Browse Search Feedback Other Links Home Home
The Talk.Origins Archive

Hellfire, crime and Natural Selection (with a pinch of sex)

Post of the Month: July 2012


Subject:    | Hellfire revisited  - evolution of religion, cooperation …
Date:       | 11 Aug 2012
Message-ID: |

Given a certain tendency to shoot at messengers, especially when as in a new study on the impact of religion on crime rate has caused a bit of a stir, and as the authors link it also to the evolution of cooperation and the evolution of religion (as one of the "social glue" type of theories that advocate that a tendency to reason about invisible powerful entities was selected for in our distant past), I thought it might be of interest to TO.

A typical pop since account is here:

The full study is open access and available here:

Plos1 has also a "comment" section, worth looking into (I say a bit more about them later)

The findings are pretty depressing. In essence: Religious beliefs do have an effect on crime rates. The nastier the religion is (fire and brimstone, emphasis on hell, punishment, eternal torture) the better as far as crime is concerned – "nice" religions (universal salvation, let's all have a hug and be friends type) also have an effect, just in the opposite direction, they actually increase crime rate – or so it seems. This summary is in effect a little bit (intentionally) careless, though this is how many have interpreted the results.

As you can see when you go to the "comment" section, that article got quite some media coverage, including, unsurprisingly, in religious newsletters – so you can expect people to refer to it, with varying degrees of accuracy representing what the article says, so worth reading for that reason alone.

OK, bit of background for those of you not having next office door neighbours who are criminologists, – the rest can skip a few paragraphs.

The idea of studying the interaction between religion and crime is as old as criminology and sociology itself. Auguste Comte formulated the theory that religion increases social coherence and deters crime – which then led him (being an atheist himself) to formulate the project to set up a "positive religion" or "religion of humanity", a purely secular religion that should take on the same role but without the metaphysical baggage.

Closer to modern notions of scientific studies was Durkheim's study on suicide, which became the classical text for the deterrent effect of religion on deviant behaviour. (Durkheim argued that religious beliefs can reduce crime – but bear in mind that he thought that crime was something potentially quite good..) It really became mainstream however in the 1950s, with the rise of "control theory" in criminology (which in turn responded to the increase in crime in western postwar societies). Early control theory is most closely related with the names of Albert Reiss (who started it) , Ivan Nye and more than anyone else Travis Hirschi. While religion is just one aspect of control in his theory, Hirschi did some ground breaking work on that aspect, and he gave the name "hellfire hypothesis" to the idea that threat of eternal punishment is negatively correlated with crime rates. (classical text: Hirschi, T. & Stark, R. (1969). "Hellfire and Delinquency." Social Problems 17:202-213)

Ever since, lots of criminologists have tried to either confirm or dis-confirm the hypothesis (as Hirschi and Stark argued themselves), creating in the process some extraordinarily good studies, but also some methodologically abysmal ones, and the methodological issues are substantial. A huge body of literature has as a result emerged from this, and it is difficult to give an overall appreciation

The result is confusing, to put it mildly, with some studies showing strong negative correlations, many neutral effects, and a few a positive correlation (religion as criminogenic)". One recurrent theme in all these studies and the discussion they trigger are issues of methodology – there is an unfortunate focus on your delinquency, a focus on convicted offenders rather than society wide sampling, the problem to get good crime data and also an absence of longitudinal studies (that for me is the biggest issue, but these are expensive – the very large cohort study my colleagues are running on youth crime and delinquency in Scotland extremely so.)

One recurrent theme is how to design the tests, and what to control for – often, initial correlations disappear when the analysis is subsequently extended or refined. This led one of the most influential researchers in the field conclude that the correlation is probably spurious – John Cochran, P Wood and B Arneklev: Is the Religiosity-Delinquency Relationship Spurious? A Test of Arousal and Social Control Theories.

Spurious in both directions that is, there is no positive or negative correlation according to this study. Chochrane's PhD btw was also on that topic, and is available free here:

More recently (2001), Colin Baier and Bradley Wright did a meta study of the 60 largest (and, methodologically, least controversial ) studies: If You Love Me, Keep My Commandments": A Meta-Analysis of the Effect of Religion on Crime Journal of Research in Crime & Delinquency, 38 (1), pp. 3-21. Their results of the meta-analysis show that religious beliefs and behaviours exert a moderate deterrent effect on individuals' criminal behaviour. However, this general trend hides significant differences, so a better conclusion could be: it all depends. (Big surprise that)

Some religious beliefs deter some forms of deviance – and more so in some societies than others. So yes, the number of drunk driving offences in Saudi Arabia is very low – they simply don't allow woman to drive <frantically ducks for cover…> Generally, one sub-theory that has some empirical support is that religious beliefs can support asceticism, and hence reduces "consumption crimes" (which can have a domino effect if you consider the relation between alcohol in particular and violent crimes) But even for that , the evidence is mixed: Chochrane and Akan (1989) Beyond Hellfire: An Exploration of the Variable Effects of Religiosity on Adolescent Marijuana Use. Journal of Research in Crime & Delinquency. 26 (3), pp. 198-255 is a typical "it all rather depends" study, with a general negative correlation between drug use and religious belief, but it depends on the belief, how it is measured, and the drugs.

Rarer are the studies that identify a positive correlation between religion and crime, but these too exist. One example is Charles Kimball, When religion becomes Lethal, (2003) which argues that dualist (struggle between good and bad) religions with a high degree of punitativeness and absolutist truth claims fare particularly badly with regards to homicide (which contributes to the explanation of the disproportionately high homicide rate in the southern part of the US, see also Ellison C.G., J. A. Burr, and P. McCall. 2003. "The Enduring Puzzle of Southern Homicide: Is Regional Religious Culture the Missing Piece?" A Homicide Studies 2003; 7: 326-352 ).

So, by and large, criminologists don't even ask any longer: "does religion deter crime?" but: "what aspects of a specific religion deter what sorts of crime under what wider social conditions?" That is the wider context of the study linked to above, the most recent offering. It does "drill down" to specific aspects of religions, in particular belief in hell and heaven. But it is more ambitious on the crime side, and looks at a variety of core crimes. What however does it actually say? Despite reports to the contrary, it does NOT claim that belief in hell reduces crime. Rather, it says that in comparison to belief in heaven, belief in hell reduces crime, while belief in heaven increases it. It "may" also be the case the belief in hell is an absolute deterrent – but the study is agnostic about it. It only compares two types of religious belief. In essence, the result is: people react better to punishment than to rewards, which again is depressing for all sorts of reasons (and has massive implications for all sorts of issues – my approach to marking e.g. is all wrong…)

Quite a number of people misread this, including one of the critics on the comment page, Paul. That one is, in a way, quite funny – as I said above, the research into the hellfire hypothesis is always going to be methodologically difficult (as is any study in the causes of crime) and there are a couple of studies around where you wonder why in hell (pun intended) the referees let that one through. Paul's own. "Cross-National Correlations of Quantifiable Societal Health with Popularity Religiosity and Secularism in the Prosperous Democracies". Journal of Religion & Society 7: 1-17 is in my opinion one of them, and I have used it in our method courses as an example of what you really should not do. So misunderstanding the article, then criticising its methodology on the basis of this misunderstanding, and complaining that one's own work is not cited enough is… cheeky I think is the technical term.

Having said that, if you take the study together with the meta-analysis by Baier and Wright (and some of the lab research by the authors of this study), it seems not too implausible to speculate that the weak negative correlation between religion and crime might in fact be much stronger for some religions (and, and that is the really depressing part, the nasty ones for that) who are however "dragged down" by the nicer ones.

What can we say about the validity of the study? They use a much larger data set than most other studies, and most importantly, they also look at data over time. That is very often not done and I'd consider it crucial – they do however in their analysis not make the most of it. So it is not a real longitudinal cohort study that would be much better to settle the issue – but closer to it than many others. They are pretty unique in combining lab studies with etiological macro studies, another great plus. The quality of the statistical work is also very good. As always, there are issues – and they acknowledge that the study allows for different interpretations in addition to the "simple causal one" they prefer on balance.

The points made by some of the commentators address some of these, and you can look at the rebuttal of the authors yourself. My first point too would have been that belief in hell and heaven are too strongly correlated, and the surveys they use too coarse grained – but their answer to this made sense to me. There is as always problems with the reliability and comparability of the crime data. Heck, even within nations, method of crime data collection, classification and recording differ widely – in the UK e.g. the British crime survey asks a sample of people if they have become victims of crime, whereas the police statistics record reported crime – the two diverge for obvious reasons.

More problematically, some of the crimes are heavily influenced by cultural norms – both in the way they are defined and the way they are enforced. Rape is the obvious example – whether marital rape is in our out differs between countries. In Islamic jurisdictions, as a haddith crime, you'd need five male witnesses of good standing for a conviction – and if the accusation fails to result in a conviction, the victim is in some countries likely to face prosecution for adultery. Social norms also impact heavily on the willingness to report this crime even where the law is more favourably disposed. Kidnapping too might be distorted by cultural norms – as a large percentage are kidnappings by a parent as part of a custody battle . How this is classified varies between countries. Human trafficking is another crime where enforcement, and classification, differs substantially, and some of it is culturally caused (police attitudes towards "foreigners" e.g. general attitudes on immigration etc etc)

The authors argue that this does not apply for the "gold standard", homicide, and the fact that all the other categories bar two diverge on the findings for homicide is a good indication that the correlation is real for them too. While it is true that homicide is for these studies seen as gold standard, it is a bit of tarnished gold. It only records cases where the unnatural death was established – Harold Shipman anyone? So countries with a better forensics/medical coroner infrastructure will have higher homicide rates for this reason alone – but that should be unrelated to religious/cultural beliefs, unless a religion prohibits autopsy. Another factor is quality of medical care – some of the reduction in homicide in the UK since the 80s e.g. is attributable to improved medical care for assault victims – its simply that more people can be saved after a knife or gun attack, and then count as assaults. So in theory at least, if you have say a religious prohibition against blood transfer, that could artificially increase your homicide rate for the purpose of a study like this one. However, while all these factors can influence the absolute numbers, the authors have a point when they argue that none of them can explain the convergence between the categories.

Personally, I'm also a bit doubtful if "countries" are not way too big as entities, especially when the in-country differences are larger than those between countries – but they do a very good job in controlling for a variety of factors that could be affected by this.

My own overall conclusion is that the study is too good to be dismissed easily and raises some interesting issues (and some worrying conclusions for policy, regardless of the religious issue). The authors also make a link to the evolution of cooperation, and with that the "social glue" theories of selection for a predisposition to evoke invisible, powerful and and omnipresent entities. Even though I'm all in favour of this line of reasoning (they cite the study by Johnson (2011) Why God is the best punisher. Religion Brain Behav 1: 77-84 that I cited before) I'm actually unsure if this specific study really fits into this debate. Our surveillance culture with omnipresent CCTV coverage should on a macro-level drown out any correlation you can establish on a macro (nation state) level – the laboratory based work by the same authors (e.g. Shariff AF, Norenzayan A (2007) God is watching you: Supernatural agent concepts increase prosocial behavior in an anonymous economic game. Psychol Sci 18: 803-809 or Norenzayan, A., Shariff, A.F., & Gervais, W. (2010). The Evolution of Religious Misbelief. Behavioral and Brain Sciences, 32, 531-532.

Home Page | Browse | Search | Feedback | Links
The FAQ | Must-Read Files | Index | Creationism | Evolution | Age of the Earth | Flood Geology | Catastrophism | Debates