More on IQ and crime: In the earlier post, Tex wrote in the comments section that IQ-crime correlation is curvilinear: really smart people and the mentally retarded are the least criminal. Here's the percent ever arrested for each number out of ten answered correctly on the GSS vocab test (for whites only):
Percent ever arrested by vocabulary score
We don't see much of drop in arrest among the dullest group (I imagine really slow people were not interviewed) but the smartest group is noticeable less likely to have ever been booked and fingerprinted. What happens if we estimate a logistic regression on the modal category--a score of 5--through the top group? This eliminates the curvilinearity problem (although it does restrict the range). The estimated coefficient is .085, with a p-value of .012. But IQ still explains less than 1% of the variation in arrest. Although it violates assumptions to use Pearson correlations, researchers do it all the time, so for people more familiar with that statistic, it is .04 in this analysis, and also .04 for the full range of IQ. I'm not impressed.