From Psychology Today:
An international study that appeared last year came to the rather surprising conclusion that countries where belief in Heaven is more common than belief in Hell have higher crime rates than countries where the two beliefs are more on par (Shariff & Rhemtulla, 2012). They argued that belief in supernatural reward needs to be balanced out with a corresponding belief in punishment to have a beneficial effect on society in terms of suppressing antisocial behavior. These findings are puzzling considering that an earlier study on international homicide rates using similar sources of data came to the opposite conclusion, finding that countries with more belief in Hell had higher rates of homicide. Critics have argued that the results of the study by Shariff and Rhemtulla are distorted by statistical problems and therefore are not valid. Discussion of the social effects of religious beliefs needs to be kept in perspective considering that the most peaceful, least violent nations in the developed world today, such as Sweden, are also generally the least religious. Personally, I suspect that belief in Hell may create more problems than it solves and that mature societies seem to do well without it.
Social scientists have developed conflicting theories about the role of religion in society. Some scholars have argued that religion functions to regulate individual behavior in order to ensure that people abide by normative social rules rather than pursuing their own selfish interests to the detriment of others. Religious institutions therefore offer supernatural rewards in the afterlife for good behavior (i.e. Heaven), while threatening bad behavior with supernatural punishment (i.e. Hell). On the other hand, others have pointed out that religion has often been a divisive force that promotes in-group loyalty and out-group derogation. That is, religion can foster an “us versus them” mentality that often escalates into sectarian violence and extremism (Jensen, 2006). Jensen has argued that religious cosmologies based on the belief in a cosmic struggle between good and evil forces, such as God and the Devil, Heaven and Hell, facilitate lethal violence by promoting intolerance and a disinclination to negotiate or compromise. Such a worldview fosters development of rigid moral boundaries with the result that people become quick to take offense and hence more willing to punish those they perceive as disrespecting their beliefs.
Shariff and Rhemtulla (2012) favour the view that religion functions to promote good behavior and suppress selfishness. However, they have argued that some religious beliefs are more effective in this regard than others. Specifically, they have argued that the threat of supernatural punishment is actually more effective in suppressing bad behavior than offers of supernatural rewards. In fact, they have argued that belief in supernatural rewards may be counterproductive in this respect. As evidence they cite an earlier study (Shariff & Norenzayan, 2011) in which students who believed in a forgiving God were more likely to cheat on an experimental task than students who believed in an angry and punishing God. They took this to mean that belief in God’s forgiveness encourages moral laxity, as people think they can get away with naughty behavior and be let off the hook so to speak. On the other hand, those who believed in a stern disciplinarian God were scared into toeing the line. (Strangely enough, participants who did not believe in God at all were not more likely to cheat than those who did. Maybe positive and negative God-beliefs cancel each other out among the believers.)
Extending these findings to national levels, they argued that belief in Hell (threat of punishment) should deter crime, while belief in Heaven, like belief in a forgiving God, might actually have an opposite effect. To examine this, they looked at national crime data for nine types of crime from 67 countries, along with data from international surveys about national rates of belief in God, Heaven and Hell. Although belief in Heaven and Hell respectively are very strongly associated as one would expect, rates of belief in Heaven tend to be higher, sometimes substantially, than rates of belief in Hell. Considering how many people find the doctrine of Hell repugnant, this hardly seems surprising. Using multiple regression they found that, as they expected, greater national belief in Hell predicted lower crime rates, while greater national belief in Heaven actually predicted higher crime rates. They still found these same effects even after controlling for a number of relevant variables, such as gross domestic product and income inequality. To further separate the effects of each belief, they subtracted belief in Hell from belief in Heaven for each country. They found that the difference between the two rates of belief was positively correlated with overall crime rates. See the graph below. In fact, if their analysis is correct, the difference between the two beliefs explains a massive 54% of the variance in crime rate, which is an extraordinarily large effect. They took these results as support for their idea that belief in supernatural reward (Heaven) in the absence of belief in punishment (Hell) is ineffective in suppressing antisocial behavior. That is, a big carrot without a big stick does not work very well. This might be because people who believe in Heaven but not in hell expect to be forgiven for their wrong-doings. They did acknowledge the correlationalnature of their study, which makes the direction of causality difficult to assess. However, they argued that their hypothesis was the most parsimonious explanation of the findings.
However, these results have been called into question. Something I found strange about the reporting of Shariff and Rhemtulla was that they did not report the simple correlations between their predictor variables, belief in Heaven and Hell, and the outcome variables, i.e. crime rates. This is interesting because there is evidence that greater belief in Heaven and Hell at national levels are each associated with higher national homicide rates (Jensen, 2006; Robbins, 2012). Similarly, a survey of first world nations found that the countries that were most socially successful – as determined by a measure that combined data from such factors as homicide rate, lifespan, poverty levels, life satisfaction and several other factors – were also the most non-religious (Paul, 2009). Specifically, the more successful nations on these measures (e.g. Sweden) generally had lower rates of belief in God, Heaven, and Hell respectively than the less successful ones (e.g. the USA). Furthermore, an earlier study testing the theory that “malevolent religious cosmologies” are associated with lethal violence, found opposite results to those of Shariff and Rhemtulla. That is, countries with higher national rates of “malevolent” religious beliefs (specifically in Hell and the Devil) had worse homicide rates, while those with higher national rates of “benevolent” religious beliefs (specifically in Heaven and God) had lower homicide rates (Jensen, 2006). (An associated finding was that intensity of religious belief was also associated positively with homicide rates, indicating that nations that are more serious about religion tend to have more killings.) This is in spite of the fact that Jensen used a similar methodology to Shariff and Rhemtulla, such as reliance on data about religious beliefs from the World Values Survey and multiple regression analysis. Jensen also noted that countries that had high levels of belief in both God and the Devil had significantly higher homicide rates compared with countries with relatively high belief in God but low belief in the Devil, and compared with the most secular nations where there is reduced belief in both God and the Devil respectively.
One scholarly commentator on the Shariff and Rhemtulla study, Blaine Robbins, argued that their results were most likely due to statistical artifacts rather than genuine trends. Specifically, belief in Heaven and Hell have an extremely high positive correlation with each other (r = .93). Using such highly correlated predictor variables in multiple regression analysis can distort the results for the individual predictors in random ways, due to a problem known as multicollinearity. Multicollinearity problems might explain why belief in Heaven and in Hell each have positive correlations with homicide rates, yet in Shariff and Rhemtulla’s study belief in Hell changes from a positive to a negative predictor, while in Jensen’s study it is belief in Heaven rather than Hell that changes from a positive to a negative predictor. Additionally, Robbins conducted her own analyses using slightly different methods and found non-significant results. This suggests that Shariff and Rhemtulla’s are probably not statistically robust.
In their response to their critics, Shariff and Rhemtulla give various reasons why they think that multicollinearity was not a problem in their analysis. However, they did not explain why Robbins’ analysis yielded contradictory and non-significant results, or why Jensen found that countries with mainly “benevolent” religious beliefs had lower homicide rates than countries with a more even mixture of “benevolent” and “malevolent” beliefs.
Shariff and Rhemtulla also argued that their use of a composite measure subtracting belief in Hell from belief in Heaven would have eliminated issues with multicollinearity. One thing I want to note about this measure is that it considers the difference between the two beliefs but does not appear to consider the absolute levels of belief for each country. Jensen’s analysis of whether nations had rates of both malevolent and benevolent beliefs compared to those that had lower rates of malevolent beliefs compared to benevolent ones did take this into account, e.g. some nations were classed as “high dualists” (high levels of both beliefs), while others were more “moderate dualists.” In Sweden for example less than half of the population believes in either Heaven or Hell, so any difference between rates of the two beliefs would only affect a minority of the population. This is in contrast to Mexico for instance where more than three quarters of the population profess such beliefs. Within each of these countries though, the difference between rates of the two beliefs is somewhat similar, even though the two nations are far apart in terms of overall belief. Speaking of these two nations, Shariff and Rhemtulla’s chart indicates that Sweden has a higher crime rate than Mexico. Considering that Mexico has one of the highest murder rates in the world, not to mention high rates of crime in general, while Sweden has one of the lowest murder rates, I have serious reservations about the accuracy of the national crime rate data used in this analysis, a concern also expressed by scholar Gregory Paul.
Even though Shariff and Rhemtulla argue that belief in supernatural punishment can benefit society by helping to suppress antisocial behavior, I suspect that that such beliefs may have undesirable social costs. Belief in supernatural punishment as severely harsh as spending an eternity in Hellimplies that some people must be extremely evil to deserve such harsh treatment. A study on belief in pure evil found that people who believed that some people are completely evil are more likely to support violent retributive policies such as capital punishment and the use of torture at Guantanamo Bay (Campbell & Vollhardt, 2013). On a related topic, another study found that people who believed in a literal Satan tended to be intolerant of homosexuals and somewhat prejudiced against minorities generally (Wilson & Huff, 2001). Belief in Hell is related to belief in evil and in the Devil, and although I do not have proof, I would venture a guess that belief in Hell at a national level is probably associated with greater support for retributive policies such as capital punishment and torture, as well as with prejudice against people who violate religious norms such as gays. Aside from being an extremely cruel thing to believe in, the idea of Hell may have done more harm to society than good.
I did my own quick and dirty analysis of the correlation between the difference in rates of belief in Heaven and Hell respectively (using data from the World Values Survey) and the national homicide rate in 65 countries (using UNODC data) and the non-significant correlation (r = .08) explained a paltry 0.6% of the variance. Admittedly, this was a very basic analysis, but still the disparity from S&R's results makes me wonder.
According to World Values Survey data, belief in Heaven and Hell in Mexico are 88.2% and 74.6% respectively, while in Sweden they are 31.1% and 8.8% respectively.
 I am aware that Eastern religions regard Hell as involving a temporary state of punishment rather than an eternity. This is another factor not addressed in Shariff and Rhemtulla’s study.