No country has an IQ below 60
TL;DR.
Becker’s national IQ dataset estimated the IQs of some nations to be extremely low (e.g. 43 for Nepal, 45 for some Western African countries, 48 for Guatemala).
These figures are clearly indefensible. The creator of the dataset says as much and currently employs a policy of setting the absolute minimum of IQ estimates to 60. Argument to authority aside, multiple lines of evidence suggest that it is unlikely that there is any nation has an IQ of below 60 due to deficiencies in general intelligence.
Failure to replicate
In every single case, the countries I highlighted previously do not score as poorly on international scholastic assessments (e.g. PISA) — which I focus on because they are much less likely to be fraudulent and tend to be higher quality samples. The following samples I cite come from either the World Bank’s dataset of harmonized learning outcomes or from my HLO+IQ dataset.
Without further ado:
Nepal: the nation scored a 340 on the EGRA (reading test) according to the World Bank, which is equivalent to an IQ of about 76 [ 100 - (500-340)*15/100 = 76 ]. On IQ tests, two Nepali samples found by Kirkegaard scored at an average of 77 and 78.
Sierra Leone: scored a 240 on the EGRA, equivalent to an IQ of 61.
Liberia: scored at a mean of 273 and 294 on the EGRA, equivalent to IQs of 66 and 69.
Guatemala (GTM): scores at an IQ of 75-81 on scholastic tests such as the PISA and LLECE (Latin American PISA clone).
Gambia (GMB): three different EGRA samples scored at a mean of 68, 73, and 81.
Nicaragua (NIC): scored at an IQ of 76 on the ERCE (Latin American PISA).
Guinea (GIN): scored at an IQ of 57 - 65 on scholastic tests administered within Africa.
Ghana (GHA): scored at an IQ of 59 - 71 on the TIMSS (math and science tests).
Ivory Coast (CIV): scored at 56-67 on some regional African scholastic tests.
South Sudan (SSD): scored at 66 and 67 in two EGRA samples.
Mali (MLI): scored the IQ equivalent of a 61 in an EGRA sample.
Mauritania: scored the equivalent of a 66.7 on the PASEC reading test.
Togo: scored the equivalent of about a 74 on some regional African scholastic tests (I don’t trust the norming the world bank did here, but that’s a score to settle another day).
When the IQs of countries that have been estimated below 60 is estimated based on higher quality scholastic estimations, the averages increase, though are still clearly far below the global average of 85.
First principles reasoning
National IQs fluctuate for three main reasons: true genetic differences in general intelligence, true environmental differences in general intelligence, and test bias. Technically there are a few more factors that could be involved (e.g. non-g ability tilts), but they are largely irrelevant at the national level, as differences between different ability factors (e.g. math and science) correlate extremely highly.
Consider that the average American Black scores an IQ of 80 in the largest, high quality datasets (e.g. NLS, ABCD, Project Talent) regardless of cohort year. Some of that is inflated by the fact that American Blacks are roughly 20% European on average, implying a genotypic Sub-Saharan African IQ of about 75 if a between group broad-sense heritability of 100% is assumed.
Environmental effects on human intelligence provably exist based on twin studies; it would be silly to think these effects do not vary between groups. Unfortunately, these environmental factors are very difficult to identify, as most of the factors that people like to study (e.g. parental education, wealth) are genetically confounded and not causally related to intelligence. All things considered, for a Black African nation to have an IQ of below 60, the environment would have to be so bad that it drops IQ by about 15 points. Does an effect this strong exist?
Probably not.
Education does increase IQ scores, but the increases are on specific abilities, not general intelligence (see Lasker & Kirkegaard or Richtie & Bates), which is what IQ tests attempt to measure in the first place. While some of the low measured African IQ could be plausibly be attributed to low levels of education, this does not imply lower levels of intelligence. Notably, even in the more developed and educated South Africa, Blacks still score at an IQ of roughly 70.
Based on priors, nutrition should have an effect on IQ, though almost every time the effect of nutrition improvements on IQ has been studied, the results have been null, meaning that it is very unlikely that there are large effects of nutrition on intelligence within the current range of nutritional quality that is observed within humans.
Parental social class is related to child intelligence, but this is due to genetic confounding, based on both molecular genetic evidence as well as twin modeling. The latter study tested the hypothesis indirectly by testing whether the cognitive abilities of parents and their children was related for non-genetic reasons using cognitive testing data of twins and parents. The answer is that they do not, as the best fitting twin models have no social transmission pathway:
The first model assumes no social transmission (m = 0), the second model assumes no dominant genetic effects (d = 0), and the third model assumes no siblings specific environmental effects (s = 0). We compared these different models by fixing non-significant model parameters to zero and determined the best fitting model based on the Akaike Information Criterion (AIC). An overview of all estimated models and their respective parameter estimates is provided in Tab. 7.
For all four cohorts, the best fitting model assumes no dominant genetic effects and no social transmission as both parameters, d and m can be fixed to zero. So, the best fitting model suggests genetic inheritance to be the dominant mechanism of intergenerational transmission of cognitive ability which is in line with hypothesis H2. Importantly, even regardless of the estimated model, there is no evidence for mechanisms of social transmission as the estimate m is very small and non significant in all of the estimated models (see Tab. 7). So, while the CTD provided some indirect support for a social transmission of cognitive ability, the NTFD results suggest that the CTD estimates of the influence of the shared family environment is upwardly biased due to assortative mating and all remaining shared environmental influences are specific to the siblings and independent of the parents (S).
While the “Flynn effect” does exist, the effect is largely because the tests are not measuring the same properties in different cohorts. The best evidence for this are the tests of measurement invariance which find that IQ tests are biased in favour of recently born cohorts. The second is that different measurements of intelligence appear to be changing at different rates: reaction time, colour differentiation, and backward digit span are all declining; crystalized ability has been constant since the 50s; and every other test has been increasing across time.
The Verdict
I think people like to post images of the super low (<60) African IQ scores because it’s funny to dunk on the Blacks. But really this is doing the opposite; it’s just hard to believe that any country actually scores below a 60 on an IQ test, and highlighting these estimates is making these datasets less reputable.
Lastly, these super low implausible estimates of certain countries do not imply that the whole dataset is rotten. Consider my national IQ map, which uses 0 geographical imputations:
Think critically. What kind of people tend to score high on IQ tests, within countries? The wealthy score above the poor, Asians score above Whites, who score above Latin Americans and Arabs, who score above Blacks. Naturally, the same pattern is found across countries. Even if these differences are environmentally caused (they are not), it would be bizarre to think that national IQ datasets are not measuring a latent reality, even if they only do so with moderate accuracy.
So, what then explains the anomalously low mean IQs observed in certain samples within certain countries? Given the rates of scientific malpractice in the third world, fraud must be considered as a potential cause, followed by test bias. As Flynn effects differ in magnitude between nations, and that these effects are largely artefacts of test bias, it would be reasonable to conclude that some differences in national IQs are caused by test bias. If the magnitude of these biases differs by sample and test, then it would be expected that a few samples have anomalously low means that do not adequately represent how intelligent a nation is.
> The creator of the dataset says as much and currently employs a policy of setting the absolute minimum of IQ estimates to 60.
This doesn't seem to be defensible. Implausible sure but how can you justify fiddling with raw data this way?
People with IQs of 60 are usually within institutions, certainly within special ed in the US. When you deal with them you know they are dumb.
Traveling to Nicaragua and Nepal, working with the people, I experience them as normal. Shopkeepers could do arithmetic - this was in the days before calculators were ubiquitous. They managed to run small businesses.
This was even true, though to a lesser degree, than Haiti. You are right to question the administration and even the meaning of the tests.