Monday, January 17, 2011

More on IQ and delinquency

Using the more appropriate Gamma statistic, I recalculated the relationships displayed in the last post between IQ and 15 types of delinquency. Gamma is similar to Pearson--it ranges from -1.00 to +1.00.

Gamma measure of association

Graffiti -.02
Vandalism .10
Lied to parents .08
Shoplift .03
Serious fight -.14
Seriously injured someone -.10
Ran away from home -.07
Stole a car -.01
Stole something > $50 -.01
Burglarized a building -.04
Used or threatened with weapon -.10
Sold drugs .10
Stole something < $50 .08
Group fight -.16
Rowdy in public .08

The associations are perhaps a bit sharper, but the only consistent negative relationship is between IQ and types of violence. The link is weak, however.

In this new study in the journal Intelligence, the authors find that low-IQ U.S. counties have higher crime rates. Macrolevel studies evidently show stronger correlations than microlevel studies. Maybe this is due to reduced measurement error. Also--official statistics capture much more serious behavior. Self-report data is swamped with minor misbehaviors, which smart kids are frequently guilty of. Keep in mind, too, that my analysis is of whites only. Including all races might strengthen the correlations.


  1. Cool. Is a Biserial correlation appropriate here? One of the variables is binary right?

  2. No, delinquency items have 5 categories, and the IQ measure has many categories.


You don't see me kissing up to my 100% German boss by telling her stories of my single German ancestor from 19th century Pennsylvania

Carl Zimmer and Razib Khan have made a big deal on Twitter about the inaccuracy of the claim that Elizabeth Warren is no more Indian than th...