I looked into the researcher's responses to the criticism. It does not exclude people who believe in both. In excluded people who believe in neither. Because ~95% of people who believe in hell also believe in heaven, trying to evaluate for the effects of belief in Hell told basically the same story as believing in heaven which was increased crime.

But when they compared people who believe in heaven w/o hell to people who believe in heaven w/ hell, they found that in a population with otherwise similar beliefs that crime went down when you just isolate for Hell.

As for the variance over time, they found ~5% variance in populations that they did have data for over the full time period; so, they accounted for this when determining uncertainty.

As for variance in national cultures, the reason I feel breaking up the results by nation is that you then get all of the other cultural factors of each nation that may be independent of their religious beliefs skewing each figure. If you average the impacts of other cultural factors across your whole sample group, then (assuming you have an adequate sample size), those other factors should average out leaving just the factors you are testing for.

As for measurement variance, that is indeed a complicating factor. But as spire.sciencespo.fr/hdl:/2441/f6h8764enu2lskk9p48i5a4sj/… shows, certain crimes such as homicide are much less prone to measurement errors, and homicide still showed a significant difference in the affect of Hell study.

One part I would agree requires more research is this is a correlational study. It demonstrates the relationship between belief structures, but does not give evidence to cause. Because it relies on averaging out extraneous variables, from the data, there is no way of knowing if religions of certain types cause certain behaviors, or if they appeal to different individuals.

While further research is needed there, I still feel it is a much more relevant study to the topic than one that is just measuring the effects of religion on crime.