Chanda Chisala and I post about each other so often that we should be employing the same agent. Properly managed, I might finally get onto a lecture tour circuit somewhere. The Shetlands, perhaps.
Below is the post to which I am referring.
My reply to Chisala’s post has hardly been prompt: one source of delay, among others, was posting about a massive new study by Hur et al. (2017) on Nigerian children in public schools which has shown early male advantage in intelligence scores, and an overall result which is unchanged from previous estimates, but is far stronger in terms of sample size and representativeness.
Whilst one can always wish for better coverage, particularly of private schools, the current picture is that on Raven’s matrices the new Nigerian generation has an IQ of 70. This is relevant to the calculation of the internationally judged smart fraction in Nigeria, in that it reduces its size and, if the test results are true, makes exceptional intellectual performance less likely in that country, and thus makes demonstrated Nigerian achievement even more of an exceptional case, thus potentially challenging the predictive value of intelligence testing. For the avoidance of doubt: a confirmed low score on intelligence tests placed against good performance in international competitions involving intelligence calls the intelligence test results into question.
Chisala is testing the veracity of African country-level intelligence test results by finding excess African performance on international Scrabble competitions. I am in favour of the general approach of looking at intellectual competitions, because real life achievements are the real test of ability, and intelligence tests are merely predictive instruments. Curious as it may seem, we are in agreement on that matter. I would prefer that the real-life tests of achievement were studied across a broader range than Scrabble, including Chess, Maths competitions, numbers of patents, innovations in science and technology and so on, but I am sure that Scrabble at competition level requires high intelligence.
I am also sure that general relationships are just that: general. There will be outliers and exceptions, positive and negative residuals on the overall trend line, but if there are too many then there will not be a general relationship. So, although the argument largely centres on Nigerian examples, the overall assessment between intelligence tests and actual achievements should cover all countries.
Chisala believes that evidence of exceptional performance is more damaging to the hereditarian than the environmentalist position. He says:
My argument is therefore not against the low IQ score estimates for African nations (by Richard Lynn, et al), but whether this reflects some restrictive racially linked genetic cause. If it is indeed basically genetic, it should practically be impossible to find any area of relative cognitive performance of Africans that is inconsistent with this large IQ deficit with whites and other groups.
If, on the other hand, the cause is basically environmental (specifically, learning resource deficiencies), then some exceptions are bound to exist and these will predictably only be found in areas in which the cognitive challenges are high but the learning resource requirements (books, well-trained teachers etc) are extremely minimal. Performance on such cognitively demanding but bookless contests will far exceed even academic areas that are light on cognitive demands but heavy on book learning (eg soft subjects like sociology etc, where you still find no Africans at the top). The genetic Racial Hypothesis predicts that the gap should be even bigger in favor of whites as you go to the more naturally complex contests like Scrabble (see Spearman’s Hypothesis.) In short, if there was to be any exception to inferior African intellectual performance, it should have been in the softer fields where there are less of the gifted math types to contend with.
I can see his general argument, though geneticists argue that the degree of heritability does not equal the norm of reaction, and a high heritability does not mean that a trait is unaffected by its environment. I see this discussion in a simpler light: if many countries show abilities which are higher than predicted by intelligence testing, then the validity of those test results are called into question. To be clear, I agree that is a valid argument against intelligence assessments at the country level if actual achievements are generally poorly predicted.
Now, performance above prediction could be from genetic causes if there were sub-groups within Nigeria (or any other country) with exceptional abilities. They need not be particularly numerous within Nigeria, any more than Jews need be numerous to be clearly above average in Europe, but they can have disproportionate success in intellectual competitions. This is the observed pattern in Nobel prizes and so on which confirm Jewish superiority in intellect compared to other Europeans. However, the genetic interpretation of exceptional performance could only be argued if it could be shown that Nigerian Scrabble or Chess or Maths competition winners were disproportionately drawn from some especially talented minority group. As far as I know that hasn’t been established, or not yet anyway.
I agree with Chisala’s other points about “softer” fields not being the best measure of national ability, though they require ability which I would probably rate higher than Scrabble, and we already have evidence of that. For example, Nigerian Wole Soyinka won the Nobel for literature in 1986.
However, I do not agree with his dismissal of the chess results as being due to a lack of training in Africa, particularly when Chisala then goes on to make a point about Kenny Solomon of South Africa being a Chess Grand Master. FIDE does not say that, because he has not reached their threshold requirement of 2500 points. Wikipedia says: “Although Solomon has never reached the required rating of 2500, he earned the Grandmaster title by winning the African Chess Championship in December 2014, thereby becoming the first chess grandmaster from South Africa and the second one from sub-Saharian Africa after Amon Simutowe.” This could have made clearer, and it shows how important it is to set thresholds to begin with, or one mixes local rankings with the ones that matter most, the international open competitions. That Africans can be excellent chess players is not in question, simply that determining Grand Masters depends on using standard ranking procedures.
The big advantage of using chess to look at the general cognitive ability of countries is that it is non-verbal competition with a clear win-lose-draw scoring system, and very wide representation across the world. FIDE has 188 member federations, WESPA (English language Scrabble) seems to have 23 nations. This makes Scrabble a slenderer basis for international comparisons.
The top 100 Chess grand master FIDE list does not appear to include an African, but I have not gone through all the names in detail, so please do so for me.
I have looked at the FIDE total scores for the top 10 players in each country, which gives a broader sample of capability. It is worth looking at this list, which allows both genetic and cultural interpretations.
Russia is No 1, followed by USA, China and then Ukraine. With a mere 3.3 million person Uruguay manages to slip in at No 75, but Faroe Islands achieves No 81 with a mere 48,199 hardy islanders, and that is despite mercury in their whale meat (traces of, but I will post about that later).
On the African continent, the first countries to come in are: Egypt 49th, Algeria 68th, Morocco 70th, Tunisia 74, South Africa 77, Zambia 88, Nigeria 91. Gabon with 1.8 million inhabitants is there at No 157, a ranking similar to Sierra Leone, Guernsey (population 63,000), Mozambique and Somalia. There is lots to discuss here, and showing all these countries by IQ and population would give a general test of the country intelligence/chess success link.
Here are the South African top players
Here are the Nigerian top players
In sum, I think that chess is a valid indicator worth studying, and Nigerian chess performance is not disproportionately good, given population size.
Back to Scrabble. Here is some background on the intellectual characteristics of Scrabble players: Toma, Halpern and Berger (2014) found that top scrabble experts have “extraordinarily high levels of visuospatial and verbal working memory capacities” and score 1.23d higher than elite college students who scored at the 93th percentile of the quantitative SAT.” We are agreed that top level Scrabble players are very bright.
My approach in testing the claims of out-of-the-ordinary Scrabble achievement was to make minimal assumptions: assume that the country intelligence estimates from Lynn or Rindermann were what needed to be tested (but now see the Becker edition updates); set the standard deviations to a common 15 points; take the UN estimates of total population; then calculate how many citizens of that country were above a certain criterion, which I placed at IQ 130. I also looked at IQ 140, but the IQ 130 figures are sufficient to illustrate the general point. I could have spelled this out in more detail, but I repeat the general principles of this approach as set out in my previous post:
1) A threshold of performance must be found that depends purely on cognitive ability.
2) The threshold must be unequivocally defined.
3) The rates at which the target group and the control group cross the threshold must be established.
4) Opportunity must exist for both groups to cross the threshold. That is, the target group cannot be restricted by political considerations from fair competition.
5) An overriding motive to cross the threshold must exist, being sufficiently strong that virtually all those capable of crossing the threshold do.
For a really productive discussion, we need to answer those points.
1) We need to agree what the threshold is. It could be getting to the finals or winning the most prestigious and representative Scrabble prize. We should find a threshold which could also be applied to other competitions and achievements.
2) The unequivocal threshold must be defined. Perhaps it should be the top 20, 30, 50 or 100 players in the world each year. I would favour the top 100 to make it easier for Africans to enter the lists. In the case of Scrabble, we might have to decide whether the competition should be in English or French, since this could influence African country totals. We need to agree a figure and a method. Top 100 players in each competition? If most countries do not participate in Scrabble competitions this limits the conclusions that can be drawn from the competitions.
3) The rates at which the target group (Nigeria or Gabon or all of Sub-saharan African countries) cross the threshold and the control group (Europe; US/Canada; Asia; rest of world?) cross the threshold need to be agreed. This should be easier, because there are lists of competition winners.
4) Are there restrictions which impede competition? Such restrictions are probably mostly experienced in Africa, but national training programs may be a contrary factor. Overall, they might predict African under-performance because of resource restrictions. Africans who move abroad could be an exception.
5) Are people motivated to play the game in question? The motivation is very probably not as uniformly strong as playing Chess in the USSR, so this is a tricky one to resolve. We would need to study the numbers of people playing competition Scrabble and Chess and other games in each country. It might be possible to estimate this. It would appear that chess is far more widely played than Scrabble.
The best form of comparison is to look at country achievement levels in the most important competition, and to work back the calculations from there. Here are the Scrabble national winning teams from the World Championship History 1991-2016: United States 7, England 6, Canada 5, Thailand 5, New Zealand 4, Australia 1, Malaysia 1, Nigeria 1. The first thing which springs to mind is that this is hardly a globally represented game, as measured by the winning teams. However, these give us threshold scores for a scattering of participating countries, and a basis for doing some calculations.
First of all, the correlation between county IQs and number of national wins is r= 0.498 which is indicative, but moderate in size. The correlation between population size and wins is r= .336 which is lower. The correlation between wins and smart fraction (Greenwich IQ 130+) is r=.794 which is strong and unlikely to be a fluke. I think it safe to say that there is a general relationship between national intelligence and national wins on Scrabble competitions when one takes population size into account. Nigeria is punching above its weight, in that it draws its Scrabble champions from a much smaller group of smart people, but it is not an impossibly small group. It is a separate question as to whether they should be spending their time on this game, but as you know, “I do not do policy”.
A few points about calculating the numbers of IQ 130+ people in each country. The simplest technique is to work out those figures for the whole population, since the numbers can then be applied to all intellectual endeavours: Maths, science, technology, Chess and Scrabble and all else. Of course, each country will have different age structures (some like Nigeria will have younger populations than others) but the same calculations should be applied to all countries.
By the way, I think it a mistake in reasoning to say that if Scrabble winners turn out to be mostly men we should not include women in our calculations. I think it is better to calculate the number of bright persons regardless of sex, and then see how they employ their talents. Women may start winning more of these competitions, and it is easier to apply the same calculations to all countries then to argue backwards from the particular characteristics of the winners of each game.
Chisala suggests that my estimate of the number of bright Nigerians be reduced in the following ways: cut my estimate in half to account for toddlers; further cut by more than half to exclude women; and further cut by half to account for half of Africa’s intellectuals having moved away from Africa. I think these reductions are excessive, and in excluding women from the general calculation of the smart fraction (because they are not in the top ranks of Scrabble players) mistaken, but I am following his line of argument just to do the figures. If Scrabble requires IQ 130+ then those three cuttings-in-half bring the smart fraction down from 5764 to 720 persons. If IQ 140 is required, then dividing by half three times brings the smart fraction down from 1336 Nigerians to 167. Chisala gets the number down to 33 so he must have used a higher discounting. I do not accuse him of “reverse engineering his concealed steps” which he levelled at me when I set all standard deviations to a standard 15 points. He has discounted in a different way, and can explain the precise reductions in our later exchanges, which may be the basis of a major movie someday, if our afore-mentioned agent earns his keep.
Chisala also asks me to do this calculation for Gabon, which has a much smaller population than Nigeria.
I don’t know why Dr. Thompson neglected to show us how his calculations would work on Gabon (the country that was most prominent in my own rough statistical argument), which has less than 2 million people and an IQ of 64 but regularly produces top world championship Scrabble players. As a defender of your hypothesis, you should normally tackle the hardest cases to show how they happily survive the biggest hits from the opposition. The math fails miserably for Gabon, even without a single correction to his assumptions.
It Gabon IQ were 64 the smart fraction would be as follows: Taking Gabon’s population as 1.8 million the IQ 130 smart fraction (standard deviation 15) would be 428 persons. Above IQ 140 would be 95 persons. However, there is a problem with calculating the IQ of Gabon. In previous publications the IQ has been given as 64, and in other places at 69, but on the Becker edition of the Lynn database we no longer have an entry for that country, because we do not have a reference which comes up to standard. It is time for a researcher to test mental ability on a large sample in Gabon.
In order to discuss all these matters (exceptional performance by individual countries) we need to carry out the same calculation for all countries, which is to work out the smart fraction for every country, and use that as the base rate for all intellectual prizes, including those achieved by women.
On the question of standard deviations, although I set them all to a common figure, I agree with Chisala that my uniform use of 15 points can be questioned, and I like his argument on this matter. A smaller standard deviation of 12, as found in African Americans by Arthur Jensen in his 1998 “The g factor” summary would give much lower figures for the Nigerian and other African estimates of numbers of smart fraction. This is a good test of whether the narrower spread of ability found in African Americans is also, as one might expect, found in Africans in Africa.
However, it was worth waiting for the new massive sample of Nigerian school children on the Standard Progressive Matrices Plus test, which finds an IQ of 70. With a raw score average of 21 and standard deviation 8.9 points, there is clearly a very wide dispersion of scores in Nigerian teenagers. The low mean score creates a slight problem, in that there will be restriction of range at the lower end. David Becker, who is in charge of the new database (see previous posts) shows that for 15 year olds (the older ages show a similar picture) the 8.9 points standard deviation of the raw scores translates to an IQ standard deviation of about 16 points below the mean, and 25 points above the mean for males; and 16 points below the mean and 20 points above the mean for females. David Becker adds: “It is always a problem in the case of nations with low scores that in this part of the scale, one raw score difference can cause 5 to 7 IQ scores difference. So even the rounding rules have a relatively huge impact.” In the case of Nigeria this massive study shows that a standard deviation of at least 15 IQ points seems appropriate. It would also be good to check all other African data, and to get more up to date figures for African Americans.
In discussions with other researchers I am trying to encourage more attention being paid to the distributions of scores, and the standard deviations. It will be good to gather new data on the standard deviations found currently in African Americans. Standard deviations are informative, and raw score distributions even more so.
Chisala also raises other matters. He points out that high African performers in the US often come from Africa itself, not local African Americans, and if the latter are brighter because they are about 20% white, that should not be the case. I think the answer lies in the selective migration of the brightest Africans. If Chisala really wants to cut my estimate of bright Africans in half, accepting The Economist’s calculation that half of them leave Africa (though some return to high posts in Africa), he cannot both do that and then be surprise when many of them end up in the US or UK.
To summarise: if we agree what constitutes good performance on a broad range of intellectual competitions, and if we measure every country in the same way to calculate their smart fractions, we seem to find a general correlation between intelligence and achievement. There are outliers, both positive and negative above and below the trend line, as with all demonstrated correlations.
I cannot show that there are African intellectual elites who have developed higher intelligence than all other Africans, though this is a perfectly plausible hypothesis, particularly for those like me who believe that there is a genetic component in ability differences. It might be the case that there are such elites, and it would be good to search for them. For example, the new Hur et al. (2017) study should be able to detect them in Nigeria, using the background data on the children. The authors should be encouraged to take a look at this.
Above all, no predictor variable can triumph over a demonstrated reality: if the general relationship between ability measures and demonstrated achievement does not hold, then the ability measures are poor predictors. At the moment, they appear to be reasonable predictors, and are the best available at the moment. They might be able to be replaced by better ones, but that remains to be demonstrated.