Note: We computed these results based on multiple versions of the ABCD data (v2.01 & v3.01) and with different inclusion/exclusion criteria. I originally posted a version based on the ABCD 2.01 data filtered for missing admixture, and other scores. However, after looking, I found a version that uses the maximum 3.01 sample with age-corrected NIHTBX scores (N = 11474). While the scores for the two versions correlate at r = .98, in some cases (e.g., Vietnamese), there is a notable difference. I have now replaced the original table with the one based on N = 11474 and moved the original table to the end of the post. For replicability and modifiability, I attached the latest R code which I had in my file.
……
In our manuscript, titled “Reply to Warne,” we present average eduPGS and NIH Toolbox composite scores from the ABCD study, categorized by ethnic and religious groups. In our analyses, we used unweighted means instead of sample weighted scores, since we were only interested in the correlation between mean eduPGS and cognitive ability. However, we also computed weighted NIH Toolbox scores, which may be of interest to some readers.
These scores were computed using the survey package for R as recommended by Heeringa and Berglund (2020). These weighted scores, reported below, represent the “neuropsychological performance” scores, measured between 2016 and 2018, of broadly representative samples of 10-year-old American children. (Though, children were excluded, by the ABCD consortium, if they were not fluent in English or if one of their parents were not fluent in either English or Spanish.) The first three columns, after the group labels, display the sample size, means, and standard deviations, respectively. The fourth column presents the scores normalized with the non-Hispanic White mean set to 100.00 and standard deviations set to 15.00. To norm scores, we pooled the standard deviations across all groups (pooled SD = 16.45) and transformed the values using the pooled SD. On a reader request, I added average years of parental education, which I previously outputted, in the fifth column.
The ethnic groups are mutually exclusive, and the specific variables used to code them are provided in the supplementary materials of the manuscript. Classifications are based on the race/ethnicity of the child as reported by the responding parent in conjunction with the nationality and immigrant status of the parents; see the Parent Demographics Survey for specific variables and the second table for definitions. To be clear, some of the definitions do not perfectly overlap with ones commonly used in the social sciences. For example, the classification “USA Blacks” refers to children who were identified as being Black, not being White, not being Hispanic, but also not having an immigrant parent or grandparent. This was done because, when computing the scores, we were interested in mutually exclusive ethnocultural groups.
Bear in mind that the sample sizes are often small and so the corresponding estimates are imprecise and also that the NIH Toolbox battery is fluid-intelligence loaded. For comparison, Sailer, in 2009, reported cognitive abilities of legal American immigrants based on the digit span backwards test. Additionally, in 2015, I summarized scores by immigrant generation and ethnic groups mostly based on scholastic tests.
| Ethnic/National Group | N | M | SD | IQ-Metric Score | Parental Education (Years) |
|---|---|---|---|---|---|
| Chinese | 81 | 116.53 | 21.02 | 111.32 | 16.95 |
| Korean & Japanese | 33 | 115.13 | 19.15 | 110.05 | 16.36 |
| White & Asian Indian | 44 | 114.66 | 14.19 | 109.62 | 16.75 |
| White & Korean/Japanese | 78 | 111.41 | 18.02 | 106.65 | 15.89 |
| White & Chinese | 77 | 109.77 | 18.16 | 105.16 | 16.42 |
| White & Filipino | 60 | 109.67 | 18.09 | 105.07 | 16.16 |
| Filipino | 51 | 107.99 | 17.53 | 103.53 | 15.5 |
| Other Asian | 52 | 106.8 | 20.17 | 102.46 | 15.91 |
| Asian Indian | 53 | 106.77 | 17.03 | 102.42 | 16.8 |
| White | 5858 | 104.11 | 16.51 | 100 | 15.45 |
| White & Pacific Islander | 25 | 103.74 | 16.9 | 99.66 | 15.47 |
| Vietnamese | 24 | 102.68 | 14.51 | 98.69 | 15.95 |
| N. Africa & Mid. East | 47 | 100.33 | 20.01 | 96.56 | 14.92 |
| Pacific Islander | 17 | 99.79 | 12.18 | 96.06 | 14.07 |
| White & Native American | 144 | 99.32 | 15.36 | 95.63 | 14.51 |
| Central & South American | 352 | 98.36 | 16.98 | 94.76 | 14.15 |
| Not Identified | 217 | 96.69 | 17.78 | 93.24 | 13.4 |
| Dominican | 38 | 95.1 | 16.65 | 91.79 | 13.99 |
| White Mexican | 775 | 95.1 | 16.37 | 91.78 | 12.8 |
| White Cuban | 151 | 94.98 | 16 | 91.67 | 13.92 |
| NH Black & White | 418 | 94.93 | 16.91 | 91.63 | 14.14 |
| Other Hispanic | 518 | 94.56 | 17.75 | 91.29 | 13.95 |
| White Puerto Rican | 133 | 94.22 | 17.23 | 90.98 | 13.74 |
| Black African | 59 | 93.84 | 13.41 | 90.63 | 14.97 |
| Other Cuban | 30 | 92.81 | 18.3 | 89.69 | 14.33 |
| Native American | 39 | 92.29 | 16.1 | 89.22 | 13.14 |
| Other Mexican | 460 | 91.82 | 16.02 | 88.79 | 11.9 |
| Black Caribbean | 51 | 91.74 | 16.79 | 88.72 | 14.26 |
| Black & Other Puerto Rican | 90 | 90.61 | 15.49 | 87.69 | 13.22 |
| USA Black | 1499 | 85.44 | 14.8 | 82.98 | 13.32 |
Survey design code
svyr <- svydesign(data=merged_df_group, id=~site_id_l, strata=NULL, weights=merged_df_group$acs_raked_propensity_score)
summary(svyr)
svyby(~NIHTBX, ~SIRE_, svyr, svymean)
svyby(~NIHTBX, ~SIRE_, svyr, svyvar)
svyby(~NIHTBX, ~SIRE_, svyr, unwtd.count)
svyby(~edu_average, ~SIRE_, svyr, svymean, na.rm=T)
Ethnic definitions
| Ethnic/National group | Defintion |
|---|---|
| Asian Indian | not Hispanic & Asian Indian |
| Black & Other Puerto Rican | not White & Hispanic & Puerto Rican |
| Black African | not Hispanic & not White & Black & parent born in Sub-Saharan African country |
| Black Caribbean | not Hispanic & not White & Black & parent born in Caribbean country |
| Central & South American | Hispanic & Central and South American (excluding Mexico) |
| Chinese | not Hispanic & not White & Chinese |
| Dominican | Hispanic & Dominican |
| Filipino | not Hispanic & not White & Filipino |
| Korean & Japanese | not Hispanic & not White & Japanese OR Korean |
| N. Africa & Mid. East | not Hispanic & parent born in North African or Middle Eastern country |
| Native American | not Hispanic & not White & Native American |
| NH Black & White | not Hispanic & Black & White |
| Not Identified | Not any of the other classes |
| Other Asian | not Hispanic & Other Asian |
| Other Cuban | Hispanic & not White & |
| Other Hispanic | Hispanic & Not any of the other Hispanic group |
| Other Mexican | Hispanic & not White & Mexican |
| Pacific Islander | not Hispanic & not White & Pacific Islander |
| USA Black | not Hispanic & not White & Black & non-immigrant family |
| Vietnamese | not Hispanic & not White & Vietnamese |
| White | not Hispanic & White & no other race marked |
| White & Asian Indian | not Hispanic & White & Asian Indian |
| White & Chinese | not Hispanic & White & Asian Chinese |
| White & Filipino | not Hispanic & White & Filipino |
| White & Korean/Japanese | not Hispanic & White & (Korean OR Japanese) |
| White & Native American | not Hispanic & White & Native American |
| White & Pacific Islander | not Hispanic & White & Pacific Islander |
| White Cuban | Hispanic & White & not Black & Cuban |
| White Mexican | Hispanic & White & Mexican |
| White Puerto Rican | Hispanic & White & not Black % Puerto Rican |
Additional information
In response to a reader’s request, I append the scores by the grandparents’ country of origin — as reported by the responding parent — instead of race. For these, cases were subset to individuals whose maternal or paternal grandparents came from the same country. Originally, we planned to correlate scores with National IQ scores but we eventually decided that reports were too sparse and also unreliable.
Version 1 based on the reduced dataset
The original table, now replaced with the version based on N = 11474, is shown below.
| Ethnic/National group | N | M | SD | IQ-metric scores | Parent Education (Years) | |
|---|---|---|---|---|---|---|
| Korean & Japanese | 16 | 117.4 | 21.87 | 112.23 | 16.27 | |
| Chinese | 43 | 115.79 | 14.67 | 110.76 | 16.96 | |
| White & Asian Indian | 41 | 114.57 | 14.53 | 109.64 | 16.72 | |
| White & Korean/Japanese | 71 | 110.88 | 18.22 | 106.26 | 16 | |
| White & Chinese | 70 | 109.37 | 18.34 | 104.88 | 16.32 | |
| White & Filipino | 55 | 109.07 | 17.57 | 104.6 | 16.19 | |
| Asian Indian | 50 | 107.59 | 17.33 | 103.25 | 16.77 | |
| Filipino | 36 | 107.09 | 17.26 | 102.79 | 15.34 | |
| White | 5614 | 104.04 | 16.5 | 100 | 15.51 | |
| Other Asian | 43 | 103.1 | 17.73 | 99.14 | 15.55 | |
| N. Africa & Mid. East | 41 | 100.12 | 20.98 | 96.41 | 15.05 | |
| White & Native American | 136 | 99.29 | 15.55 | 95.65 | 14.56 | |
| Central & South American | 314 | 98.25 | 16.92 | 94.7 | 14.43 | |
| Vietnamese | 12 | 96.04 | 14.45 | 92.68 | 15.94 | |
| Not Identified | 202 | 95.76 | 15.5 | 92.42 | 13.83 | |
| Dominican | 36 | 95.6 | 16.56 | 92.27 | 14.25 | |
| White Mexican | 669 | 95.39 | 16.84 | 92.08 | 13.2 | |
| NH Black & White | 397 | 95.16 | 16.69 | 91.87 | 14.21 | |
| White Cuban | 146 | 95.06 | 16.16 | 91.78 | 13.97 | |
| Other Hispanic | 484 | 94.89 | 17.87 | 91.62 | 14.06 | |
| Black African | 57 | 93.8 | 13.44 | 90.63 | 14.79 | |
| White Puerto Rican | 124 | 93.74 | 16.65 | 90.57 | 14.05 | |
| Native American | 36 | 93.23 | 15.58 | 90.1 | 13.14 | |
| Other Mexican | 366 | 92.68 | 16.33 | 89.6 | 12.25 | |
| Black Caribbean | 46 | 91.66 | 17.19 | 88.67 | 14.3 | |
| Other Cuban | 28 | 91.53 | 17.48 | 88.55 | 14.55 | |
| Black & Other Puerto Rican | 75 | 90.91 | 15.94 | 87.98 | 13.68 | |
| USA Black | 1426 | 85.51 | 14.77 | 83.04 | 13.43 |
I can’t remember if it was your paper on the ABCD or the similar (but localized) Philadelphia longitudinal database, but my casual impression was that the black families seemed skewed downscale and the white families seemed skewed upscale, as if the researchers had made a special effort to recruit poor blacks and/or had let whites with an interest in the human sciences volunteer more and hadn’t much pursued less educated whites.
Or maybe whites have just gotten a lot more higher education on average by the ABCD in the 2010s compared to, say, the well-known participants in NLSY79 database cited often in “The Bell Curve.”
I was highly skeptical of the results at first. Now that I have read about the rigorous scientific methodology applied, I am confident in the quality of this work.
I was still wary of trusting the conclusions of the investigation but thank god you posted this comment making me better understand the implications. I assumed the conclusion is supposed to be that black kids are genetically less intelligent but the IQ score rather points to differences in education instead of inherent cognitive capabilities.
Anyways, interesting study but I was led here by a Twitter user posting this to support the Bell Curve argument…lmao
Am I right to think you do not support this flawed correlation Steve, out of curiosity?
The analysis does not conclude that IQ differences are due to education. There is not even strong evidence that education causes intelligence. As for genetic effects, they exist.
Who would believe or suggest that ‘education causes intelligence’? — that is preposterous — that intelligence results in a higher level of educational attainment is, in contrast, a very reasonable belief (and largely correct).
However it is true that many believe ‘IQ differences are due to education’, because determination of IQ is, literally, done by taking a test — and it is widely thought that IQ tests are culturally biased etc — so it follows naturally that a person with more education, or one who comes from a family with educated parents or where learning is valued and encouraged, will do better on an IQ test — this is why it’s important to circulate findings like the one in a subsequent article here: ‘It is striking that Asians with parents who did not complete high school score 100 points above blacks with PhD parents.’
Education does not affect IQ. I graduated highschool in 2003. I have no formal education beyond that. Yet, I’m a teacher, and trainer with an FSIQ of 150+. Education gives you tools, and perspective. It doesn’t give you horsepower.
I’m actually flabbergasted by the results. I seriously did not realize there was over a full deviation between races.
A few points or even 15 isn’t a big deal. You can make up for that with a good work ethic, focus and passion without any doubt.
But from 85 -110 is extremely concerning.
Out of sheer curiosity, is there any study done between black children, IQ and black children raised by white family IQ?
I believe the flynn effect is correct. And that if we were to take somebody with an average IQ of 85 today and put them back in 1960 they would be an average student. So obviously can’t be fully genetic we don’t evolve that quickly. But something is off here and that’s not something we can just ignore if it’s accurate.
There is this study (Table 2) but unlike the authors’ conclusion the findings actually support the hereditarian position (of partial genetic cause).
White socio-economic status is likely slightly inflated by missing about 30% of rural kids due to their not living close enough to brain scanners:
“As a consequence, neuroimaging research centers are more likely to be located in urban areas resulting in a potential under-representation of rural youth.”
https://www.sciencedirect.com/science/article/pii/S1878929317301809
There is an obvious typing mistake in table 2, Asians instead of Americans.
Central & South American Hispanic & Central and South Asians (excluding Mexico)
Thanks.
Do you know if Pakistanis are included under “Asian Indian” or are they entirely excluded? Would love to see data for ourselves!
In the U.S., Pakistanis are generally counted by federal bureaucrats as South Asians, but not Afghans or Iranians. The Khyber Pass is the dividing line.
See: “Additional information”.
I added the scores broken down by grandparents’ country of origin in cases when either both maternal or both paternal grandparents came from the same country.
Bear in mind: Scores are not by race; sample sizes are often tiny; there is likely outcome-relevant response bias; there is exogamy-dependent attrition.
The average IQ of North Africans and Middle Easterners is 96?
Richard Lynn gives an average of 84 for racial differences in intelligence in 2006.
Is this due to migratory selection?
Do these people do well in the United States? Academically and professionally?
In Europe, Arabs are a very big problem, especially in France, so I imagine it’s not the same selection process…
Arab migrants to France are two types: illegal immigrants who were generally unemployed and work in low wage sectors, and those that are middle class and work at high and selective institutions, most of these people pass very hard government exams in France.
Look at the sd and you can see that they have the second highest. This means that the Lynn may have been speaking about the median or have a very large sample. Among the arabs I know across the world, there are many extremely smart people and many that are the opposite.
Also, the sample is about 15% twins because of the obvious fascination of twins for human science. I’m guessing that white twins tend to be fairly heavily due to older mothers using fertility treatments, who tend to be more upscale.
So, the 17 point gap between whites and non-Hispanic non-immigrant non-mulatto African-Americans might be slightly inflated by the white sample being slightly more metropolitan and upper middle class than the national white population.
But, yeah, it suggests the country is not making much progress at closing the long-standing white-black IQ gap.
I found a version based on N = 11474, which uses the max 3.01 sample. See note at the top. (We changed versions halfway through the admixture analyses, which is why the case numbers in “Linear and partially linear models” are lower than in subsequent papers. As a result, I have different versions of analyses laying around.) But the BW gaps are hardly changed. (Bear in mind the “Black” here refers to non-immigrant family (up to grandparents) African Americans. So, this is a somewhat different group than reported by say NAEP.
More specifically, 44% (or more) of the US Black sample is from the South versus 24% of the non-Hispanic White sample. By the way, twins are 18% of the total sample.*
“The South is also the region with the highest concentration of the single-race U.S. Black population: 59% of that group lives there as of 2021. The Midwest is home to 17%, while another 15% live in the Northeast and 9% live in the West.”
https://abcdstudy.org/scientists/data-sharing/baseline-data-demographics-2-0/
https://www.pewresearch.org/social-trends/fact-sheet/facts-about-the-us-black-population/
Dev Cogn Neurosci. 2018 Aug:32:16-22.
Recruiting the ABCD sample: Design considerations and procedures *
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6314286/
Here’s a technical question: if I’m using Excel’s Normdist function to calculate the percentile that groups would fall at in the white distribution, I enter the mean as 100. But should I enter the standard deviation at 15 or as 16.39, or as 16.5? For example, if I use a standard deviation of 15, the African American average of 83 comes out at the 13th percentile. But if I use 16.5, it comes out at the 15th percentile. Any thoughts?
The IQ-metric scores were standardized to the white distribution, so when calculating the percentiles of different groups in the white distribution the total sample standard deviation shouldn’t be used(note: in this sample, the white and black sd are roughly equal, and the larger sd for the total sample is a consequence of this being a diverse sample in terms of ethnic differences in cognitive ability).
We used the SD pooled across groups (i.e., the weighted average of each group’s standard deviation), not the total sample SD. So, the between-group variance is factored out.
Well, I reported the untransformed scores and SDs because, sometimes, it’s also informative to look at variances. For cut-off percentiles, you might want to take into account the group-specific variances. That said, if you use the IQ-metric scores, set the SD to 15.
Thanks.
Sample weights are designed to adjust for selectiveness(it’s larger without sample weights), and a 17 point gap seems to be in line of what you see in other fairly representative samples(1.13 d in the NLSY97, various college/graduate admissions tests(ACT, SAT, MCAT, LSAT etc.), including in states where’s it’s mandatory for students, and IQ test standardisations(when g is modeled instead of full scale IQ scores, like in Frisby and Beaujean 2015). Also when you’re looking at academic tests(PISA/TIMMS/PIRLS/SAT etc.), it’s better to use composite scores because of the imperfect correlation between math/science/reading(because it’s worse to score badly on multiple tests compared to just one).(Schneider, J. W. (2016), https://web.archive.org/web/20210220193517/https://assessingpsyche.wordpress.com/2016/02/17/the-composite-score-extremity-effect/).
Your findings are very interesting. Just one comment. It seems to me that the sample sizes are too small to represent their groups. Would you let me know why each sample size is sufficient to represent each group?
A better question to ask should have been: whether these subgroup means are consistent with the means reported in other studies.
You can plainly see that the Japanese, Koreans, and Chinese as children in general seem to be very bright for their age.
I was keen to see the genetic markers used to distinguish between the various racial groups, since many popular online information sources (i.e. Britannica) claim that no such markers have been found. So I was glad to see the link that appeared to answer my question about the genetic markers: “The ethnic groups are mutually exclusive, and the specific variables used to code them are provided in the supplementary materials of the manuscript.” Unfortunately, the linked page returns a Page Not Found error at https://osf.io/exafy. Searching osf.io didn’t bring up useful results.
Please let me know where I can find the supplementary materials.
Thanks!
Might be this one here.
It looks like there’s no file with the supplementary tables mentioned in the paper on that page. Some files were deleted june 4. Please add the supplementary files back if possible.
Thank you
Link fixed.