One of the most fascinating correlates of IQ is conservatism.  I find this correlate fascinating because IQ tests justify anti-black racism (a conservative value) by showing that dark skinned people have lower IQs (on average) yet at the same time, IQ tests condemn racism by showing that anti-black racists also have low IQs.  So one can’t use IQ tests as a racist weapon without impugning their own likely IQ.

The other reason conservatism is such a fascinating correlate of IQ is that it’s one of the few variables that correlates positively with IQ between groups, but negatively with IQ within groups. That is, higher IQ races tend to be more conservative than low IQ races, yet within each race, I’ve argued that higher IQ people tend to be less conservative.  For someone who likes to statistically predict IQs from biodemographic variables, this is an incredibly useful property.

Why?  Because one of the biggest problems with predictive statistical models is multicollinearity. In other words, when trying to guess a person’s IQ based on many variables, you run into the problem of the variables being positively intercorrelated and thus not adding new information.  For example, if I know you’re homeless, I can guess with 90% certainty that  have a low IQ because 90% of America’s homeless are below 100 IQ (white norms).  Now if I also know you’re also high school dropout, the odds of you having a low IQ increase (because you now belong to two low IQ groups), but the odds don’t increase as much you’d think, because the mere fact that you’re homeless already suggests you’re probably uneducated.

On the other hand if I know you’re black, I might guess your IQ is below 100, but I can’t be certain, because millions of blacks have high IQs, but if I know you’re also conservative, the odds of you having a low IQ increase precipitously, because conservatism is not something I could have guessed from your blackness.  Just the opposite: blacks tend to be liberal, so conservatism predicts low IQ completely independently of blackness.  Statistically, one would need an especially low IQ to belong to belong to the two low IQ groups that have almost zero overlap.