I wonder if race data is just wildly inaccurate and that's what makes single motherhood such a good predictor--because it's a proxy for race.
If you read the whole article (it's long, I know), it identified that race was the greatest single variable for predicting homicide. It notes that once you conduct multivariate analysis it becomes clear that single motherhood is the actual greatest predictor of homicide and that black households are headed by single mothers at extraordinary rates.
E.g. nuclear families of any race? Very little homicide. Child of a single mother? Lots of homicide.
Wouldn't surprise me if different counties have different criteria for how they categorize race, in which case one has to be careful aggregating or comparing data between counties to ensure the result isn't nonsense.
Another thing is affirmative action policies really muddy socio-economic status as well. The academic ability would be wildly different for the kid of an engineer and the kid of a diversity consultant. However, they are likely to be in the same income level. Handing out college degrees does make anyone smarter or better behaved, but earning a college degree (if that can even be done these days) does demonstrates a level of intelligence, planning, and commitment.
(post is archived)