Measured general intelligence predicts differences in violent criminality even within families: of siblings raised in the same household, the one with lower IQ is more likely to commit violent crime.
1) Re the possible SES confound, IQ and crime are correlated among brothers raised together as well: 2) Saying that "IQ has a test-retest correlation of about 0.8" is like saying that all cars in the world have a top speed of 120 mph. Test-retest
It is worth noting that IQ predicts violent crime even *within families within populations* (ie violent crime can differentially be predicted among children who grow up in same SES by their differential IQs), so unlikely that it wouldn't predict beyond GDP