Explaining the connection between never going to church and criminality: Razib at Gene Expression made an interesting point about my GSS finding that 43% of men who never go to church have been arrested:
The main question I would have are the affects of the background environment; in many socially conservative environments the expectation of involvement in a church is very strong and unchurched status could be a signal for anti-social tendencies. I know whereof I speak, I grew up for a while in a 3/4 Republican 99% white region of the Mountain West and those who were unchurched were often those who were "up to no good" (a small minority were secular liberals, but only a very small minority). My own prediction would be that this would be a more common phenomenon in a very religious country like the United States.
One way to test this hypothesis is to see if the relationship between church attendance and the risk of arrest is stronger in regions of the United States where people worship more frequently. The General Social Survey divides the country into nine divisions. I calculated mean church attendance for each region (scores range from "never" (=0) to "more than once a week" (=8)):
Mean church attendance score
East South Central 4.51
West South Central 4.28
West North Central 4.13
South Atlantic 4.10
East North Central 3.91
Middle Atlantic 3.67
New England 3.60
My next step was to estimate the association between arrest and attendance for each of the nine divisions: I did this with logistic regression (sample sizes ranged between 460 and 2,306). I then calculated the Pearson correlation between these logit coefficients and the mean attendance scores displayed above. It is .44. This means that the connection between arrest and never going to church is stronger in areas where churchgoing is most common. So Razib might be right that in religious areas many of the well-adjusted folks feel like they should go to church, leaving a high percentage of antisocial people among the ranks of non-attenders.