Thank you so much for your support! We met our yearly fundraising goal within 12 hours of yesterday’s post. We look forward to finishing and publishing these analyses.
In Lasker, Pesta, Fuerst, and Kirkegaard (2019), we found an unstandardized beta for European genetic ancestry, when predicting g, of .85 among African Americans (model 2; Table 6). Simply put: a 100% increase in European (vs. African) ancestry was associated with a 0.85 d increase in intelligence. We interpreted these results as strong support for a partial hereditarian model. As did others in the HBD sphere.
Bird (2021a), in contrast, argued that our regression analyses suffered from omitted variable bias. Notably, he did not disagree that the results would support a hereditarian model were they robust.
Given the 2.053 d (or 30.8 point) measured test score difference between continental Africans and Europeans which Bird (2021a) adopts, genetic effects alone, based on our results, would represent .85 d /2.053 d = 41% of the phenotypic difference. Expressed in terms of variance explained, this would be (.85 d)^2/(2.053 d)^2 = 17.14%.  However, this is relative to an average within-groups heritability for g of 66.5% for this specific sample (Mollon et al., 2018; Pesta, Kirkegaard, te Nijenhuis, Lasker, & Fuerst, 2020). Since the expected differences are proportionate to the within-groups heritability, the variance explained would be predicted to be around 17.14%/66.5%*50% = 12.88% conditioned on a heritability of 50%.
Now, based on his analysis of SNP data, Bird (2021a) estimated a variance explained of 12% given a heritability of 50%. Thus, these two very different methodologies (global admixture analysis & SNPS Fst comparisons) derive very similar estimates conditioned on the same heritability coefficient.
But Bird (2021) goes on to interpret his result as “no support for a hereditarian hypothesis”. Well, one could define a ‘hereditarian hypothesis’ such that these magnitudes do not support it. But, in that case, one could just cite our own widely discussed research results against it. In this case, Bird (2021b) should then also state that, “Lasker et al. (2019) also found ‘no support for the hereditarian hypothesis of the Black–White achievement gap’ and, in fact, Fuerst is strongly supportive of an environmental model, despite what some disreputable sites claim.”
I won’t complain. I am sure that being labeled an environmentalist will not hurt my career prospects. However, don’t call me a hereditarian for arguing X but then go on to argue X and also call that ‘no support for a hereditarian hypothesis’.
 To convert between variance metrics, such as R^2, and linear metrics such as r, you take the square-root of the former or the square of the latter. The difference between variance and linear metrics can lead to misinterpretations, since variance metrics do not align with our intuitive sense of distance. Phil Birnbaum (2006) gives the following example: if you were playing baseball and you made it to second base, you might think you made it 2/4 = .5 or one-half of the way home, but in terms of variance metrics you really only ran 2^2/4^2 = 4/16 =.25 or one-quarter of the total variance to home base. This is why, in context to the continental African and European differences discussed, a between-group variance of 17.14% is equivalent to a real-world percent explained of sqrt(17.14%) = 41%.
 Originally, I reported an average heritability for g in the TCP sample of 81.5; the correct value was 66.5 (White = 72%; Black = 61%). The text has been updated.
 As for which estimates to use, a point which Bird (2021b) touches on, ideally one would employ both within-groups broad-sense heritability and total genetic variance between populations so to calculate the broad-sense between-group heritability and the total expected differences. This is insofar as one is interested in the overall differences, not predicting offspring values from parental ones or testing specific evolutionary models. Now Bird (2021a) cites Polderman et al. (2015). For adults (age 16 to 65), Polderman et al. (2015) give meta-analytic MZ and DZ correlations of .68 and .28 (Figure 3; High-level cognitive functioning), which, using Falconer’s formula, yields a meta-analytic broad-sense heritability of 80%.
Of this, most of the variance is additive genetic; nearly all the remainder is due to an unknown mix of active gene-environmental covariance and dominance variance. Now, if for methodological or theoretical reasons, one uses within-groups narrow-sense heritability and additive genetic variance between populations, one simply derives the expected differences due to additive genetic differences. That can be useful for certain purposes, however, it will underestimate total genetic differences (unless, unexpectedly, in this case, the genetic variance components go in discordant directions between populations). Regardless, since global admixture results will relate to broad-sense heritability, one needs to adjust the heritability when comparing the results of Bird (2021) to those of Lasker et al. (2019).
Bird, K. A. (2021a). No support for the hereditarian hypothesis of the Black–White achievement gap using polygenic scores and tests for divergent selection. American Journal of Physical Anthropology.
Bird, K. A. (2021b, February 12). Still No Support For Hereditarianism. Accessed at: https://kevinabird.github.io/
Lasker, J., Pesta, B. J., Fuerst, J. G., & Kirkegaard, E. O. (2019). Global ancestry and cognitive ability. Psych, 1(1), 431-459.
Mollon, J., Knowles, E. E., Mathias, S. R., Gur, R., Peralta, J. M., Weiner, D. J., … & Glahn, D. C. (2018). Genetic influence on cognitive development between childhood and adulthood. Molecular psychiatry, 1-10.
Pesta, B. J., Kirkegaard, E. O., te Nijenhuis, J., Lasker, J., & Fuerst, J. G. (2020). Racial and ethnic group differences in the heritability of intelligence: A systematic review and meta-analysis. Intelligence, 78, 101408.
Polderman, T. J., Benyamin, B., De Leeuw, C. A., Sullivan, P. F., Van Bochoven, A., Visscher, P. M., & Posthuma, D. (2015). Meta-analysis of the heritability of human traits based on fifty years of twin studies. Nature genetics, 47(7), 702-709.
Which means that the Human Phenome Diversity Foundation’s (HPDF) annual fundraising drive has commenced.
Our goal is $2,500.
We have some great projects which we would like to support this year if we can afford to.
Last year’s fundraising helped finance an important admixture paper, currently under review, which is up at biorxiv.
We would like to continue to fund genetically informed research with your support.
If you care to see this research done, you can donate at the HPDF’s official gofundme charity site. Donations are tax-deductible since the HPDF is a 501(c)(3) organization.
Also, the HPDF now has an associated corporate Kraken account, so you can donate directly with cryptocurrencies, too:
Kevin Bird has a paper out in which he claims, more or less, to evidence “insignificant” race differences. There is a lot there to criticize: misinterpretations, odd analytic choices, a crucial wrong formula , etc.
Maybe I will write a formal critique.
For now, it’s sufficient to point out that the results strongly agree with a hereditarian model:
- The predicted differences, given the genetic divergence in the educational and intelligence SNPs, are medium to large given reasonable estimates of broad-sense heritability (H2).
- While there is inconsistent evidence of divergent selection (for this pairwise comparison), there is zero evidence of homogenizing or stabilizing selection.
To illustrate point (1), Table 1 shows the expected BGH given the 30.8 point continental European-African difference which Bird adopts along with the expected phenotypic gaps when environments are equal (i.e., when BGH is set to unity). I use the lowest Fst value Bird reports in his table. Proofs are provided for the formulas used in the .doc file.
Table 1. Expected Between Group Heritability (BGH) and Expected IQ Point Differences between Europeans and Africans Given Different Values of the Genetic Intraclass Correlation (r and r_c) and H2 assuming an eduSNP Fst =.111 from Bird (2021; Table 1; 1301 clumped EA SNPs)
|H2||r||t_observed||BGH||t_expected||Expected IQ difference||Cohen’s Interpretation|
|H2||r_c||t_observed||BGH||t_expected||Expected IQ difference||Cohen’s Interpretation|
Note: H2 = broad-sense heritability; r = intraclass genetic correlation; r_c = intraclass genetic correlation corrected for mathematical constraints on Fst; t_observed = intraclass phenotypic correlation i.e., phenotypic variance between groups (given d = 2.053); BGH = between group heritability; t_expeced = expected phenotypic variance between groups when environments are equalized; Expected IQ difference = expected IQ differences when environments are equalized; Cohen’s Interpretation = conventional interpretation of effects sizes.
You can argue that one should use narrow-sense heritability, instead of broad-sense, contra Jensen (1972; 1998), then lowball the estimates, and finally take advantage of statistical illiteracy and portray the differences as ‘small’ or ‘insignificant’ by emphasizing the portion of variance explained. However, the expected differences (which are equal to sqrt(BGH) x observed phenotypic differences) are still medium to large. Of course, Bird (2021) argues that the differences could go either way with equal likelihood. This would be true if you knew nothing else. However, in his prior analyses, he uses polygenic score (PGS) weights, and the eduPGS weights are directional. For the same set of eduSNPs the PGS differences are shown below:
Table 2. Mean MTAG-based PGS for CEU and YRI Calculated using population-GWAS and Within Family Betas.
|W/ population-GWAS||W/ Within Family Betas|
|CEU (N = 99)||YRI (N = 108)||CEU (N = 99)||YRI (N = 108)|
|p-value (Welch’s Two Sample t-test)||< 0.0001||< 0.0001|
|p-value (Welch’s Two Sample t-test)||< 0.0001||< 0.0001|
|p-value (Welch’s Two Sample t-test)||< 0.0001||< 0.0001|
Note: SNPs were filtered for MAF >0.01 for both CEU and YRI. Scores represent standard scores calculated using the standard deviation in the total sample. Sample sizes for the t-test were N = 99 for CEU and N=108 for YRI.
Thus, it makes no sense to say that the expected difference could go either way, with equal probability, when the eduPGS weights indicate a direction. When this is recognized, the only option is to declare that the eduPGS is biased between populations. This is possible, of course.
However, this leaves the evolutionary default or null, which is that differences will be commensurate with neutral variation. As Rosenberg, Edge, Pritchard, & Feldman (2019) note: “One key component of the inference of polygenic adaption is the use of an appropriate null expectation for polygenic scores distributions and phenotypic differences…[P]henotypic differences among populations are predicted under neutrality to be similar in magnitude to typical genetic differences among populations.” The authors, of course, go on to cite Lewontin and slyly note that differences “are small in comparison with variation within populations”. But, of course, “large” differences in the conventional sense are also “small in comparison with variation within” (e.g., .80d = 14% variance). And while the evolutionary default is directionless, the totality of the behavioral genetic and psychometric data assembled on this topic points one way.
 See, for example, equation 4 in Bird (2021). However,
total between phenotypic variance = phenotypic variance due to genes + phenotypic variance due to environment
which can be rewritten, in linear metrics, as PD^2 = GD^2 + ED^2 or PD = sqrt( GD^2 + ED^2 )
Since, BGH = phenotypic variance due to genes / total between phenotypic variance
BGH = GD^2 / PD^2 and, therefore, GD = sqrt(BGH)*PD
This is approximated but underestimated by 2*PD^2 * sqrt(2/pi) (*15) which Bird (2021) uses.
e.g., sqrt(.12)*30.8 = 10.67 (correct) versus 2*sqrt(.12)*sqrt(2/pi) (*15) = 8.29 (Bird, 2021)
 While the use of narrow-sense heritability is recommended for Qst-Fst comparisons and the assessment of directional selection, narrow-sense heritability, and the corresponding narrow-sense Qst underestimates between-group genetic variance by not taking into account non-additive genetic variation between populations, along with active gene-environment covariance (which is commonly classed as a genetic effect; Sesardic, 2005). Thus when it comes to calculating the expected difference due to genes, it makes sense to use the broad-sense heritability, at least for an upper-bounds estimate, as hereditarians have done (e.g., Jensen, 1998).
In 1969, Harvard Educational Review published a long, 122-page article under the title “How Much Can We Boost IQ and Scholastic Achievement?” It was authored by Arthur R. Jensen (1923–2012), a professor of educational psychology at the University of California, Berkeley. The article offered an overview of the measurement and determinants of cognitive ability and its relation to academic achievement, as well as a largely negative assessment of attempts to ameliorate intellectual and educational deficiencies through preschool and compensatory education programs. Jensen also made some suggestions on how to change educational systems to better accommodate students with disparate levels of ability.
While most of the article did not deal with race, Jensen did argue that it was “a not unreasonable hypothesis” that genetic differences between whites and blacks were an important cause of IQ and achievement gaps between the two races. This set off a huge academic controversy—Google Scholar says that the article was cited more than 1,200 times in the decade after its publication and almost 5,400 times by December 2019. The dispute about the article centered on the question of racial differences, which is understandable as Jensen’s thesis came out on the heels of the civil rights movement and its attendant controversies, such as school integration, busing of students, and affirmative action. Jensen questioned whether it is in fact possible to eliminate racial differences in socially valued outcomes through conventional policy measures, striking at the foundational assumption of liberal and radical racial politics. His floating of the racial-genetic hypothesis was what set his argument apart from the general tenor of the era’s scholarly and policy debate.
In this post, I will take a look at Jensen’s arguments and their development over time. The focus will be on the race question, but many related, more general topics will be discussed as well. The post has four parts. The first is a synopsis of Jensen’s argument as it was presented in the 1969 article. The second part offers an updated restatement of Jensen’s model of race and intelligence, while in the third part I argue, using the Bradford Hill criteria, that the model has many virtues as a causal explanation. In the fourth and concluding part I will make some more general remarks about the status and significance of racialist thinking about race and IQ.[Note]
Last updated: 4/18/2018
I was asked to meta-analyze a century (1914-2014) of IQ/Academic achievement and racial admixture (genealogy, gestalt racial appearance, and color) research. There is a lot out there, especially when one takes into account MA & PhD dissertations. To this end I am posting $20 (negotiable) bounties for each of the following dissertations (to be paid in bitcoin):
–Snider, J. G. (1953). A Comparative Study of the Intelligence and Aptitudes of Whites and Nezperce Indians (Doctoral dissertation, University of Idaho).
–Zimmerman, H. E. (1934). The Indian’s Ability to Learn Mathematics According to Degree of Indian Blood. MA, Kansas State Teacher’s College, Pittsburg.
–Rainey, C. D. (1932). A study of the Salem Indian High School, comparing the cultural background, the intelligence scores, and the per cents of white blood, and the classroom grades (Doctoral dissertation, Willamette University).
–Ross, D. D. (1962). A comparative intergroup study of the academic achievement and attendance patterns between the full-blood type and the mixed-blood type Oglala Sioux Indian student in the secondary department of Oglala Community School, Pine Ridge, South Dakota (Doctoral dissertation, Chadron State College).
The full dissertations are not needed, but just a copy (or photo) of the relevant tables and/or correlation matrices along with the following sample characteristics: country of sample, first order administrative unit of sample (e.g., North Carolina), group type (e.g., school, college, random stratified), ethnic group, age range, sample size, description of the ancestry index, cognitive tests used, statistical methods for comparing admixed groups (e.g., means & SD, correlations, Chi-square)
This should be an easy job if you are at one of the schools at which there is a copy of the dissertation. If interested email at firstname.lastname@example.org. I will update this list as I go along.
Given the central role that testing plays in the American educational system, most datasets that we have on racial and ethnic differences in cognitive ability include only children, adolescents, or young adults. Most of the economic and social effects of cognitive differences are, however, produced by the working age population, so it would be useful to have test scores from older adults as well. The PIAAC survey of adult skills conducted by the OECD provides excellent data for this purpose. Continue reading
Regression to the mean, RTM for short, is a statistical phenomenon which occurs when a variable that is in some sense unreliable or unstable is measured on two different occasions. Another way to put it is that RTM is to be expected whenever there is a less than perfect correlation between two measurements of the same thing. The most conspicuous consequence of RTM is that individuals who are far from the mean value of the distribution on first measurement tend to be noticeably closer to the mean on second measurement. As most variables aren’t perfectly stable over time, RTM is a more or less universal phenomenon.
In this post, I will attempt to explain why regression to the mean happens. I will also try to clarify certain common misconceptions about it, such as why RTM does not make people more average over time. Much of the post is devoted to demonstrating how RTM complicates group comparisons, and what can be done about it. My approach is didactic and I will repeat myself a lot, but I think that’s warranted given how often people are misled by this phenomenon.
There are a couple of new, well designed, obtainable, surveys out — with ancestry, MRI, and cognitive data – which should allow for the (dis)confirmation of certain conjectures of ill repute:
–Neurodevelopmental Genomics: Trajectories of Complex Phenotypes (age 8-21, N ~ 10,000)
–The Brain Genomics Superstruct Project (age 18-35, N ~ 1,500)
For example, Greg Cochran likes to go on about how major ancestry groups often differ in crude brain morphology, and how these differences probably explain a significant chunk (> 20%) of bio-ancestry related differences in CA. I doubt it. But if he agrees to specify the analytic strategy, I will try to get the data and run the analyses.
I did look through the PING survey (age 3-21, N ~ 1,500) – which might not be very informative owing to the age structure. Going by this, Greg seems to be more or less correct about some of the endo differences and probably about their origins. As an example, Figure 1 & 2 show the B/W diffs for intracranial and total brain volume by age. (AAs are picked out for illustration since they are the largest non-White ethnic group, showing the biggest deviation from Whites.) And Figure 3 shows the relation between brain volume and ancestry in the self-identified AA group; the results were basically the same for intracranial volume, etc. — and so not shown.
Yet, as seen in Table 1 & 2, CA was more or less uncorrelated with these particular endophenotypes (r = 0.07-0.08); unsurprisingly, CA explained virtually no endo differences, and vice versa. Yet, CA was strongly (negatively) associated with both African and Amerindian ancestry – and also, though to a lesser degree, with Oceanian.
Perhaps there is a more sound way to run the numbers? Or a better way to take into account age? Dunno, it’s not my position to defend.
Thought criminal extraordinaire, Steve Sailer commented recently on Foreign Policy’s article, “Brazil’s New Problem With Blackness.” Money quotes:
These policies didn’t eliminate race, but they did affect how it came to be classified. The marker of race drifted away from a binary consideration of a person’s ancestry and became increasingly based on one’s appearance.
What ultimately binds these definitions together is an awareness that the less “black” a person looks, the better — better for securing jobs, better for social mobility. The widespread acceptance of multiracial identities in Brazil coexists with steep racial inequality — a contradiction that the sociologist Edward E. Telles has called “the enigma of Brazilian race relations.”
Eleven experts comprised the panel, among them UFPel administrators, anthropologists, and leaders in the wider black community of Pelotas. They received strict guidelines from the Public Prosecutors Office: “Phenotypical characteristics are what should be taken into account,” read the instructions. “Arguments concerning the race of one’s ancestors are therefore irrelevant.”
“In Aug. 2016, after it had become clear that the law left room for fraud, the government ordered all departments to install verification committees. But it failed to provide the agencies with any guidance.
The Department of Education in Para, Brazil’s blackest state, attempted to fulfill the decree with a checklist, which leaked to the press. Among the criteria to be scored: Is the job candidate’s nose short, wide and flat? How thick are their lips? Are their gums sufficiently purple? What about their lower jaw? Does it protrude forward? Candidates were to be awarded points per item, like “hair type” and “skull shape” …
But black activists say such measures are unavoidable.
When you allow your national policies to be guided by sociological theories, like those of Telles, you are bound to run into this type of mess.
Below are regression results, based on the Pelotas Birth Cohort (n = ~ 2850), for genetic racial ancestry, interviewer and interviewee-reported color (corr), and three SES indicators. In this 1982 birth cohort, independent of European ancestry, it can be seen that there is no consistent negative association between interviewer rated “black” appearance and outcomes. That is, in Brazil, the average race of one’s ancestors is more relevant than stereotypical race-associated phenotypic characteristics. (Note: the sample sizes for “Yellow” and “Indigenous” were small, so those estimates are fairly unreliable; also, neither an East Asian nor Amerindian ancestry component was included).
(Source: F. Hartwig, personal communication, March 4, 2016; full results)
So, if one is interested in addressing historic race-related inequalities, it would be more efficient and just — since (dis)advantages are mostly being passed along lines of descent — to positively discriminate according to objectively determinable biogeographic ancestry, not subjectively assessed stereotypical racial appearance. And it’s hard to see how requiring 23andMe reports would be more intrusive than having a 12 member panel examine applicants for nose width, lip thickness, craniofacial morphology, etc. to see if they are sufficiently African-looking.
Of course, this isn’t going to happen any time soon, since the conclusion that ancestry with respect to major racial or descent groups is relevant to social outcome needs to be evaded, even at the expense of good science and quality social policy.