Monday, September 3, 2012

On Crime and Punishment

Quite regularly, I take a look at the "Graphic Detail" blog that's published each business day on The Economist's website. Many of the graphs, maps and infographics that they produce are rather interesting.
Today's one is taken from a recent study, "Divergent Effects of Beliefs in Heaven an Hell on National Crime Rates", published by Azim Shariff and Mijke Rhemtulla  in the open-access journal, Plos One.
Here's the abstract from that paper:

"Though religion has been shown to have generally positive effects on normative ‘prosocial’ behavior, recent laboratory research suggests that these effects may be driven primarily by supernatural punishment. Supernatural benevolence, on the other hand, may actually be associated with less prosocial behavior. Here, we investigate these effects at the societal level, showing that the proportion of people who believe in hell negatively predicts national crime rates whereas belief in heaven predicts higher crime rates. These effects remain after accounting for a host of covariates, and ultimately prove stronger predictors of national crime rates than economic variables such as GDP and income inequality. Expanding on laboratory research on religious prosociality, this is the first study to tie religious beliefs to large-scale cross-national trends in pro- and anti-social behavior."
 You've probably guessed that the authors aren't econometricians!

Oh yes - here's the graph in question:

Now, to be fair to the authors, they do acknowledge (in the "Limitations and Future Directions" section of their paper):

"First and foremost, these findings are correlational, and thus reverse-causation and third variable explanations need to be discounted before causal claims can be firmly endorsed."
 However, they then (immediately) continue:
"However, at least two reasons suggest that a causal effect of these religious beliefs on crime is a plausible explanation for the pattern of results.

First, obvious third variable candidates such as differences between countries in national personality, wealth, wealth distribution, and general religiosity show no indication of driving the effects. Second, numerous lab studies have established direct causal effects for religious beliefs on both pro- and anti-social behaviors. The possibility remains that the lab effects and the international crime rate effects are entirely unrelated, but parsimony suggests that both are, at least to some degree, a reflection of the same underlying causal story. Nevertheless, future research would be beneficially directed towards addressing possible alternative explanations."
My interpretation of what they're saying here - "we believe there's a causal relationship".

If you take a look at this study, I think you'll find quite a few things that warrant an eye-roll.

I'll be interested to see your comments about how you would have undertaken the statistical analysis differently!

© 2012, David E. Giles

1 comment:

  1. I think I'm more interested in how "lab studies have established direct causal effects for religious beliefs on both pro- and anti-social behaviors."! What did those experimentes entail, subjecting pyschology grad students to sermons (or the obverse making them read some non-fiction book about evolution or something similar)?

    In terms of how you would go about the study, it is very difficult. There are all sorts of problems to deal with given such data. You have problems just in terms of identifying unique effects due to aggregation bias and a seemingly endless number of potential ommitted variables, especially when the effect of interest is not reasonably exogenous of other socio-economic and demographic factors.

    You also have just more generic data problems in general. Frequently reported crime types aren't necessarily similar between nations. I'm skeptical off-the-cuff in their list of crimes that drug crime and theft are consistent between nations, these things frequently aren't even consistent within the US! You also have the problem that you only have a weird variety of nations reporting useable crime statistics, so you end up with a strange sample of mostly European and American nations, and a smaller proportion of African and Asian countries. It is just a mess to deal with.

    To be fair though one could find mainstream articles in prominent criminology journals that utilize similar data. The recent president American Society of Criminology, Steven Messner, is one of the preeminent criminologist's looking at similar cross-national, macro level explanations of crime.

    IMO I would say the strongest articles in said field utilize time series approaches, looking at the same set of nations over a period of time. They also preferably only predict homicide data (it is really the only crime type which near universal consistent definitions and regularly considered to be accurately reported). The effects are interest are also (preferably) plausibly exogenous of other socio-ecomonic and demographic factors. You really can't reasonably control for all such factors, and so the best you can do is hope your "effect of interest" is not confounded with such other factors.


Note: Only a member of this blog may post a comment.