Adolescents in the ADD Health Study were asked if they had intentionally injured someone in the past year. I estimated logistic regression models to see if: 1) non-Asian minorities (NAMs) were more likely to answer yes, and 2) are the racial/ethnic differences explained by lower IQs.
The coefficient for blacks is 0.34 (p = .001) and 0.24 for Hispanics (p = .013). The numbers indicate that members of both groups are more likely to have injured someone. When a measure of IQ is added to the equations (it's a vocab test), the results look like this:
Logistic regression coefficients
Hispanic model (n = 6,150)
Black model (n = 6,154)
* p < .05, two-tailed test
When IQ is entered into the model the Hispanic effect disappears. In other words, lower IQs explain why Hispanics are more likely to be violent. On the other hand, entering IQ into the model for blacks reduces the relationship between being black and violent, but it does not eliminate it. According to ADD Health data, other factors are necessary to fully explain high levels of violence by blacks.