Tuesday, July 17, 2007

Race, class, and crime II: Readers have wondered what the link between race and crime (discussed below) would be if additional factors were controlled. Someone mentioned using mother's education for social class, and another asked about IQ. Myself, I was concerned that age might change the results, so here are the logistic coefficients for race and several controls, with ever having been arrested as the dependent variable:

Logistic regression coefficient (t-statistic in parentheses)

Black .53 (4.17)
Mother's education .01 (.39)
IQ -.02 (-1.00)
Male 1.62 (15.36)
Age -.03 (-8.98)

Number of valid cases 4,831

As a rule of thumb, the t-statistic needs to be at least 2 for the relationship to be considered statistically significant. All thrown in together, neither mother's education or IQ exert an significant net effect on arrest. Using the t-value as an indicator of the magnitude of the relationships, the impact of race is not as strong as sex or age. But keep in mind something I wrote in the comments section of the previous post: there is research comparing self-reports with police reports that suggests that black males underreport being arrested. Evidently, they don't trust interviewers.

No comments:

Post a Comment

Meta-analysis of clinical trials: Eat walnuts

I am always looking for easy eating choices that are good for you. This new meta-analysis of 26 clinical trials looked to see if walnuts ma...