In my last post on the American Enterprise Institute, I lumped Charles Murray in with James Glassman, Karl Zinsmeister and Lynne Cheney, as someone who had contributed to the loss of the AEI’s reputation for scholarship (John Lott, the main subject of the post is in a class of his own). Regular reader Jack Strocchi objected, and I thought that this would be a good time to set out my views on Murray, and more particularly on The Bell Curve with Richard Herrnstein, which is, I think, his most significant work.

The Bell Curve got a thorough hammering on statistical grounds when it came out (this Heckman review is one of the more favorable but is still pretty damning. But the thing that most annoyed me when I read it was their discussion of the Flynn effect, namely that average scores on IQ tests have risen steadily over time, by amounts sufficient to wipe out the differences between racial groups on which Murray and Herrnstein rely. As Thomas Sowell points out in this review (reproduced by Brad de Long), it’s hard to see how any claim that differences in IQ test scores observed in Western societies are mostly due to genetic factors can stand up in the face of this observation. But Murray and Herrnstein slide straight past it, saying that they are concerned with contemporary inequality not with time trends. This is about as reasonable as a “nurturist” deciding to ignore twin studies on the grounds that most people aren’t twins.

Update Marginal revolution and Aranda Blog have more on the Flynn effect. And Brad de Long says “pretty damning” is too kind a description of Heckman’s review. He prefers “the impeccably right-wing Jim Heckman flayed Murry and Herrnstein alive and hung their skins on his office door”.

I think there are two reasonable interpretations of the Flynn effect. One is to agree with Murray and Herrnstein that IQ test scores measure “intelligence” and to conclude that intelligence is primarily determined by environmental factors, and particularly by education.

The second is that IQ tests scores do not, in fact, correspond to what we normally mean by intelligence but to something narrower, and more environmentally malleable. The simplest version of this would be “test sophistication”, the fact that people who have done IQ tests in the past do better. The problem is that the administration of IQ tests to kids was more common 50 years ago than today.

A more plausible claim is that IQ tests are essentially academic achievement/aptitude tests. They are ‘content-free’ in that they don’t rely on information from specific courses in the manner of school assessments, but they do test skills that are promoted by school education and by an environment where most people have high (by historical standards) levels of education. They are good predictors of university performance (thanks for this link, Jack). Since academic aptitude naturally rises with education we would expect to get the obvious virtuous circle – the more schooling there is, the higher is IQ and the more kids are capable of benefitting from a university education.

While it’s implausible that the population is massively more intelligent than it was a couple of generations ago, it’s not unreasonable to suggest that capacity for academic achievement has been rising steadily. Fifty years ago, someone who suggested that the majority of people were capable of getting a university degree would have been regarded as utopian. And despite the dumbing down of the last decade or so, the content of the average degree is probably tougher now than fifty years ago.

On the other hand, I imagine that, if there were tests that depended, even subtly, on familiarity with agriculture, scores would have plummeted over the same period.

I’m happy with the idea that intelligence is more than IQ. But of course, the evidence on which Murray and Herrnstein rely is all about IQ scores. It shows, about as unambiguously as is possible, that changes in the environment can substantially increase the average IQ score of a population with a broadly constant genetic endowment.