-populations.

I’m not an expert on statistical reasoning; but of course every reader knows that evidence of correlation between two variables A and B is not conclusive evidence of a causal relation existing between those variables. There may be a common factor that causes variation in both A and B, or the correlation may be flatly spurious. It is necessary to frame a credible and testable hypothesis about how the putative causal condition leads to the putative effect. On the other hand, finding that a proposed mechanism is not in fact valid does not establish that there is not a causal connection between the variables; it only mandates that we then search further for a mechanism that does in fact obtain. Only if we are ultimately convinced that there is no possible mechanism would we come to the conclusion that the correlation is spurious. So my position in the posting under discussion is simply that the authors have not established a mechanism linking income inequalities to bad health outcomes that satisfies me.

One can be opposed to extreme and rising income inequalities in one’s society – as many of the readers of The Spirit Level are – without being convinced of the particular causal connections that the authors assert. And this matters quite a bit; if we want to reform society, we need to have well grounded theories of the causal processes that lead to the outcomes we want to reduce or avoid.


Shane Leavy said...

Interesting stuff indeed. I also found The Spirit Level interesting, while questioning some aspects of their methodology. One point I pondered was that they seemed to use "income" and "status" as interchangeable concepts. That is, they seemed to presume that a high income individual must also be high status.

I question this. In history there are plenty of examples of low or medium wealth individuals with much higher status in society, like Christian priests/preachers or poor samurai in feudal Japan.

In fact I wonder if people like Pickett and Wilkinson, by implying that low income people must also be low status, are making things WORSE, making low income people feel worse about themselves. I wrote about this here, suggesting:

"Perhaps this gives us an idea of why the US scores so poorly in many of The Spirit Level's indicators. The US simultaneously has high inequality, low social mobility and a culture based on the idea that any individual can succeed - the American Dream. In such a society, failure must be difficult to deal with; low status an indication of personal defeat.

By contrast an unequal culture with a more conservative stratification might actually be less stressful. There, a miserably poor farmer, for example, can still feel proud of his very humble successes since expectations are already extremely low."

Anonymous said...

Correlation is not causation. I am strongly opposed to our inequal and unjust society, but inequality is not a satisfactory explanation for the health problems of modern societies. American doctor George Beard proposed in 1869 that there was a disease called "American nervousness" due to the hectic pace of modern life. He seemed to focus on harried wealthy people who just couldn't cope, and his main treatment was electric shocks.

Christianity, Judaism and Islam all said in their formative period that we owed much to the poor and the sick. Now we have Ayn Rand, Paul Ryan and the dynamic duo of Bachmann & Palin claiming to be Christians and saying that the poor are moochers, who are to blame for their own problems. Rand didn't claim to be Christian of course. P & W are too simplistic.

Peter T said...

Fair enough, but my recollection of Wilkinson's argument is that there is a direct causal link, well-attested, between social inequality and ill-health (for an unforgettable account of some early research on this in baboons, see Robert Sapolsky's Memoirs of a Primate). Social inequality in modern societies is strongly correlated with economic inequality, and most of the factors you mention are also strongly correlated with economic inequality (eg racial attitudes and the rural/urban divide). So one would expect that economic inequality would be correlated with ill health and much else. And the authors demonstrate that indeed it is.

This is not to say that a policy of direct income transfers would be an appropriate remedy. For one thing, these are unlikely to be implemented or sustained in the absence of a supportive social and political climate - which the US clearly lacks. But it does tell us that, over the longer term, lessening inequality will lower the load on the health care, welfare, corrections and similar systems.

AllanW said...

@ Anonymous

Please expand on what you mean by ‘inequality is not a satisfactory explanation for the health problems of modern societies’ in order for me to understand the comment sufficiently. Thanks.

@ Peter T

Your comment is most clear and I wholly agree with it. The one thing I’d add is in relation to this;

‘This is not to say that a policy of direct income transfers would be an appropriate remedy’

Entirely correct because Wilkinson and Pickett think that the dominant causal mechanism concerns social standing and disparities in expectations of lifestyle achievements which cannot be eradicated or ameliorated in solely material ways. A brief discussion of which aspects was held on ‘Economist’s View’ here;

eric said...

@DL: if we want to reform society, we need to have well grounded theories of the causal processes that lead to the outcomes we want to reduce or avoid"

This is far from obvious. FDR arguably reformed society, and he arguably did so by trial and error. I agree that stress alone sounds insufficient to explain large discrepancies in health outcomes, but I cannot imagine that large money inequality does not lead to many other inequalities as well (health care inequities, pollution inequalities, food inequalities, etc.), and so have large indirect effects. We don't have to be able to trace every cause and effect link to decide to try to reduce inequality.

Misaki said...

"serious and systemic sources of separation ... that might be expected to have the same kinds of psychological and physiological effects on individuals: racism, ethnic hatred, sexism, economic conflicts that do not have to do with inequalities, and so on."

One significant difference is that these do not involve signalling as a goal. They are, mostly, things you can do nothing about, and each faction will generally feel content with their own characteristics.

Income however is seen by many as a positive thing, and moreover due to its complexity and scarcity serves for many as an indicator of personal abilities or even of one's worth to society. People instinctively try to conform to standard signals in a society, and internal conflict results when a signal is seen to be unreliable which does not result for the other sources of separation in a society mentioned because they are not seen to be an obtainable certification of some other aspect of performance.

This can be seen not only in people who 'choose' to be poor when the alternative is doing something seen as unethical or inefficient, but also those who are able to easily send a high signal and feel they have no way to communicate the factors the signal is meant to indicate. Such as in this documentary:

chatten said...

Dan: I think I understand your response to The Spirit Level better now. Granted, laying all social problems to one problem caused by inequality is simplistic. But the research on the interconnection between social position and health is very strong (look at the work of Michael Marmot) and so it makes sense that making the distance between social levels worse would also make health and well-being gnerally worse. Racism and discrimination generally can operate independently of inequality, but also can be closely linked to it (as I think it is with African-Americans). The bottom line here is that even though Wilkinson and Pickett can't prove inequality causes all the problems they discuss, the case for both direct and indirect connections is strong. Their work does suggest there's general political economic situation that links high inequality to a host of negative conditions, such as more violent crime, corruption, inadequate spending on social services and discrimination.

Faustina said...

It seems that the link between income inequality and health disappears if racial segregation is taken into account (at least for the US). In communities dominated by blacks income inequality does not correspond with bad health. In communities with a white majority, income inequality influences black people in a negative way. To me this indicates that it is wrong to assume that income equality is the same as status equality. Social cohesion must be something else. Link: and
The last article is just an abstract, unless you pay.