Psychology professor Eric Turkheimer is remarkably inconsistent on a variety of topics in his field of work and he’s prone to unscientific, idle speculation on a whim. In the past he has acknowledged that group differences should not be thought of differently than individual differences; he has since reneged on that. In the past he has applied his pet hypothesis, the Scarr-Rowe effect, to suggest that race differences in intelligence are environmental rather than genetic; he has now reneged that.
Eric recently tweeted a highlight from Turkheimer, Harden, D’Onofrio & Gottesman (2009). In this tweet he highlighted the section
The hypothesis that the heritability of cognitive ability might vary with socioeconomic status (SES) was first forwarded by Sandra Scarr in a 1971 paper published in Science.
in order to justify his tweeted statement that “[The study of the Scarr-Rowe effect] was always about SES, not race.” In his quoted article, he goes on to say about that 1971 paper
[H]eritabilities were found to be consistently higher for Whites than for Blacks, and for higher compared to lower socioeconomic groups, both across races and between them.
This idea, that variance components might differ by level of SES, and thus by race if races differ in their level of SES, has been discussed ad infinitum in the literature. One of the first references comes from Jensen (1968), who created these figures:
He also wrote
One way of testing the hypothesis that a particular segment of the population is intellectually handicapped because of its position on the environmental continuum would be to carry out a heritability study within this segment of the population. If the hypothesis represented by Figure 2 has any merit, heritability estimates should be significantly lower for groups reared in the more disadvantaged part of the environmental continuum. Here, then, is one feasible means of directly testing the hypothesis that Negroes perform below most other groups on tests of intelligence and scholastic achievement because of environmental rather than genetic differences.
In his tweets, Eric wrote
Sure you could say the hypothesis “implies” that there would be a racial difference in heritability, just like it “implies” that there would be a h2 difference between Connecticut and New Mexico. Race is the hbd obsession, not mine.
There is no reason to expect a race difference, other than the extent to which race is a poor marker for SES. And it is a poor marker. Using data at hand from Louisville Twin Study, race and SES are correlated r=.37, R2=.13. Blacks in twin studies are often not all that poor.
So in a field in which (I admit) most analyses are already badly underpowered, you are using an indicator of the main construct with a validity coefficient of .37. So it’s a pointless analysis of a pointless hypothesis.
Race is something which Eric has written about from the very beginning of his career (purposefully, from 1990 on). He is not unfamiliar with the concept, nor is he unwilling to bandy about the Scarr-Rowe effect as relevant to race, as he writes in his anti-Murray piece in Vox
The heritability of intelligence, although never zero, is markedly lower among American children raised in poverty. Several interpretations of this fact are possible. The one we find most persuasive is that children raised in those circumstances are unable to take full advantage of their genetic potential because they do not have access to the high-quality environments that could support it.
Indeed, directly stating its relevance to race, Scarr (1971), whom Turkheimer cites, writes
To the extent that the same environmental factors are assumed to affect the development of IQ in the same way in both black and white populations,
predictions can be made about the sources of racial differences in mean IQ scores. If certain biological deprivations (such as low weight at birth, poor nutrition) are known to be more prevalent in lower class groups of both populations and more prevalent among blacks than whites, then the two models can make differential predictions about the effects of these sources of environmental variance on the proportion of genetic variance in each population. Given a larger proportion of disadvantaged children within the black group, the environmental disadvantage hypothesis must predict smaller proportions of genetic variance to account for differences in phenotypic IQ among blacks than among whites, as whole populations. Since the genotype distribution hypothesis predicts no differences in the proportion of genetic variance for social class groups within the races, it should predict the same proportions of genetic variance in the two races…. Environmental disadvantage is predicated to reduce the the genotype-phenotype correlation in lower-class groups and in the black group as a whole.
So from the very beginning, at least, what Turkheimer acknowledges as the Scarr-Rowe hypothesis’ origin was concerned with racial differences. Based on his usage of it in various articles which he has either authored or coauthored, including the Vox piece, as something to contradict genetic racial differences, he clearly knows this to be the case. Despite selling his famous 2003 paper as evidence for this proposition, he has acknowledged (via personal communications with other researchers) that there were no racial differences in the heritability of IQ in that study. It’s abundantly clear then that he understood these things and implicitly — despite it being something he knew was untrue — laid out what can be called Turkheimer’s syllogism:
- Races differ in mean socioeconomic status;
- Social classes differ in mean heritability, thus;
- Races differ in mean heritability.
And yet, he has the audacity to write that this is not implied by the Scarr-Rowe hypothesis, citing the article which he has repeatedly claimed originated it — and which disagrees with him — as proof. The proof becomes even sillier when we read deeper into Turkheimer’s own 2009 article which he tweeted from, wherein he contradicts himself by writing
[T]here is also a need to return to the clear theoretical focus that Scarr brought to her early work on this subject in 1971. Now that software is readily available, it would be possible to re-analyze practically every twin analysis that has ever been conducted, with the familiar variance components moderated by socioeconomic status, or by age or gender or race.
Turkheimer, in his unending quest to obfuscate the study of the Scarr-Rowe effect as it pertains to race differences, also remarks that blacks in twin studies are not all that poor, with a correlation (presumably point-biserial) between race and SES of 0.37 for his Louisville Twin Study. Correlations translate to effect sizes, and in this case, his r of 0.37 is a Cohen’s d of 0.80. This is somewhat smaller than the difference in four major cohorts assessed between 1988 and 2009 which have been put into the table below
For these studies, “race and SES are correlated at” r = 0.31, taking the simple average d (0.6515) and converting it to a value of r. This is remarkably similar to the correlation between academic achievement (basically a measure of IQ) and socioeconomic status:
If Eric’s remark that race is a weak indicator of socioeconomic status is true, and considered relevant, we might also question the relationship between IQ or other achievement measures and socioeconomic status generally, since it is comparable to the correlation he gave. But if we want to be more serious, we can look at the correlation between ancestry (i.e., genetic race) and socioeconomic status. Two recent studies have reported this exact thing. In the former, the relationship between a socioeconomic status measure made from “household income, guardian 1 educational level, guardian 2 educational level, guardian 1 occupational level, and guardian 2 occupational level” and African ancestry was -0.30; in the latter , a measure made from parental education, the most important parental socioeconomic variable for predicting children’s IQ, correlated at -0.41 with African ancestry. The relationships between socioeconomic status and IQ in these studies were 0.42 and 0.41 respectively. These numbers are remarkably similar to Eric’s and they come with racial group differences of the average size (~1 d). It’s hard to say these are minor or insignificant, and yet they’re of the same magnitude as the general IQ-socioeconomic status correlation, which he apparently thinks the Scarr-Rowe is just about.
Eric’s game is easy to play. In Scarr’s original study, which, as Eric tells us, was about socioeconomic status and certainly not race, and which constitutes, to him, the first test of the Scarr-Rowe effect, socioeconomic status differences were established by trichotomizing socioeconomic status within each group. The differences in verbal, nonverbal, and total scores between blacks and whites were 0.79, 0.74, and 0.87 d in total; the differences within each group between social classes (lowest and highest scores used) were 0.33, 0.20, and 0.28 for blacks and 0.69, 0.70, and and 0.86 for whites. In this sense, the predictions of the Scarr-Rowe hypothesis, as Jensen (1968) would have stated them, turned out to be true: blacks were relatively constrained in variance compared to whites, consistent with their worse average environments, as measured by their SES. Moreover, the differences between the top and bottom SES brackets within the black group were smaller than the difference between the black and the white groups in total, for each category. The SES index used in the study divided the highest segment of the black group from the lowest segment less so than it did the black from the white group. If SES is the relevant locus for investigation, then it’s hard to say why race shouldn’t be when it’s associated with similarly sized IQ and socioeconomic differences.
If the Scarr-Rowe hypothesis is correct, then why, when the average socioeconomic differences between the top and bottom thirds of the socioeconomic distribution are around the size of the race difference in socioeconomic status and the heritabilities vary within each race accordingly, are there no differences in heritability by race? This is a quandary, an anomaly, and castigating researchers for noting it is wholly unscientific and mean-spirited.
It’s obvious what Eric is doing is not being consistent. Contrary to what he tweeted, race differences are an excellent vantage point for assessing the Scarr-Rowe hypothesis and this was part of the original intent of the hypothesis — a possible explanation for social class and thus race differences — either way. The idea that there is no connection between the two is not only just tenuous, it is wrong, and both his work (especially in 1990 and 1991) and his popular statements (which are becoming increasingly common) prove that. He should take a note from Scarr (1971), a study he might not have ever read:
Dislike of a genetic hypothesis to account for racial differences in mean IQ scores does not equal disproof of that hypothesis. Evidence for genetic or environmental hypotheses must come from a critical examination of both explanations, with data that support one.
