Religiosity and crime

I was stunned to read in We Are Doomed by John Derbyshire the argument that since the crime rate is high in religious Mississippi and low in irreligious Oregon, religion does not reduce crime. I wouldn't be surprised to see that kind of reasoning by a superficial culturist thinker, but by the Derb?! (Heather Mac Donald, a normally strong analyst herself, has made similar comparisons). Just as I regained composure, he goes on to make the same argument comparing Nigeria and Japan! Holy crap! I expected the next comparison to be between humans and Martians!

(Perhaps Derbyshire is thinking that religion is impotent since Oregon and Mississippi show that other factors like race, ethnicity, and subculture can overpower religion, but that is not clear in his writing.)

If we follow Derbyshire's logic, then religious individuals in Oregon or Mississippi should be more likely to be criminals. After all, persons, not states, commit crimes. 

I can't isolate Oregon and Mississippi with the GSS, but I can look at the Pacific and East South Central divisions--the least and most religious regions, according to the church frequency question on the GSS. I calculated logistic regression cofficients for the relationship between frequency of church attendance and ever having been arrested. The coefficient for the Pacific region is -.18 (N = 1,353, p < .001). For the southern region, it is -.33 (N = 666, p < .001).

Perhaps we should focus on a northern region instead of the Pacific region which includes the high crime state California (a very irreligious state, by the way). The coefficient for New England is -.30 (N = 460, p < .001). All of the coefficients are the same or larger if we limit the analysis to whites. When we focus on real-life individuals, the irreligious ones are at higher risk of arrest.