When newspapers or generally mainstream media talk IQ, which occasionally they feel compelled to do, the reporting is often less than complete. You might term it desperate if you don’t want to call it dishonest. After all this is a topic where just mentioning an unequivocal fact might have very negative consequences on your career as a journalist.
In this post I want to point out several types of „lies by omission“ that one should look out for in these circumstances.
Trick number one: Compare childhood IQ, not adult IQ.
Childhood IQ is much more malleable than adult IQ, so this is a way to circumvent the fact that in the long term IQ has almost zero shared environment effect . This trick can be used to show the efficacy of interventions (just don’t ever follow up on the kids) or to prove that disparities in abilities are influenced by some environmental difference. But it goes further than that. Some ethnic disparities in IQ grow with age. For Arab kids this is called the „Simber effect“ . In some studies this effect also shows up for the black-white gap. So by looking only at children, you can effectively hide a big part of an IQ gap that explains disparities in income and other life outcomes.
Trick number two: Ignore the g-factor.
The g-factor is the part of IQ that is both the heritable part and the predictive part. The rest of IQ is basically a measurement problem, which among other things, leads to the Flynn effect. The Flynn effect, the rise of IQ scores observed in many countries in the last century, does not increase g. The increases for the different subtests are anti-correlated with how predictive they are of the g-factor. By omitting these little facts, the Flynn effect can be used to invalidate all kind of observations about heritability and predictiveness of IQ and of the relative permanence of IQ gaps between groups.
Trick number three: Don’t mention the sample size or other sample attributes.
Social science is full of small scale studies that do not replicate. Especially small scale studies with politically expedient results. So there is always something to cite, to prove your point of view. Here is the rub, the main IQ results are extremely robust and often have massive sample sizes. The gap between white and black Americans for example is estimated by a meta study with N=6.000.000!  So studies that „refute“ such a result with sample sizes of a few hundred or maybe even just a few dozens should sensibly dismissed out of hand. Especially because these small studies often suffer from egregious sampling problems.
 Wilson effect
 Simber effect
 Ethnic differences