I was at the beach with my crumbling copy of Arthur Jensen’s 1980 landmark book Bias in Mental Testing when a sentence on page 493 made me smile:

College GPA, in contrast to final grades in armed forces training schools, is probably almost as highly loaded on level II ability (after correction for attenuation) as are the predictive tests

Level II ability is mental ability that loads on g (the general factor of intelligence), such as abstract reasoning, conceptual ability etc, as opposed to Level I ability which is stuff like rote memory.  Jensen believed that the working class and certain minorities were normal in level I ability but below average in Level II ability.

I have found that he SAT has a g loading of around 0.7 in the general U.S. population, and so Jensen’s comments implies college GPA has a g loading of almost 0.7 (corrected for attenuation) if we could drag all American young people into college, and the same college at that.

But what on Earth does he mean by corrected for attenuation?  I would expect college grades to be more reliable than SAT scores given that they are based on many tests and assignments, not just one.

But perhaps the attenuation comes from the bias of the professors themselves.  Just as different supreme court judges can reach different decisions about a case, different law professors will give the same essay an A or C depending on whether they agree with the legal arguments.  And similarly for different English professors, depending on whether they agree with your literary tastes. So it’s this bias in scoring that largely makes college GPA a worse measure of intelligence than the college admission tests themselves.

One reason math and computer types tend to have higher IQs than the rest of us is not just that these subjects are more g loaded, but they’re also more objective.  The computer program either runs or it doesn’t.  Less room for bias on the professor’s part.  Jealous of their ability, other students call computer geeks “autistic”, because unlike math or computers, you can seldom be proven wrong about your social intelligence because you can never get into someone else’s head and prove you correctly read their mind, so social IQ becomes an easy way for people to convince themselves that they’re smarter than computer geeks at something.

Back to Jensen’s quote: why did reading this make me smile?  I’ve long been fascinated by the idea that smart people get to the top but I feared this was just a self-fulfilling prophecy of the testocracy we live in and says little about the intrinsic predictive power of the IQ itself.

But the fact that GPA is almost as g loaded as predictive tests means that even if we removed the testocracy (college admission tests), it would have little effect on who gets to the top in America, because the very act of learning the skills you need to function in elite occupations is itself almost as g loaded.  The predictive power of IQ is not something society artificially created.