A short blog about a blog: more reasons why national IQ databases are not useful

A recent blog by Russel T Warne (https://russellwarne.com/2020/07/19/thoughts-on-low-national-iqs-intellectual-disability-and-data-quality/) makes some sensible points that appear to have been prompted by some of the furore associated with the now-retracted Clark et al paper that colleagues and I critiqued recently. He begins by saying that IQ scores do not measure intelligence with great validity, that IQ scores are susceptible to many influences including bias, practice, education etc, but can be useful when interpreted carefully, in context, with consideration for culture, language, visual and auditory acuity and history, as one must do when using a cognitive ability test clinically. I entirely agree. I infer from reading his blog that the author also believes that cross-national comparisons would be particularly prone to these problems. Again, I would have to agree. He goes on to complain a bit about critics of national IQ datasets, saying that if they are unhappy they should go out and collect better data. I agree with Will Gervais (see here) that this is not only irrelevant to the science, but is also just not how science works. Critics may well choose not to do that because they have little faith such research would be useful, and/or they have other more pressing problems to work on. This is obviously a perfectly legitimate position to take.

Anyway, never mind…. As I said, this was a rather tangential point. The author goes on to say “I do not think that average national IQs are worthless… I think that these scores are an approximate rank order to how well the citizens of each nation have been trained to solve formal cognitive tasks”. In other words, they are basically a measure of the outcome of a certain kind of formal education. This would be consistent of course with ample evidence that IQ test scores improve with practice, can be taught, tend to increase over generations and so forth. With that sort of interpretation it would be natural to argue that studies of associations between country-level IQ and, say, economic activity, point to the importance of education for economic development. Hard to argue with that.

Funnily enough, the defenders of these IQ databases, who do seem to believe they capture intelligence, often point out that national IQ scores correlate strongly with school achievement. They see this as evidence of the validity of the national average IQ scores. I have always found this odd, because IQ tests are not usually considered to be measures of educational achievement, so this either undermines the validity claim, or supports the claim that national IQ scores validly measure something else. The latter is indeed what Russell Warne argues in his blog. So, the simplest interpretation of such correlations is that they provide evidence that cross-national IQ differences substantially reflect educational differences.

I should point out at this stage that I completely agree with many commentators, including the authors of the retracted Clark paper, that much of the data in the infamous Lynn/Becker etc database is laughable scientifically, particularly for African and middle/low-income countries elsewhere. But that’s not the point I wanted to explore in this blog.

So, setting that aside for a moment …

As someone who would certainly never claim to be an IQ expert, I was quite surprised to discover that even if you restrict your focus to Europe, where I would guess the data quality for national IQ is better than in other parts of the world (though probably far from perfect), national average IQ still correlates quite highly with national age 15 PISA reading scores (from the World Bank database). See the graph below…

I wouldn’t want to make strong claims about the validity of the IQ data here I must underline – I have taken this from the data used by Clark and colleagues, and have not done exhaustive checks on the provenance of the national IQ estimates for Europe. Nevertheless, the PISA reading scores are likely to be reasonably valid. However you look at it, it seems quite clear that these national IQ scores are closely linked to the quality of a country’s education system, as they closely track the reading achievement levels of a country’s average 15 year old.

Bear in mind that, given what we now know, national differences in any complex trait are extremely unlikely to reflect causal genetic differences. But, to put it bluntly, across European countries such a claim would surely be even more absurd. Bulgaria compared to Hungary or Poland? Or Greece compared to Estonia? There are roughly 10-point differences in national IQ here, apparently, and modern-day eugenicists want us to believe that differences like this are mostly due to genetics. Setting aside the fact that no experts in population genetics take these kinds of eugenic claims seriously (e.g., see some of Ewan Birney’s comments here), I find these data alone extremely difficult to reconcile with them. So, I would take all this as further reason to believe that IQ test performance at national level, to the extent it is meaningful at all and when it is measured reasonably well, largely reflects social/systemic factors like educational quality (as well as some bias, noise, etc), and not intelligence as most of us would understand it. That they correlate with other country-level indicators like economic development (e.g., PISA reading score in Europe correlates .74 with GDP), civil society and so forth just points to what many progressive thinkers and indeed sensible governments believe – namely that education is critical for economic and social development.

So does this lead me to be favourably inclined to national IQ data? No. Firstly, the data that exist currently are, as many have noted already, mostly very dubious, particularly for less wealthy parts of the world, and in many countries they are so bad as to be almost funny if they weren’t so damaging. But what if sampling and measurement issues could be resolved? Still no, for two reasons. First, right now too many actors in this area are desperate to torture IQ data to make anti-progressive and often overtly or covertly racist claims. You can never divorce science from politics and ethics, and I would not want to be a part of something that could be so misused. Second, if Russell Warne and the arguments I made above are more or less correct then we have much better, more reliable, more meaningful and mechanistically much more informative ways of measuring national educational quality and achievement, to help us understand their role in social and economic development. IQ adds nothing, other than confusion and the potential for misunderstanding and misuse.

4 thoughts on “A short blog about a blog: more reasons why national IQ databases are not useful

  1. We also have PISA and TIMMS scores for some lower income countries. They correlate well with IQ scores for those countries, much like within Europe.

    We also have PISA scores for racial groups within the U.S.: once again, they correlate well with average IQ scores for the different racial/ethnic groups, just as the federal government’s NAEP school test finds the same racial patterns as IQ tests.

    The most plausible explanation is that both school achievement and IQ tests are influenced by both nature and nurture. Similarly, higher IQs and higher school test scores contribute to higher per capita GDP, but also that higher per capita GDP contributes to higher test scores.

    1. Thanks Steve. Yes, absolutely, especially agree about the circular relationships between education and social and economic development. The main point for me is an old one: behavioural geneticists have always pointed out that heritability within a population says nothing about causes of between-group variability. The European data just provide a very clear example of that – there is no way those country-level differences are caused by genetic differences. My favourite example was Greece (now comparatively poor EU nation with struggling state infrastructure, but birthplace of contemporary civilisation), versus Estonia. Very large differences in PISA scores; obviously absurd to claim genetic differences explain any meaningful proportion of that.

      1. “there is no way those country-level differences are caused by genetic differences”

        Perhaps, but how do we know? For example, there are differences in average height between European countries.

        Some of that is undoubtedly due to nurture: The Dutch have fairly recently sprouted up to be very tall, but I don’t recall any stereotypes about the Dutch being tall when I was young 50 years ago. So something the Dutch have been doing in recent generations appears to have increased their height for non-genetic reasons, although nobody was sure what they are doing different than say the Danes the last time I checked into it.

        On the other hand, the much poorer Balkan peoples were tall even back then: in 1950 anthropologist Carleton Coon published a study of the Ghegs of northern Albania entitled “The Mountain of Giants.”

        How materially deprived were Albanians in 1950? There’s an anecdote from WWII: two American soldiers drove their Jeep far up into an Albanian village, where the headman greeted them with a feast. When they came out, they found their Jeep had been turned over. But the villagers weren’t hostile, they just explained, “We wanted to find out whether you were riding a boy Jeep or a girl Jeep.”

        I can recall Bill Walton’s toughest opposing center in the NCAA basketball tournament in the early 1970s was BYU’s Yugoslav import Kresmir Cosic.

        And today, now that they’ve had a couple of decades of peace and capitalism, Balkan people are probably the best white athletes in the world. A large fraction of the top white NBA players now have names ending in “-ic.”

        People familiar with the Balkans tell me that the really tall and robust people aren’t the lowlanders from the river valleys but instead are the giants from the mountains.

        So, apparently, genetic differences can have real world effects within Europe, especially where mountains are barriers to gene flow.

      2. No doubt genetic differences make a difference to individual outcomes. But do they explain big difference in country-level outcomes? Not impossible that’s true, just very unlikely (in my opinion) – there is enormous genetic mixing going on in Europe and has been for hundreds if not thousands of years. Many pockets of complex admixtures of genetic ancestries within countries too, and lots of evidence of relatively short-term changes in educational outcomes dependent on schooling policies and economic circumstances. You can believe it if you want to, but I have a very hard time believing that the Finns are just naturally genetically better endowed for intellectual work than the Greeks, Italians or Moldovans – all the evidence points away from that, even if far-fetched genetic hypotheses can be entertained. Finland was one of the poorest countries in Europe pre-WWII, with low educational standards – huge improvements in the last 60 years, much of it in the last 20. Do you know of any credible genetic evidence that explains these sorts of national differences (and changes)? Not twin studies – they don’t say anything about between-group differences or temporal changes)? I do like the Dutch height example – as you say – it’s a recent phenomenon. Good example actually of the fact that within-population heritability says nothing about change or between-population variability. 80% heritability within a population, but 20cm increase in height over 150 years. Some modest proportion of that could be genetic selection, but not much I would think.

Leave a Reply to Steve Sailer Cancel reply

%d bloggers like this: