« HONEST SIGNALS | Gene Expression Front Page | Behold, and the Lord God Muddled the Mind of Man! »
September 04, 2003

SES and IQ Heritability

Earlier godless posted about the fascinating new study which appears to show that IQ heritability varies significantly with socioeconomic status (SES). Here are my thoughts...

Back-to-school pop quiz:  Why do poor children, and especially black poor children, score lower on average than their middle-class and white counterparts on IQ tests and other measures of cognitive performance?

That's the lead question in a Washington Post article about a new study by researchers at the University of Virginia.  This study appears to show that IQ heritability varies significantly with socioeconomic status (SES):

Until recently, [lead researcher Eric] Turkheimer and others said, research had indicated that the "heritability" of IQ - that is, the degree to which genes can explain the differences in IQ scores - completely dominated environmental influences.

But it turned out that virtually all those studies on the heritability of IQ had been done on middle-class and wealthy families.  Only when Turkheimer tested that assumption in a population of poor and mostly black children did it become clear that, in fact, the influence of genes on IQ was significantly lower in conditions of poverty, where environmental deficits overwhelm genetic potential.

Specifically, the heritability of IQ at the low end of the wealth spectrum was just 0.10 on a scale of zero to one, while it was 0.72 for families of high socioeconomic status.

The study itself used 320 pairs of twins.  Twin studies are great for this kind of research, because comparing the correlation of IQ between identical twins, which share environment and genes, with fraternal twins, which share environment but not genes, allows the degree of heritability to be accurately determined.

This would be a very important finding if true - and would go a long way toward explaining the surprisingly low average IQs of many third-world countries (see IQ and Populations for more).  It would also give hope to those who feel improving living conditions in poor countries would enable them to become competitive in the global workforce.

However, it is worth pointing out that this study contradicts earlier studies looking for the same thing.  The WP article mentions Robert Plomin, a behavioral geneticist with the Institute of Psychiatry at King's College of London, who has been seeking genes linked specifically to intelligence.  Plomin said his own unpublished work involving 4,000 pairs of twins has not produced the same results as Turkheimer's.  Similarly, the National Longitudinal Survey of Youth study famously used as the basis for many of the conclusions drawn in Richard Herrnstein's and Charles Murray's classic book The Bell Curve did not find these correlations, despite specific efforts to correlate SES with IQ.

The U of V study does differ from other work in one important aspect.  Instead of seeking correlation between SES and IQ directly, the researchers were seeking correlation between SES and the heritability of IQ.  Why is this different?

Well, if SES and IQ correlate, and IQ is substantially heritable, then it implies that poor populations are stuck in a sort of vicious circle.  Their poorness implies low average IQ, and their low average IQ implies a low average IQ among their children, which in turn implies their children will be poor.  (That's an over-simplification, but first-order this is the result.)  That's a pretty tough circle to exit, especially if the poor populations also have a higher-than-average birthrate.

On the other hand, if SES and heritability of IQ correlate, then in poor populations IQ is not significantly heritable (the heritability figure of 0.10 means essentially there is no correlation from one generation to the next).  This would break the vicious circle.  Poor populations might indeed have a low average IQ, but their children need not if their socioeconomic conditions are improved.  This conclusion supports efforts such as Project Head Start, which attempts to improve the lot of poor young children by giving them food, books, and exposure to positive learning environments.

Despite this hopeful conclusion, it should be noted that studies which have attempted to validate the effect of Project Head Start have invariably suggested it is not helpful.  But again, these studies have measured correlation to IQ, not correlation to heritability of IQ.

As the article indicates, this research suggests a fruitful avenue for future study:

The next big challenge is to find out what it is about socioeconomic status - a measure that includes not only income but also parental education and occupational status - that contributes to [heritability of ] IQ, so social programs can more effectively boost those factors.

The WP article is balanced and well worth a read.  I'm going to try to get the study itself to learn more...


Posted by ole at 09:45 AM

"The U of V study does differ from other work in one important aspect. Instead of seeking correlation between SES and IQ directly, the researchers were seeking correlation between SES and the heritability of IQ."

You get to the heart of the matter. Very astute of you, I must say.

Posted by: Hardbitten Sceptic at September 4, 2003 11:59 AM

Nobody's seen the actual study yet, just the WaPo article by a general purpose reporter who clearly is ignorant of the history of IQ analyses, as I point out here:


John Jay Ray points out here that "restriction of range" is the most likely explanation:


Posted by: Steve Sailer at September 4, 2003 04:25 PM

It is essential to rule out range restriction. If there is little variance of IQ in the low income subsample, there is little variance for any indendent variable or variables to explain. If range restriction is a problem in this analysis, then it is likely that a sample larger than 320 would be needed for this type of study.

Posted by: StateUniversityProf at September 5, 2003 11:39 AM

May be worth noting that in Teasdale and Owen's long-term studies of the 'Flynn Effect' among Danish army recruits, the rise in IQ over the last 40 years has been much more notable among the low-IQ than the high-IQ part of the distribution. This may suggest that environmental factors are more important in depressing scores at the low end than in raising scores at the high end - which I guess is consistent with higher heritability at the high end. But I'm no expert, so don't shoot me if I'm wrong.

Posted by: David B at September 5, 2003 11:51 AM

The idea makes a certain amount of sense. For example you can assume that high-SES people are more consistent in creating a good environment, by avoiding things like lead poisoning for their kids, smoking while pregnant, and feeding their infants formula. All of those things can also be done by the low-SES cohort, but they would probably be less consistent and so create a much more variable prenatal/early postnatal environment...

Posted by: bbartlog at September 10, 2003 06:31 PM