IQ, the Flynn Effect, and Africa
Over at Aretae's blog a discussion has flared up about IQ, the Flynn effect and Africa. It is fairly well established that the typical African IQ score is on the order of 2 standard deviations lower than the typical American. The question is what this means for governance and growth. Aretae argues that based on the Flynn effect, Americans in the 19th century probably had a similar IQ to the average African today. Therefore it is doubtful that low African IQ is the cause of Africa's woes. My reply exceed blogger's character limits, so I post here instead.
Both IQ and the Flynn Effect are very poorly understood. The first point to correct is this piece of conventional wisdom, repeated by Aretae: "On every test we've got...and every way we know how to measure things, IQ has been going up for at least half a century, and probably a lot longer than that." This is wrong.
The gains in scores vary enormously by particular test. For instance, picture arrangement has gone way up, while arithmetic has barely budged. See this chart:
Let's step back a moment and think through what IQ and intelligence mean. To do so, we will discard imprecisely defined words, and introduce three new terms. We can define three aspects of what we commonly call "intelligence":
cognitive power - the ability to solve novel problems involving abstract reasoning. What we commonly mean when we say, "he is sharp, he understands everything we explain to him really fast".
specialized cognitive skills - a specific skill involving abstract symbol manipulation - computer programming, calculus, writing poetry, etc.
base cognitive skills - a skill of abstract symbols that form the base of more complex skills - arithmetic, vocab, reading comprehension, etc.
Cognitive power is really hard to measure directly, because you have to control for how much the person already knows. The art of designing an IQ test is to create a test of a skill that most people have roughly equal exposure to. If everyone has roughly the same exposure to a cognitive skill, then differences in ability will tend to be a function how quickly each person learns - cognitive power.
In general, then there are two ways to design an IQ test: a) give a base cognitive skill test for a common skill that everyone in the group has large amounts of exposure to (vocab, arithmetic, word matching) b) give a base cognitive skill test for a skill that no one really practices (raven matrices, block manipulation).
Thus an IQ test is a test of a base cognitive skill, that proxies as a test for cognitive power. It will be a better or worse proxy, depending on how even the exposure to the cognitive skill is. If we give an English vocab test to both Americans and Japanese, that test will be a horrible proxy. But if we give a game of Tetris to both Japanese and Americans, it might be quite a good proxy. If we give that game of Tetris to Americans from 1950 and Americans in 2010, then it will suck as a proxy.
I don't think the Flynn effect represents a real rise in cognitive power. a) the rise in scores varied greatly by test, we'd expect the opposite if cognitive power rose b) for the scores that rose the most, I can think of plausible explanations. There is much greater test awareness today, I grew up playing block matching games both on paper and on the computer. But since knowledge of basic English and arithmetic were already saturated, there has not been much change on those tests. c) from talking with family members, I think my parents and grand parents were every bit as quick as my siblings and cousins d) if you look at the speeches and newspapers from 50+ years ago, they are just as sophisticated as our modern papers.
So my conclusion is that Americans of 50 or 150 years ago had similar cognitive power, and similar cognitive skills in the areas that matter (literacy and basic math). Thus they were able to create high levels of governance and growth.
As for Africa then, we have no idea how much of the 30 point IQ differential is due to lack of education in cognitive skills, how much due to environmental deficits lowering cognitive power, and how is due to genetic differences in cognitive power. But it is quite clear that, unlike 1850's Americans, Africans have substantial deficits in very relevant cognitive skills (literacy and math), not just test taking skills (like the Raven's matrices).
From the perspective of good governance and growth, the exact reason for the 30-point differential doesn't really matter. Whatever the reason, finding an internal elite to govern the country wisely and drive entrepreneurial and technological growth will be extremely difficult. The problem is made much more difficult by various historical events and the introduction of forms of government entirely unsuited to Africa's demographics and culture.
Comments
|
1 point by Devin Finbarr Jul 20It's difficult to make a judgement about the overall sophistication of the population from reading speeches, because you have to adjust for what portion of the population was actually hearing/reading those speeches. Speeches in the 1800's were perhaps mostly "heard" via newspaper, and thus may have been targeting a more sophisticated audience. Speeches today are meant to be seen on TV and thus are targeting a much dumber audience. Worse, speeches today are often meant to be replayed as sound bites, making them even more simplistic. |
|
|
2 points by aretae Jul 20Your comments always exceed the limits. :-) My big problem is that the Flynn effect Seems to be concentrated in those intelligence tests that are most heavily g-loaded, or in your terms, "cognitive power". My understanding is that the evidence goes precisely against what you're suggesting right now, and that your "cognitive power" is precisely the thing we have the most evidence for increase in, while skill is highly different. Second, g is highly correlated with all positive outcomes in life. If g is going up, it means people should be better at everything that there are incentives to be better at. And I'd bet that basic arithmetic isn't a weakness (poor, unschooled folks usually do great with all useful, self-benefiting forms of arithmetic -- better than Americans most likely, on anything concrete and near used skills), though probably literacy is weaker. My conclusion is that Americans of 150 years ago had much lower cognitive power, better skills in areas that matter ( due to no negatively-effective schooling, strong self-reliance, home-education, relevance to life), and access to the Anglo-enlightenment traditions of newspaper reading, low government involvement, individual bootstrapping, small clan-less families, and big anti-corruption norms. And that the traditions were most of the difference.
|
|
|
2 points by Devin Finbarr Jul 20My big problem is that the Flynn effect Seems to be concentrated in those intelligence tests that are most heavily g-loaded, The core mistake the intelligence researchers make is assuming that 'g' loading is a property of the test, rather than a property of how the test interacts with the environment. So as I said in my post, if you are comparing across two modern countries with different languages, a Raven's test or a game of tetris might be most g-loaded. As in, the Raven's will most accurately correlate with what we think of as "cognitive power". But if you are comparing Americans from the 1950's to the 2010's than a vocab test might be the most g-loaded. Using tetris as a test will suck. Americans of today have far, far more exposure to visual puzzle games, so it makes sense that we would score better on those tests. No one can numerically calculate what the most g-loaded test is for comparing 2010 Americans to 1950 Americans. We have to look at a variety of environmental factors, and make a subjective judgement about which test was least affected by environment. 20 IQ points is a huge difference. From the newspapers, movies, memoirs, that I've read it's utterly implausible that Americans in the 1930 had an average cognitive power equal to a person that we now think of 80 IQ. Sure schools suck now, but it's not like they were great in the 1800's or even early 1900's. Most people got just enough school to read and do a bit of math. From then on they were self-taught. It's pretty much the same today, except that we have to waste far more hours in school. Most people get their vocab |
|
|
1 point by Andrew Jul 20Yes. Spectacular post + follow-up attempt at making *yet more clear* the game of isolating cognitive powers between peoples. I challenge Aretae to show he's actually understanding rather than -- or at least before -- digging in heels. :P |
|
|
2 points by Koanic Jul 20In my opinion, the black maximum genetic potential average IQ is 85. The lower scores seen in Africa are due to a lower genetic IQ in one subregion, pygmies I think, and malnutrition + disease for the rest. I think you're underestimating the partial reality of the Flynn effect due to improving nutrition in the West and other places for the poorest. I think your explanation of Flynn as a testing artifact is interesting. It would be very difficult to control for historical test exposure. |
|
|
2 points by Devin Finbarr Jul 20Based on my limited knowledge of biology, there is no reason to think a priori that limited nutrition would affect height/weight and cognitive power equally. In fact, it makes sense that the body would restrict resources for height long before it would restrict the nutrient flow to the brain. So a priori I don't think that nutrition would necessarily affect cognitive power. And for the reasons I listed in my post, it seems there is more evidence on the side that cognitive power has not changed much at all. |
|
|
1 point by Andrew Jul 20I'm curious about this thought: "it makes sense that the body would restrict resources for height long before it would restrict the nutrient flow to the brain." My first thought is it depends on where the architecture of DNA expression is most flexible, rescalable. When signaling what to grow and how much in a body, are there more intricate dependencies involved in brainbuilding or in heightchoosing? We "know" (I haven't personally researched it) that adult height is much affected by childhood nutrition, so that's at least significant evidence for blueprint flexibility on height and therefore for your thought. Actual people obv vary (from each other, if not from that their other possible, differently raised selves) however much as they do on both dimensions, though... |
|
|
1 point by Devin Finbarr Jul 20My knowledge of this area is mostly just speculation. I've heard a person argue before that different parts of the brain (verbal vs non-verbal for instance) may respond differently to lack of nutrition, which might explain the the weird Flynn effect scores. I'm not sure how much evidence there is for that claim. |
1 point by Foseti Jul 20
I also think the Flynn effect is probably non-existent. Either way, it's wrong to rely on it too much given that it is still so misunderstood.
I'd probably go further, if pushed. Devin says: "if you look at the speeches and newspapers from 50+ years ago, they are just as sophisticated as our modern papers."
I read old speeches, books and newspapers regularly. If one does this, one is struck by how much more sophisticated they are than their modern equivalents - they're so much more sophisticated I hesitate to use the word "equivalent."