A fascism of statistics

Last night, in a lengthy programme on Channel 4, Rageh Omaar explored the highly controversial issue of whether there are innate differences between the levels of intelligence of different races. In each thread of his investigation, Omaar allowed the proponents of the most provocative theories to set out their case, before seeking to demolish their arguments. From my own perspective, he did so extremely convincingly, by showing that the genetic differences between races are too minimal to produce the dramatic differentials in intelligence that have been suggested, and by pointing to the 'Flynn Effect', which could imply that IQ tests are less of a direct measure of intelligence, and more a measure of an individual or group's "adaptation to modernity". However, as the programme progressed, I could just imagine the self-anointed experts on the field of IQ (many of them statisticians) saying "this is proving nothing, he's missed the point entirely, anyone who thinks that IQ tests do not directly and accurately measure intelligence simply does not understand statistics".

And that, I think, is the essence of the whole problem with the debate over the validity of IQ tests, regardless of whether it relates to race, or merely the 'classification' of individuals by intelligence. There's a kind of 'statistical fascism' at play, whereby the statisticians refuse to seriously engage with any of the powerful counter-arguments to the prevailing wisdom that IQ tests are meaningful measures of intelligence - unless it is done on their own terms, with detailed reference to 'meta-analyses' and 'regressions' and all sorts of other incomprehensible language. Given the huge importance of this subject to everyone, it simply can't be right that authoritative discussion of it is restricted to such a narrow area. For one thing, anyone who has ever taken an IQ test (I've taken the American SAT, which in its old form closely resembled an IQ test) won't need a grounding in statistical theory to intuitively understand that it simply can't be a pure measure of intelligence - your score will also be significantly affected by your level of motivation to do well, your composure under severe pressure of time, etc, etc. The statisticians would respond by pointing out that, if an individual takes several different IQ tests, the outcome will typically be remarkably close every time. But this doesn't really address the point - if someone's composure and motivation levels remain fairly constant every time they take the test, it follows that the scores would remain constant as well. This constancy does not constitute proof that IQ tests accurately measure intelligence.

Ah, the statisticians might respond, you're overlooking the high level of correlation between IQ scores and success in life, as measured by educational attainment and income levels. But doesn't an individual's composure under pressure and motivation to succeed also play a significant role in determining their life chances? If an IQ test is partly measuring those things, it's hardly surprising there would be such a correlation.

In fact the buzz phrase of statisticians in this field appears to be "intelligence tests are the most accurate of all psychological tests". Just goes to show how ropey all the others must be.