It is good when an article giving data on African scholastic attainment and intelligence draws attention and comment. Data is what we thrive on.
There is much in this article to be discussed. I left an immediate comment which, for ease of reference, I copy below.
Thanks for your article. I think that the problem which bedevils much research on intelligence is the difficulty of getting representative samples. It is for that reason that birth cohort studies and other full population studies are so important. Even small rates of selective attrition can affect the overall means, and where samples are small and unrepresentative less reliance can be placed on results. For example, looking at the UK cognitive data you show above, the sample size for White British is 145,000 and the immigrant groups are at best 4000 or far less. Although one can have reasonable confidence about the error terms for the White British assessment, one would have to accept larger error terms for the smaller groups. Those latter groups could only be seen as representative of their genetic and cultural societies of origin if we can obtain proper measures on representative samples in their countries of origin. PISA and TIMMS have some data (though few African countries have ever participated in such studies) and the general pattern is for immigrant groups to reflect the same averages as in their countries of origin. As a general rule, single country immigrant histories are only indicators and not always representative, and currently the UK is a magnet for immigrant groups, particularly elites.
There is certainly great merit in looking at elite genetic groups in Africa, but so far it has been hard to get many reliable representative studies (mostly they tend to be nation based). If genetic based intelligence research can be done in Africa that would go a long way to elucidating gene-intelligence effects.
I went back to the article and I thought it worthwhile to make further comments, purely because it raises interesting questions about African intelligence. I have tried to find the original sources, though that has sometimes been difficult. There are other matters about which I have reservations, like the interpretation of regression to the mean, which I will skip for the time being. Most of what I am going to say has been covered in previous posts, so if you have already read those, please bear with me while I make some basic points.
I will look at the claims Chanda Chisala makes about African performance in the UK because it is a big part of his argument, and also because it is easier for me to comment on the educational system here.
Chisala observes: “The first report that caused some consternation in the IQ blogosphere indicated that black African pupils were apparently catching up with British white pupils on their GCSE tests and that in fact, they had already overtaken them at the lower end: the poor black kids were now performing better than poor white kids (The Guardian, 2010).”
The reference is slightly dated (true of some of the other references) and there are results for 2013 showing further African improvement, so it makes more sense to use the most up-to-date figures. The most recent in 2015 carries a caution that it is not comparable to previous years, so I might deal with it separately later. Instead I will use the recent but comparable account from the Department for Education First Statistical Release for January 2014.
Here is the relevant highlight in the executive summary:
“Pupils of any black background have seen a large improvement. The percentage of black pupils achieving 5 or more GCSEs at grade A* to C or equivalent including English and mathematics has increased by 3.5 percentage points since 2011/12. The gap compared to the national average has narrowed by 1.7 percentage points since 2011/12.”
First of all, what does it mean to achieve 5 or more GCSEs at grade A* to C? The GCSE is the standard UK scholastic examination at age 16. A* is the highest mark and the C is the lowest mark which still counts as a pass. Students of very low ability do not usually take the test, but the coverage is otherwise good. Here are the scores achieved in the 2008/9 to 2012/13 period.
“FSM eligibility” refers to poorer children who get Free School Meals (shown in blue) with the majority of the sample shown in red. On the left are the scholastic results which include general “softer” subjects. 85% of the school population get a pass mark, which by some process has gone up from 73% in 2008/9. That means that currently students with an IQ of 84 and above get the qualification. This is not of itself highly informative about scholastic ability. The C grade is very low and raw scores giving the actual percentage scores would be much better.
On the right are the slightly more demanding scores which include English and Maths (normally the core of any educational system). This shows that 65% of the school population achieve the required level, which has also risen from 54% in 2008/9. The present results mean that reading, writing and arithmetic can be achieved by students of IQ 94 and above. UK results on international tests like PISA have not gone up over this same period. It appears likely that the GCSE test has been made easier. The pass rate does not reveal the raw scores. I do not know why they suppress these essential data points. For example, giving pass rates denies us a look at the real score distributions. Do any of my readers have influence in these official statistical publication matters, and are they able to get the raw results?
Here are the racial comparisons:
This histogram nobly shows the pass rates for the 3 R’s from 2008 to 2012, thus giving the game away. There has been significant grade inflation. Pass rates are open to “gaming” as described in my previous post. We need to see the distributions for the raw scores. However, the racial comparisons are still instructive. Several groups have higher pass rates than the White British. This is certainly worth attending to. (A not-very-good measure can still contribute knowledge, particularly if no other data are available, though that is not the case here).
This is what the Government document says:
“Chinese pupils remain the highest attaining ethnic group. The percentage of Chinese pupils achieving 5 or more GCSEs at grade A* to C or equivalent including English and mathematics GCSEs or iGCSEs is 17.5 percentage points above the national average. The percentage of Chinese pupils achieving 5 or more GCSEs at grade A* to C or equivalent including English and mathematics GCSEs or iGCSEs increased by 1.6 percentage points in 2012/13, having decreased by 2.1 percentage points in the previous year. Despite this, the overall attainment gap for this indicator between Chinese pupils and the national average continues to remain at its lowest level, and has narrowed by 3.4 percentage points since 2008/09.
Pupils from a black background remain the lowest performing group, although they have shown the largest improvement. The percentage of black pupils achieving 5 or more GCSEs at grade A* to C or equivalent including English and mathematics GCSEs or iGCSEs is 2.5 percentage points below the national average. This gap has narrowed by 1.7 percentage points since 2011/12 but over the longer term has narrowed by 3.7 percentage points since 2008/09.”
“Gap narrowing” is obviously the focus of interest, but the metric used is misleading, because the size of the apparent gap is highly sensitive to the pass rate. Make the pass rate go up and the “gap” artificially seems to reduce as a consequence of the cut off used. Displaying the two distributions makes this clear. Try using Emil’s visualiser for tail effects to see how a pass rate can lead to different areas under each population distribution, and thus different gaps, whilst the mean differences remain the same.
The original and more detailed explanation of how to close the racial learning gap is given by Griffe du Lion here:
What the Chisala article does not give is sample sizes, which would allow us to work out confidence levels for the individual racial group results. However, the article then goes on to give sample sizes from the GL report on the Cognitive Assessment Test, which is for a very comparable sample. This assessment using cognitive tests is much better, and the sample sizes, means and standard deviations can be used for hypothesis testing.
Although representation for the White British is of course very good, the numbers for immigrants are understandably far smaller. Indian and Pakistani are over 3000, Black African at 2,197 students are more numerous than Black Caribbean at 1,978 which seems to indicate that the last decade of African migration has surpassed that of West Indians over 60 years. “The 2011 UK Census recorded 1,904,684 residents who identified as "Black/African/Caribbean/Black British", accounting for 3 per cent of the total UK population. This was the first UK census where the number of self-reported Black African residents exceeded that of Black Caribbeans”. It would be good to know which Africans are in the Black African category, and particularly what proportion is sub-Saharan.
I do not see how these figures could be considered representative of their countries of origin without further comparison with good local samples. However, the figures are informative. Sub-Sarahan African intelligence test results have been much debated by intelligence researchers, and the estimates range from about IQ 70 from Richard Lynn to IQ 80 from Jelte Wicherts. The key argument is about the representativeness of samples. The tests seem to be OK, much to popular surprise. Humans in all continents appear to solve basic problems in the same way. Africans have the same cognitive operating system as other continental groups. There are power differences, but not operating system incompatibilities. The most favourable comparison for immigrants in these data would be to concentrate on non-verbal intelligence, least affected by language and probably least affected by education. Here are three results drawn from the above Table 4.
White British Non-verbal IQ 101.4 (14.4)
Black Caribbean Non-verbal IQ 94.6 (12.9)
Black African Nov-verbal IQ 94.8 (12.1)
These scores are almost a projective test, in that they invite different interpretations. The Africans and Caribbeans have the same mean score and almost the same standard deviation, although the Africans have an even more restricted range than the Caribbeans. This restriction suggests some selection for the African group, but both Black groups have a more restricted range of ability than the larger White population. This phenomenon was noted for African Americans by Jensen, who said that the sd for African American intelligence was “approximately 12” which is borne out in this case. As you know, this has big implications for expected numbers at the highest levels of intellect, in that there should be even fewer than predicted on the basis of the lower mean. This is another testable finding.
If we take African IQ 75 as a temporary compromise between Lynn and Wicherts, the findings suggest that recent African migrants to the UK at IQ 94.1 are 1.27 standard deviations above the mean of their populations (assuming sd 15 in the local population, but if local population had a sd of 12.1 they would be even more high selected at 1.57 sd above their home mean) which might be plausible if the brighter ones were more likely to migrate. If one knew how they had been selected (say by looking at family backgrounds or prior scholastic attainments) it might be possible to do some calculations to test the hypothesis that brighter African students have migrated to the UK. However, the standard deviation for Africans is narrower than for White British (see later for some social class results).
The other approach would be to look at the racial composition of winners of national tests of higher achievements, say A* in Mathematics or Maths Olympiads or Physics prizes or whatever. By looking at how well various groups are represented calculations can be made about the mean ability of the populations from which they are drawn. That method depends on knowing the base population, and it would be unfair to count all Africans world-wide as the base population. UK census figures for immigrants would be the best alternative, but could be biased by selective migration of elites. (In general, international comparisons of representative samples of the scholastic abilities of immigrant groups do not show such elite effects, and the national average is a good estimate).
Griffe du Lion explains how to calculate the mean IQ of a sub-group without testing all of them with an intelligence test:
Even in the midst of this African intelligence debate, I cannot resist a digression. Have a look at the standard deviations for boys and girls on all three cognitive tests. Boys are more various than girls as per the usual findings on sexual dimorphism.
Now to the estimates of the abilities of particular groups of Africans. Much of this in the Chisala article comes from a 2006 government report, and the one table selected (actually Fig 4) gives the numbers of Africans by nation of origin, showing Somali, Nigerian and Ghanaian in descending order. Not shown in the Chisala paper is that the social profile of African immigrants is probably bimodal. They have almost as many parents in the professional ranks as the UK average, but also a very large number of unemployed persons. It is an odd distribution, suggestive of at least two different sources of immigrants as regards social status.
I cannot find a direct reference for the next figure in the article, about free school meals, but it is not crucial to Chisala’s argument. The “poverty” argument complicates cause and effect, and is particularly problematic for recent immigrants.
The next table Chisala shows is from “Back to basics: Towards a successful and cost-effective integration policy”, a publication from the “Institute for Public Policy Research, the UK's leading progressive thinktank”. (I do not know if it is a good thing to lead in these matters).
It takes some searching to find that Chisala has picked Table 6.1 on page 43. They use the familiar “5 A*-C GCSE grades” which have shown considerable grade inflation over that period. Here is the full description of how the table was put together “Table 6.1 presents new analysis for England showing educational performance at 16 (in GCSE examinations) by extended ethnicity code.” It would be good to have a little more detail, and at least to have sample sizes for each ethnicity for each year. The authors give the results as “mean percentage difference from England mean”. Without a methods statement I am not sure what this means. Since the table is about pass rates, the actual pass rates would suffice.
At face value these are interesting results. However, on a matter of this importance we need more than face value. Reference to a specific publication with methodological details is required.
The next table is about pass rates by language spoken at home, and comes from a Lambeth Borough publication.
The methods description is much better (page 8 with a good map of African languages). They are quite clear they picked the schools with the best scholastic results. “This research is an ethnographic study of outstanding schools in an inner London Local Authority. Two complementary methodological approaches were therefore adopted, each contributing a particular set of data to the study.” These approaches were GCSE results for all students in the Borough (which provides a good comparison) and case study research which is interesting but not standardised, so of marginal importance to the main debate.
So, we have one London Borough (with many African students) from which the best schools have been chosen, and the results are given as pass rates. Incredibly, although the publication is credited to the Borough’s Research and Statistics Unit, there is no mention of sample sizes anywhere that I can find. If they have better, more detailed results, they do not make it evident. We simply do not know how representative these children are of their populations of origin.
My summary of this publication is that there are some very bright African kids in Lambeth (and everywhere else in the world), but it is hard to drawn many conclusions from that statement.
Then a statement which drew a lot of interested comment: The superior Igbo achievement on GCSEs is not new and has been noted in studies that came before the recent media discovery of African performance. A 2007report on “case study” model schools in Lambeth also included a rare disclosure of specified Igbo performance (recorded as Ibo in the table below) and it confirms that Igbos have been performing exceptionally well for a long time (5 + A*-C GCSEs); in fact, it is difficult to find a time when they ever performed below British whites.
This bold claim is followed by a set of bright red and blue histograms showing that “African Ibo Africans” are the best students with a 2007 pass rate of 88%. Strong stuff. The link to this report is a Powerpoint presentation which indeed shows that very same figure. However, there is no hint of its provenance. I am sure it exists somewhere, but the link is not to a primary source, as would be the usual academic convention. An earlier slide in the same Powerpoint presentation shows that the Lambeth GCSE pass rate rose from 29% in 1998 to 62% in 2008. Either teachers doubled in effectiveness, or some brighter children arrived, or the test got much easier, perhaps by the C grade being given more generously.
So, that is the slide which launched a thousand website comments about the Ibos. The Ibos may very well be Africa’s finest, but an article should allow interested readers to get to the original source without wasting time guessing where it might be.
I deduce it may be Demie, Feyisa; McLean, Christabel EDUCATIONAL STUDIES 33(4):415-434 · DECEMBER 2007.
Essentially, they chose high performing schools, saying: Selection of high-performing schools was a prerequisite because the research concerns itself with educational success. So, it is a deliberately non-representative sample.
Here is Table 1 from that paper, showing that the pass rate for the case-study schools is higher than that for the rest of the Borough. More important, the African pass rate for all schools is the same as the national average.
Here is Figure 1 giving the results for their chosen case-study schools.
Remember, this measures whether students can get a C grade. Absolutely no sample sizes are given in this paper, nor in the research note from which it is drawn. It is hard to draw any conclusions from this report as it stands. I cannot find the figure Chisala gives in his article, but by now I am getting weary.
It is hard to track down the selected article slides to an actual publication, from which one can get the methods, selection criteria, sample sizes and descriptive statistics. There is material there, no doubt, but it is far from the clear signposting required in research. Academic papers may be boring, and the peer review process tedious, but they have some merits when you wish to check the facts.
Chisala mentions Prof Simon Burgess’s work. Here is my comment on that paper:
You will see that “progress” measures tend to favour those who start with low scores.
Prof Burgess has been offered the right of reply, as always, and has acknowledged the offer.
This is a curate’s egg of an article. The Cognitive Assessment Test data are interesting, informative, and worthy of further study. The GCSE pass rate results are poor by comparison: crude and less informative. The pass rates by African language, and by implication by African genetic sub-groups are hard to track down, but at the same time, they are tantalising.
If evolution applies to all of us, that includes Africans. It is perfectly possible that some African sub-groups are brighter than others. Any geographic group which takes care to encourage bright persons to marry one another could see positive effects after 20-30 generations of consistently strong selection, if those brighter parents have more surviving offspring, which is likely. The hereditarian position (ie genetic probably accounting for 50% of the variance) would be strengthened by finding reliable evidence of genetically brighter groups in all populations including Africa. The Al-Rashidis used to tell me “We are the Jews of the Arab world”. If such groups exist on any continent they will work hard, have commanding positions in commerce and learning, and will be very picky about marriage choices.
One cannot argue that a whole population is bright on the basis of a few prize winners, however emotionally appealing the impact of the individual case. However, if a particular genetic or cultural group is over-represented among the winners of a tough intellectual competition (out of proportion to their population numbers), then that can be good evidence that they are drawn from a group which has high ability. Potentially, there are undiscovered tribes and even extended families out there with outstanding intellectual abilities, and they are worth searching for. There should be no impediment to Africans championing African ancestry as the cause of African intellectual superiority, and supplying the supportive evidence.
Perhaps I have spent too long on this particular article, but I do not think that the tables and figures presented by Chisala are sufficient proof that particular African groups are among the brightest persons on the earth. They may be, but it would be prudent to require the usual sorts of evidence, evidence which is hard to obtain from the article. As the lawyers say “We need further and better particulars” and only then can the case proceed.