In 2002 Richard Lynn and Tatu Vanhanen’s book IQ and Wealth of Nations estimated the IQs of 185 countries. Critics accused them of cherry picking sources, using unrepresentative samples, comparing and combining samples tested on wildly different tests taken decades apart, and daring to think IQ could be measured cross-culturally. And yet despite nearly two decades of opprobrium, those national IQs remain a landmark, cited in countless peer reviewed articles and repeatedly revised.
One way Lynn has validated his numbers is by showing their high correlation with international exams like Programme for International Student Assessment (PISA). Another independent data-set against which Lynn’s numbers can be tested (assuming he already hasn’t done so) is the IEA‘s Trends in International Mathematics and Science Study (TIMSS). Ostensibly an achievement test, the math section resembles an IQ test, and the test is scored so that most countries average between 400 and 600.
Using the score distribution of UK students as a reference group (see technical note below), I converted the scores from 39 countries to IQ equivalents. My source for the TIMMS scores is exhibit 1.2 in this report.
TIMMS score (8th grade math; 2015)
Korea, Rep. of,
Hong Kong SAR
United Arab Emirates
Iran, Islamic Rep. of
Consistent with Lynn’s hierarchy, we find that East Asian countries cluster around the top (Japan IQ 112 to Korea, Repub of, IQ 116), followed by white majority countries (New Zealand IQ 95 to Russian federation IQ 103), followed by Dark Caucasoid countries (Saudi Arabia IQ 73 to United Arab Emirates IQ 90) and lastly sub-Saharan countries (South Africa IQ 73 to Botswana IQ 77). And while Lynn’s data was ridiculed for declaring entire countries “mentally retarded”, it’s perhaps a sign of higher quality data that no country in this data-set averaged below IQ 70 (though most of the poorest countries chose not to participate).
On page 95 of the report, we’re told that only 10% of England’s 8th graders could score 625+, 36% could score 550+, 69% could score 475+, and 93% could score 400+. Subtracting these percentages from 100 gives the following percentiles: 90, 64, 31, and 7 which can be converted to the following IQs: 119, 105, 93, and 78. Now that we have the IQ equivalents of four TIMMS scores, we can make a linear equation converting TIMMS to IQ which is IQ = 0.18(TIMMS score) + 6.5:
The NAEP provides ethnic averages and percentiles in both reading and math for 8th graders in 2019. I chose 8th graders because they are the oldest age group for which they have nationally representative samples, since 12th graders only include those who have not yet dropped out of school. Note: scores are reported on 0 to 500 scale.
American Indian/Alaska native
American Indian/Alaska native
Although the NAEP is not an IQ test, the correlation between IQ tests and scholastic achievement tests is about as high as the correlation between two IQ tests, making them statistically equivalent in the general population. Further, the main reason people care about racial IQ gaps is because they translate into racial learning gaps, so converting to IQ seems appropriate and the advantage of using the NAEP to infer group IQ gaps is the excellent sampling this data has among subjects who have spent their whole lives learning these skills.
American Indian/Alaska native
For technical details on how these scores were converted to IQ, see technical note below.
The reading, math, and composite NAEP scores were converted to IQ by equating the white NAEP means with 100 and the white NAEP SDs with 15. The reading and math SDs were estimated by subtracting the 90th percentile NAEP scores from the 10th percentile scores and dividing by 2.53 (the bell curve Z score difference between these percentiles) .To determine the white mean of the composite score, we simply add the reading and math means, which gives 564. The white SD of the composite score was crudely estimated by assuming the reading and math correlation among all white 8th graders taking the NAEP is the same as the correlation among all college bound 17-year-olds taking the SAT (r = 0.67 according to Herrnstein and Murray). Using the formula for calculating the composite SD (from page 779 of the book The Bell Curve by Herrnstein and Murray):
I am extremely honored that Davide Piffer (who has a blog) was kind enough to give our community an exclusive interview. While the leading geneticists in academia have explained only about 10% of the variance in IQ (or its proxy education) at the individual level, Piffer working on his own has reported near perfect correlations between the mean IQs of entire ethnic groups and their polygenic scores, making him a rock star in the HBD community. Virtually no one else on the planet is doing this kind of cutting edge research (at least not publicly).
In retrospect it makes perfect sense that aggregated data should correlate much better than individual level data. Imagine you visited every country in Eurasia and asked only the first person you met in each country their height. Such a small sample size (n = 1) from each country would tell you nothing about which individual country was taller than which, but if you averaged all the heights from the European countries and compared them to the average heights from the Asian countries, you’d learn a lot about which continent was taller. That’s because the small sample size at the level of individual countries is multiplied by the large sample of countries in each continent.
It’s the same with genomically predicting IQ. The small sample of single nucleotide polymorphisms (SNPs) sampled in each individual is multiplied by the large number of individuals sampled in each ethnic group, so while individual predictions are weak, group predictions are strong because individual error cancels out in the aggregate.
Below is my exclusive interview with Piffer. The interview has been lightly edited to remove typos and other mistakes. I began by asking him about table 5 in a 2019 paper he wrote. My statements are in red, while Davide’s are in blue.
PP: I’m very impressed by your work. But the correlation between PG score & mean IQ is so high in table 5 of Piffer (2019) that it seems too perfect. What would you say to skeptics who think you cherry-picked SNPs or manipulated your formulas to get such perfect results?
DP: Thanks. I didn’t cherry pick SNPs. I used the polygenic score provided by Lee et al and you can see that different PGS construction methods lead to same results… I used EA, EA Mtag, etc, weighted and unweighted..they all give same results. Also my paper replicates my previous findings and what I had predicted from theory years ago. The IQs aren’t cherry picked either because I used the same as I used in previous papers to avoid post hoc results.
PP: In table 1 of Piffer (2019), Peruvians & Colombians seem to have higher polygenic scores than the black populations, yet in Figure 11, Africa scores higher than the Americas. So who has higher polygenic scores: sub-Saharan Africans or Amerindians?
DP: Peruvian and Colombian aren’t pure. They are substantially mixed with Europeans. The groups in figure 11 are natives, so they better reflect the unadmixed population. Also the latter are from low coverage genomes with fewer markers so less reliable. I am working on a high coverage version of same datasets but it will take a while due to my limited funds.
Do you have some basic experience in bioinformatics? I am just looking for someone who could run the code on their laptop because it’s taking me a week to impute each chromosome. So I need to run it on multiple computers. But hey no bother…I will do it myself, it will just take it longer.
PP: No sadly I do not have experience with bioinformatics. But I can ask my blog & twitter readers if anyone has such experience and is willing to volunteer their time.
On table 5 of Piffer (2019) the African American PGS (GWAS sig) is 1.836 lower than the NW European PGS. But since African Americans are only 76% non-white (Bryc et al. 2015), can we roughly infer that un-mixed blacks would be 1.836/0.76 = 2.416 below NW Europeans, giving them a PGS score of 46.834?
DP: yes…also you have unmixed native Africans in the other tables. Kenyans, Yoruba, Mende Sierra Leone, etc
PP: In table 5 Latinos have a PGS (GWAS sig.) of 48.654. Do you think this could be used to estimate the PGS of unmixed Amerindians because according to Bryc et al, 2015, Latino Americans are 65.1% white (mostly southern European), 6.2% black, 18% Amerindian, and 11% unassigned, though the unassigned is broader East Asian/Amerindian so should probably be counted as Amerindian. Since you report the PGS for Southern Europeans and since I estimate the the PGS for pure blacks at 46.834, using simple algebra, I estimate unmixed Amerindians would have a PGS of 47.510.
DP: yes, but you should also cross-check these with the other table with scores for Peruvians and Mexicans and see if they converge.
PP: Good point. In one of your data sets you find a 0.57 correlation between PGS and latitude. Do you agree with Lynn’s cold winter theory of how racial differences in intelligence evolved?
DP: in part, yes. but it doesn’t explain the low Amerindian IQ because Native Americans were in Siberia during the Last Glacial Maximum and then they moved to North America at the end of it, which is also a cold region…So I think most of the differences are due to farming and civilization
PP: Well Lynn argues the anomalies can all be explained by population size. Low population races like Arctic people, Amerindians, Australoids, Bushmen, & pygmies have lower IQs than their climates predict because there weren’t enough positive mutations. Meanwhile high population races like East Asians, whites, South Asians, and West Africans have higher IQs than their climates predict. This would also explain why Neanderthals had lower IQs than their climates predict.
DP: but these SNPs are common among the races..the differences are explained by these common SNPs, not pop specific mutations. pop size is probably related to it through higher competition for resources selecting for higher IQ.
PP: I see…so then it was probably farming and civilization as you say. Just as cold climate boosted IQ because it was a novel environment to adapt to, so was farming, civilization and the literacy and numeracy requirements it imposed. Of course Amerindians also independently created civilization but most remained hunter-gatherers.
DP: yes… plus we don’t know how many of these SNPs are just life history or personality traits like C. stuff that farming selected for. most of them are related to g but a subset will also be related to conscientiousness. Emil et al in their Psych paper vetted their association with g in a sample though so I guess they must be genuine associations with IQ for the most part.
PP: Yes, because no one has given a huge sample (n = 1 million) of genotyped people a highly g loaded test. A perfect study would get a sample of 1 million people (from all over the world) and give them an extremely culture reduced test with many subtests to maximize g loading (i.e. block design, draw a person in the sand, name as many body parts as you can in 1 minute in your own language, pictorial oddities etc) and then enter the composite score, DNA and human development index of each person into a computer and have machine learning create a multiple regression equation predicting IQ using HDI & genomic variants as independent variables. By using such a diverse and global sample, one finds the genomic variants that correlate with IQ everywhere and thus are most likely to be causal.
PP: Now that the neanderthal genome has been published, why haven’t you tried to estimate their polygenic score? Richard Klein argues that before about 50 kya, modern humans and neanderthals had similar intellect, but suddenly around 50 kya there was a genetic brain change that allowed modern humans to leave Africa, colonize every continent, replace neanderthals & invent art & complex technology. Testing this hypothesis was the main motivation to sequence the neanderthal genome so there’s enormous interest in their intelligence, even in mainstream science.
DP: yes that’s the next step…we’re analyzing genomes from Bronze age now, but Neanderthal would be good. But funds are limited for this kind of research and I am not working in academia.
PP: Above you rejected Lynn’s population size mutation theory on the grounds that all races have all the known IQ related genomic variants, however it also seems you have no high coverage genomes from low population isolated groups like pygmies, bushmen, australoids, arctic people & pure Amerindians. Is it plausible that high coverage genomes of these groups would show they are missing some of the IQ enhancing mutations that appeared in the last 15,000 years?
DP: What I am saying is that you can see a difference even at the common SNPs in their frequencies. I cannot rule out that they are also missing these mutations but that would be an additional factor.
PP: Do you agree with John Hawks’s theory that positive selection in the last 5000 years has been a hundred times faster than in any other period of human evolution because of the explosion of new mutations & environmental change? This is the exact opposite of Gould who argued we have the same bodies and brains we’ve had 40,000 years ago and all subsequent change has been cultural not biological.
DP: from a purely theoretical point of view, yes, but one would need to study ancient genomes to empirically vet that hypothesis.
PP: Is there any strong evidence in support of Michael Woodley’s theory that white genomic IQ has declined by 10 or 15 IQ points since the Victorian era?
DP: I computed the decline based on the paper by Abdellaoui on British [Education Attainment] PGS and social stratification and it’s about 0.3 points per decade, so about 3 points over a century.
It’s not necessarily the case that IQ PGS declined more than the EA PGS..if anything, the latter was declining more because dysgenics on IQ is mainly via education so I think 3 points per century is a solid estimate
Thank you Davide Piffer for this interview. As mentioned above, you can find more of Davide’s thoughts on his blog.
Davide Piffer looked at 2,404 genomic variants found to predict education (a rough proxy for IQ) and used these to create polygenic scores of eight ethnic groups reared in First World conditions. He then compared the polygenic scores with the mean IQ of each group and found a 0.979 correlation.
The line of best fit allows us to predict the mean IQ of any group from their PGS (GWAS sig.):
Mean IQ = 9.31(PGS (GWAS sig.)) – 358
Given the 0.979 correlation, genotype predicts IQ remarkably well: Finnish 102, Ashkenazi 108, Southern Europe 99, Estonia 100, NW European 100, African American 83, Latino 95, East Asians 105.
So while our genomic predictions of IQ remain poor at the individual level, Piffer is showing we can predict the mean IQs of ethnic groups with incredible precision, at least when they’re all reared in similar countries.
Because we have only found a tiny fraction of the genetic variants associated with IQ (or its proxy education), the margin of error for predicting any one person’s IQ remains high. But when you try to predict the average IQ of an entire ethnic group, the overestimates and underestimates cancel each other out, and there’s a near perfect correlation between the mean polygenic score and the mean IQ.
With all the talk in the news about a potential war, it’s a good time to ask what war was like 80,000 years ago, as brilliantly depicted by one of my all time favorite movies, Quest for Fire (1980)
There were no guns so people (and I use that term loosely) would stab with spears, throw rocks or simply wrestle. Instead of dropping bombs on cities, people would try to drop boulders on folks on sitting around a camp fire by pushing it off of an above cliff.
The tribes in Quest for Fire can be divided into three main levels. 1) those smart enough to make fire (potential IQ around 80),
2) those smart enough to maintain fire but not smart enough to make it (potential IQ around 70),
and 3) those not smart enough to make or maintain it so they must steal it from more advanced tribes (potential IQ around 50).
Today every human population has mastered fire so we no longer fight wars over that, and instead (as Lion of the Blogosphere has implied) the World is divided into countries smart enough to make nuclear weapons (potential IQ around 100), countries smart enough to maintain nuclear weapons (potential IQ around 90) and countries smart enough to do neither (potential IQ around 80).
Quest for Fire as a culture fair test of fluid verbal IQ?
Another interesting feature of this film is that it could serve as a rare example of a of verbal IQ test that is both culture reduced and fluid (as opposed to crystallized). Since most of the dialogue is from no-known language ( a new language based on Indo-European roots was specifically created by Anthony Burgess ), high SES people can’t rely on their fancy education and must infer definitions on the spot.
If one scores much higher on an English vocabulary test than they do on a test like this, it implies either they were educated beyond their ability and/or cognitive decline (since their fluid verbal IQ was presumably good in the past to have acquired high crystallized verbal IQ).
Just from watching the above clip, readers can test themselves by defining the words “wogaboo” “dominyai” and “Ka Ka Ka”.
I recently watched Luce (2019) and I proclaim it one of the best films of the year. The film is about a light skinned African child soldier who is adopted by upper-class white parents and blossoms into the star of his high school. The white teachers and peers crown him their Golden boy because he is bright, articulate, polite, athletic, and has a nice smile. He is constantly asked to give speeches to the entire school, and much like Obama, held up as an example of the American dream.
However his history teacher, portrayed flawlessly by Octavia Spencer, begins to worry that Luce is too good to be true. As a dark skinned overweight black woman like Oprah, she had to claw her way up the ladder using hard work and brains, not having the luxury of being a light skinned male with upper-class white parents.
Realizing this teacher is a problem, he mysteriously starts charming her mentally ill younger sister, even suggesting the teacher invite her to school to see one of his speeches.
Of course the last thing this dignified teacher wants is for the white suburban school to know she has a schizophrenic sister with what appears to be an IQ around 70, resulting in one of the most graphic and humiliating scenes in movie history.
Was this all part of Luce’s master plan? The film doesn’t say, forcing the viewer to decide whether Luce is a misunderstood victim of society’s expectations, or a charming sociopath manipulating everyone.
The film is so good that a racist might assume it was written and directed by whites, but in fact the director and writers are black. When I learned this, I immediately suspected (correctly) that the director and co-writer was born in Africa, because such talent is more likely to be found among elite immigrants than the native population of any race.
However the Nigerian director (Julius Onah) gives much of the credit to his co-writer JC Lee who looks like a scrawny giggling Australian aboriginal with ripped jeans, though I suspect he’s African American.