Distant Correlations

When one learns that George Monbiot has murder, venereal disease and marital breakdown in mind, one would be well advised to lock up daughters, cats, dogs, ferrets and other domesticated animals, and hope for the best. When one learns, however, that he has these things in mind only in abstract consideration of their connection to religion, one may relax a little, though one might want to have a word with the local vicar, lest he become excited; for Mr Monbiot is in bold mood: “The evidence is clear that murder, venereal disease and marital breakdown are all more common in religious cultures”. (George Monbiot, “My heroes are driven by God, but I’m glad my society isn’t” The Guardian, 11th October 2005.)

The evidence that Mr Monbiot finds so clear comes from a statistical research paper published in the Journal of Religion & Society, wherein a Mr Gregory S. Paul writes,

In general, higher rates of belief in and worship of a creator correlate with higher rates of homicide, juvenile and early adult mortality, STD infection rates, teen pregnancy, and abortion in the prosperous democracies. The most theistic prosperous democracy, the U.S., is exceptional, but not in the manner Franklin predicted. The United States is almost always the most dysfunctional of the developed democracies, sometimes spectacularly so, and almost always scores poorly.

Gregory S. Paul, “Cross-National Correlations of Quantifiable Societal Health with Popular Religiosity and Secularism in the Prosperous Democracies: A First Look” Journal of Religion & Society, vol. 7, 2005.

Now, as anyone who has ever been able to tell the difference between a scattergraph and a blunderbuss should know, correlation is a measure of association, not causation; and Mr Paul, unlike Mr Monbiot, is careful to stress this. (If we were to take associations further, in order that we might begin to estimate casual relationships, ultimately we would have to seek explanations outside statistical analyses.) For associations themselves to be meaningful within statistical analyses, it is necessary to take into account the manifold factors that could plausibly impinge on the data; otherwise the associations may be meaningless. Mr Paul, however, though he has been careful to stress only association and not causation, has singularly failed to consider other factors that might plausibly affect his data, as statistician Scott Gilbreath makes clear:

There are many socio-economic data series that vary widely across the eighteen countries [in Mr Paul’s data-set] and that plausibly have a significant impact on social conditions, e.g., income distribution, proportion of GDP spent through government, social and cultural cohesion, fertility and mortality rates, age structure of the population, etc., etc. Failure to look at these and other exogenous data would introduce bias into the results, further calling them into question.

Scott Gilbreath, “From Our Bulging How Not to Do Statistics File”, Weblog: Magic Statistics, 27th September 2005.

Worse still, however, is that Mr Paul has not subjected his data to statistical techniques such as regression analysis, which ought to be used in order to determine correlation coefficients and goodness-to-fit. As Mr Gilbreath points out:

This is simply inexcusable in a research project involving statistical analysis. I have never seen anything like this–either in my professional career or in my university studies of statistics and econometrics.

The three reasons listed [by Mr Paul in his paper] for skipping the regression analysis are all bogus. High variability of degree of correlation is precisely why goodness-of-fit is estimated. If high variability appears to be an issue, that’s all the more reason to run regressions. The second reason, potential causal factors for rates of societal function are complex, is another reason why regression analysis needs to be conducted: to assess the impact of the unspecified correlated (not necessarily causal) factors. Furthermore, this reason seems to contradict what Mr Paul said earlier: The cultural and economic similarity of the developing democracies minimizes the variability of factors outside those being examined. The third reason, that the purpose of the study is not to establish causal relationships, is a red herring. Regression can be used to estimate supposed causal relationships, but it can also be used to calculate correlation coefficients without any implication of causality. In any case, if Mr Paul is worried to avoid causal inferences, there are other equally valid techniques of calculating precise correlation estimates. (ibid.)

Without a proper statistical analysis of the data, an association can be found between almost any two variables. It is in the understanding of this meaninglessness, that Sean Gleeson at the Gleeson Blogomerate has written a fine parody of Mr Paul’s research, happily finding that,

a higher proportion of visits to the Gleeson Bloglomerate seems almost to guarantee the elimination of malaria . . . [and] that in every single country beginning with U plus Colombia, those societies with higher degrees of Gleeson-visiting enjoy more gooder goodness in every single quantifiable basic measure to be studied. These findings suggest avenues of further research.

Sean Gleeson, “Cross-National Correlations of Quantifiable Societal Goodness with Gleeson readership in Countries Beginning with U”, Weblog: Gleeson Blogomerate, 30th September 2005.

If Mr Paul cannot subject his data to the statistical analyses necessary for an understanding of the strength of correlation between religious belief and social dysfunction, what then is the point of his research? He claims that it is “not an attempt to present a definitive study that establishes cause versus effect between religiosity, secularism and societal health”, but since he has made no serious attempt to demonstrate an association between these things, we are left with the suspicion that it is little but fodder for the causal ruminations of newspaper columnists.

And so we come back to George Monbiot, who on the basis of this bogus study is in no doubt that the “evidence is clear that murder, venereal disease and marital breakdown are all more common in religious cultures”.

Now, if the degree to which a culture is religious can be said to be a function of the numbers of the population that adhere to religion, then Mr Monbiot’s proposition is easily refutable. Britain, for instance, has become less of a religious culture since the 1960s, and yet at least one factor – namely, the murder rate –, which Mr Monbiot cites as positively correlated with religious culture, has risen. The number of homicides per million has more than doubled since the 1960s, from 6.8 homicides per million in 1965 to 14.1 in 1997 (Joe Hicks & Grahame Allen, “A Century of Change: Trends in UK Statistics since 1900”, Houses of Parliament Research Paper 99/111, 21st December 1999.), and yet belief in God – which for most people is affixed to religion, and thus is some overall indication of religious belief – has nearly halved, from 77% in 1968 to 44% in 2004 (YouGov poll for The Telegraph, 27th December 2004). I do not suggest a causal link, only that the facts are at odds with Mr Monbiot’s statement.

I am not surprised, however, that Mr Monbiot finds the evidence clear; for we should all know by now that Mr Monbiot views evidence through the ogle-piece of credulity, in whose confident sights the world is shrunk to fit. We might concur with him, however, when he says, “self-doubt is more likely to be absent from the mind of the believer than the infidel”. The shame is all the greater, therefore, that he cannot see the truth of this in himself.