Are Richard Lynn's national IQ estimates flawed?

Trying to gauge the average IQ of Sub-Saharan Africa.

Written by Noah Carl.

Back in 2010, Richard Lynn and Jelte Wicherts had a written debate over the average IQ in Sub-Saharan Africa. Lynn argued it was closer to 70, while Wicherts argued it was closer to 80. Much of the debate concerned the representativeness of samples. Wicherts suggested that Lynn had excluded studies purely because they yielded higher average IQs. Lynn insisted those studies had been excluded because they were based on unrepresentative elite samples, such as private school students.

The debate was never resolved in the sense that neither side conceded much ground. Lynn is now deceased (though others have continued his work, notably David Becker). And Wicherts appears to have moved on to other topics. A skim of his Google Scholar page indicates he hasn’t published anything on national IQ since 2010.

Away from the arcane pages of academic journals, the debate continues to rage on social media. Every so often, one or other version of the national IQ map goes viral. Then a bunch of critics jump in to point out that the underlying figures have been debunked. Even some hereditarians count themselves among Lynn’s critics. Describing his work on national IQ as “profoundly flawed”, Nathan Cofnas has stated that we should not be “making excuses for him”.

Over the last year, criticism of Lynn’s work on national IQ has entered a new phase. A group of academics led by Rebecca Sear has been campaigning for his articles to be retracted on the basis that they are not merely flawed but “explicitly motivated by white supremacist ideology”. (Unmentioned is the fact that Lynn deemed East Asian countries to have the highest average IQs, suggesting it would be more accurate to call him an “East Asian supremacist”.) The campaigners have already managed to get one recent paper by Heiner Rindermann retracted. And they are almost certainly the force behind Elsevier’s decision to “review” all of Lynn’s research. This, in turn, may be what led Elsevier to install two poorly suited editors at Intelligence without consulting the editorial board, thereby prompting mass resignations.

To this day, practically all the criticism of Lynn’s work on national IQ centres on his estimates for Sub-Saharan Africa.¹ Why? One reason is obvious: they’re the lowest. Many people are incredulous that the average IQ in Sub-Saharan Africa could be more than 30 points lower than the average IQ in Western Europe. And in light of various provocative things Lynn said during his life, they assume he must have fudged the numbers to make Africans look bad.

But there’s another reason. The quality of the data from Sub-Saharan Africa is lower. There are fewer large representative samples with adequate information for norming. Obviously, if the data from Sub-Saharan Africa were of high quality, there would be less room for dispute over the resulting average. But since they aren’t, different researchers can interpret them in different ways.

It’s important to note that this problem is hardly unique to IQ. For practically any variable you might care to measure, the quality of data from Sub-Saharan Africa is lower. Which shouldn’t really be surprising. In societies that are poor, highly corrupt and often stricken with civil war, it’s harder to get your hands on large, representative samples. As Garett Jones likes to say, “developing country IQ data is about as reliable as developing country GDP data”. In 2010, Ghana revised its GDP figure up by 60%. In 2013, Nigeria revised its GDP figure up by 89%. Naturally, this doesn’t mean we throw out the African GDP data.

So, who is right: Lynn or Wicherts? I’m inclined to say Lynn. This isn’t because Wicherts is an activist whose opinion I don’t trust. Unlike most of Lynn’s critics, he’s a serious researcher and an expert in psychometrics. (He once reviewed a paper I submitted to Intelligence under his own name, and all his comments were fair and reasonable.) Rather, it’s because other measures of cognitive ability derived by economists also point to an average IQ for Sub-Saharan Africa of about 70.

The latest example is Sarah Gust and colleagues’ measure of “student achievement”, introduced at the end of 2022. These economists combined data from various student assessment studies (such as PISA and TIMSS) to calculate average test scores for 159 countries. The figure for Sub-Saharan Africa was 303. The figure for the UK, which is often used as a benchmark, was 503. Since the standard deviation is 100, this means that Sub-Saharan Africa scored two standard deviations below the UK. Two standard deviations is 30 IQ points. So if the UK’s average IQ is 100, Sub-Saharan Africa’s is 70.

Incidentally, the UK was by no means the highest-scoring country. That accolade goes to Singapore, which achieved an average of 560. At the other end of the scoreboard, we have Niger with an average of just 207. That’s three and a half standard deviations (or 53 IQ points) lower. Clearly, very large differences in test scores between countries are possible even when none of the data come from Richard Lynn. So far as I’m aware, no one has accused Sarah Gust and her colleagues of having a racist agenda.

By way of illustration, the chart below plots national IQ against student achievement transformed onto the IQ scale. Countries in Sub-Saharan Africa are shown in red. The grey line is y = x.² As you can see, countries in Sub-Saharan Africa are distributed either side of the grey line, rather than being concentrated below it. Their national IQs are not systematically lower than their student achievement would predict.

The relationship between national IQ and student achievement (r = .81).

The next chart plots national IQ against student achievement but replacing national IQs for countries in Sub-Saharan Africa with the value of 80. These countries are now concentrated above the line. Their national IQs are substantially higher than the national IQs of other countries with similar student achievement. In addition, the overall relationship is weaker.³

Replacing national IQs for countries in Sub-Saharan Africa with the value of 80 (r = .73).

At this point, Lynn’s critics may raise the objection that scores derived from student assessment studies cannot be used to validate average IQs because those studies do not actually measure cognitive ability but some distinct construct, such as educational attainment.⁴ There are several responses to this.

The first is that Wicherts himself used such scores to argue that Lynn had underestimated the average IQ in Sub-Saharan Africa. His conclusions were different because fewer data were available in 2010, the methods for combining them had not been refined, and he used the regression line as the yardstick rather than y = x. The second is that a general ability factor accounts for the vast majority of variance in students’ PISA scores. As Russ Warne noted in a recent paper, “This supports Lynn and his colleagues’ decisions to use educational achievement tests as a source of data for estimating mean national IQ.” The third is that intelligence and educational achievement are, in any case, strongly correlated at the individual level.

To be absolutely clear, 70 is not an estimate of the genotypic IQ of Sub-Saharan Africa. We are simply talking about average measured IQ. Lynn explicitly argued that Sub-Saharan Africa’s genotypic IQ is around 80. Whatever the true figure, it is undoubtedly higher than 70, with the difference being explained by factors like health, education and familiarity with testing. Nobody is suggesting that under ideal conditions the average Sub-Saharan would achieve a score of 70. Ironically, it is environmentalists who tend to conflate measured IQ with genotypic IQ in this context, even though they’re the ones who believe intelligence is so heavily influenced by the environment.

Indeed, it really doesn’t make sense for them to argue that the average IQ in Sub-Saharan Africa is as high as 80. We already have abundant evidence that black Americans score about 85 on IQ tests, as compared to 100 for whites. If the average IQ in Sub-Saharan Africa is 80, this would mean the massive difference in environment between Sub-Saharan Africa and the US reduces IQ by only 5 points, yet the comparatively small difference in environment between black and white Americans somehow reduces it by 15 points. The point has been made a number of times, including by Lynn himself, but environmentalists still don’t seem to get it.

Which suggests they aren’t trying to build a comprehensive model that can explain all the data. They’re simply dealing with each datapoint on an ad hoc basis. Interestingly, three prominent environmentalists once compared themselves to “members of a fire brigade, constantly being called out in the middle of the night to put out the latest conflagration, always responding to immediate emergencies, but never with the leisure to draw up plans for a truly fireproof building”. It’s now forty years later. The “fire brigade” is still being called out and plans for a “truly fireproof building” remain on the drawing board.

In summary, Lynn’s estimate of the average IQ in Sub-Saharan Africa looks to be about right. This doesn’t mean his work on national IQ is unimpeachable. A key remaining issue is the lack of explicit inclusion and exclusion criteria for samples. However, it does mean his critics’ main complaint is wrong. And it completely undermines the case for “reviewing” all his articles, let alone retracting them.

Noah Carl is Editor at Aporia.

---

Consider supporting Aporia with a paid subscription:

You can also follow us on Twitter.

1. No one is outraged that he estimated the average of IQ Japan at 107.

2. I used Lynn and Becker’s measure of sample size and quality-weighted national IQs, dubbed “QNW”.

To transform student achievement onto the IQ scale, I divided each estimate by 100, multiplied by 15, and then added 23.97 to make the sample means equal. After transformation, both variables had a mean of 86 and a standard deviation of 12.

I believe y = x, rather than the regression line, is the appropriate yardstick, since the latter assumes that student achievement has been measured without error.

3. When replacing national IQs for countries in Sub-Saharan Africa with the value of 70, the correlation is almost identical to the original one (r = .80 versus r = .81).

4. A recent paper claimed that measures derived from student assessment studies are “not well-suited as proxies of intelligence”. However, it doesn’t stack up.