The Son of man shall send forth his angels, and they shall gather out of his kingdom all things that offend, and them which do iniquity; and shall cast them into a furnace of fire: there shall be wailing and gnashing of teeth. ~ Matthew 13:41-42
Men, forgive Him, for He knows not what He has done. ~ Jose Saramago, The gospel according to Jesus Christ
Science does not always attack religion. Recently, scholars have begun to theorize and collect evidence for the idea that religion can be good for us, individually and collectively. They find, for example, that participation in religion is related to happiness, cooperation, and volunteering (Norenzayan & Shariff, 2008). How does religion achieve these positive outcomes? One possibility is that (at least some) religions encourage moral behavior and discourage immoral behavior respectively by offering ultimate rewards and punishments. If justice is slow and uncertain on earth, it is even slower but perfectly reliable in the world to come. And the stakes are high. Bliss and agony are infinite in both magnitude and duration. A popular hypothesis is that through the invention of religion and beliefs in supernaturally administered justice, societies have gained greater social control without investing much. Indeed, the religious masses pay to maintain the bureaucratic infrastructures that represent their religions (and those don’t pay taxes).
Shariff & Rhemtulla (2012) explored the connection between religion and crime. They estimated crime rates in 67 countries from publicly available sources and harvested data from international surveys to estimate the percentage of people who believe in heaven and the percentage of people who believe in hell. Their findings are intriguing. Belief in hell shows a small negative correlation with crime (-.13) over countries, not enough to conclude that fear of eternal punishment is a potent deterrent. At the same time, belief in heaven shows a small positive correlation with crime (.16). Again, this is not enough to conclude that the expectation of supernatural compassion disinhibits would-be criminals. As one can imagine, the beliefs in heaven and hell are positively correlated with each other. This correlation is a whopping .93, and it makes all the difference when one examines the unique predictions of each variable. There are a number of ways to do this. Shariff & Rhemtulla use multiple regression. Here, I use partial correlations. The conclusions are the same.
Assuming that the correlation between the belief in hell and crime rate is of greatest interest, we can compute the partial correlation by statistically controlling the third variable, that is, the belief in heaven. The result is -.74, and it can only be described as massive. One way to put this into words is to say that the crime rate is low in countries where the belief in hell is even more prevalent than one would expect from the high prevalence of the belief in heaven. Remember, these two types of belief are almost perfectly correlated with each other because they are both facets of the same type of theist religion. When this common element is statistically removed, only residuals are left, that is, the beyond-expectation levels of the belief in hell. These residuals are what predict the drop in the crime rate.
The correlation between the belief in heaven and the belief in hell cannot be 1 for this to work. If it were 1, there would be no residuals. The observed correlation of .93 is very close to the ceiling. With a correlation of .96 between the two types of belief the partial between hell and crime would be -.99. If the belief in hell were not so highly correlated with the belief in heaven, the partial correlation would be more modest. If, for example, the two types of belief were correlated at .8, the partial correlation would be -.43. As you can see, the statistical air is thin in this theist region where beliefs in heaven and hell go together. Small changes in the magnitude of this correlation beget large changes in the unique contribution of one belief to the prediction of the crime rate. If the two types of belief were negatively correlated with each other, the partial correlation could approximate 0 (e.g., it is .001 if the correlation between the two beliefs is -.79, the lowest it would get). At this end of the spectrum, large changes in the correlation between beliefs beget small changes in the partial correlation.
What these correlations do not reveal, but Shariff & Rhemtulla are frank about, is that overall, the belief in heaven is more prevalent than the belief in hell. Thank god, you might say. But wait. It is the belief in hell that (if it has causal power) keeps crime rates down. The belief in a forgiving god raises the crime rate. When the belief in hell is at its highest, it is as high, but not higher, than the belief in heaven. At that point, the effects of the two beliefs cancel each other out. The differential effect of the two beliefs is observed only when the belief in heaven is more prevalent than the belief in hell, in which case the crime rate is higher. In other words, the most we can expect from an ambivalent theist religion is that its opponent forces nullify each other. Sharif & Rhemtulla’s data show that the crime rate of a society without religion is just as low as the crime rate of a society with a religion that inspires both a fear of hell and an equally strong hope for paradise. When differences do occur, paradise wins and the crime rate goes up.
This is not good news for the evolutionary story according to which carrot-stick religions have evolved because they support social cohesion at low cost. Such religions seem like a waste. Indeed, they seem counterproductive. Extrapolating from Shariff & Rhemtulla’s data, religion would work as a deterrent of crime only if there were greater fear of eternal punishment than hope for salvation. This, however, sounds like a stressful way to live; perhaps more stressful than living with the risk of being robbed down here on earth.
Norenzayan, A., & Shariff, A. F. (2008). The origin and evolution of religious prosociality. Science, 322, 58-62.
Shariff, A. F., & Rhemtulla, M. (2012). Divergent effects of beliefs in heaven and hell on national crime rates. PLoS ONE, 7(6), e39048.