« Paleoanthro Primer | Gene Expression Front Page | The precautionary principle's utility.... »
March 13, 2005

Education and ethnic groups in England

In a previous post I discussed educational performance in England by different ethnic groups in 2002. Razib recently drew attention to some new statistics. These go up to 2004, and are available in full here (490Kb pdf file). A document covering 2002 and 2003 is available here (254Kb).

Since 2002 there have been two interesting changes. In 2003 the statistics began to include results for mixed-race children, and for 2004 they also include results for qualifications at age 18, whereas previously they only went up to 16.

There is a lot of complicated information in these reports, so I have tried to extract some key points most useful for ethnic comparisons…

For the benefit of readers outside the UK, education here is compulsory for children from ages 5 to 16. There are some differences in the education system and qualifications between England, Wales, Scotland and Northern Ireland. The data discussed here relate only to England. In England the progress of children is assessed at the end of three Key Stages, at roughly ages 7, 11 and 14. In the last year of compulsory education (age 15/16) nearly all children take examinations for the General Certificate of Secondary Education (GCSE). After age 16 a majority of children continue in education for at least another two years. At around age 18 most of these take examinations for the General Certificate of Education (Advanced) (the A-level), usually in 2 to 4 subjects. A-levels or equivalent qualifications are normally required for entry to universities.

The data for 2004 cover all of these levels (except Key Stage 3, results for which have been delayed). I have selected data for Key Stage 1, as this is the earliest measure of achievement; for GSCE, which is the last stage for which the data covers a complete age cohort; and for A-levels.

Two limitations of coverage should be noted. First, the data cover only children in state-funded schools. About 6% of children in England attend private (independent) schools and are therefore not included. These are mainly children of parents with high income. The data are therefore probably somewhat depleted in the higher levels of children’s ability. It is possible that this affects different ethnic groups to different extents. It might be supposed that private education would be disproportionately ’white’, but this cannot safely be assumed. Black and Asian middle-class parents are keen on private education, especially as it is often the only way of getting good discipline and/or single-sex schooling.

The second limitation of coverage is that the A-level data only relate to those children who stay in school to take A-levels. They exclude those who do not take A-levels at all, and those who leave school but take A-levels in post-16 colleges. This may introduce some distortion with respect to ethnic groups.

As to the content of the various qualifications, I cannot claim to know much about them. The Key Stage tests assess the achievement of ‘expected’ levels of progress at each stage. It is intended that with good teaching the great majority of normal children should be able to pass the Key Stage tests, and the proportions reaching the target levels are in fact over 80%. In other words, the ’hurdles’ are set relatively low. One consequence of this is that the range of pass rates for different ethnic groups is rather restricted (from around 75% to 90%). GCSE is somewhat more selective. GCSEs are graded from F to A (with starred A* recently introduced as a mark of excellence). The usual measure of ‘satisfactory’ performance is to get passes in at least 5 subjects at Grade C or higher. For GCE A-levels there is a complicated ‘points’ system depending on the number of passes at each grade. See the pdf file for more details.

Like all academic qualifications, one would expect performance on Key Stage tests, GCSEs, and A-levels to be positively correlated with IQ. However, IQ is certainly not the only factor involved in performance, as is shown by the fact that in the educational tests girls perform substantially better than boys, although the average IQ of boys and girls is much the same (see N. Mackintosh IQ and Human Intelligence, pp.182-98 for discussion). Another indicator is the fact that ‘traveller’ children (gypsies, etc), whose education is frequently disrupted, perform far below other ethnic groups. So the educational data give at best a very imperfect guide to IQ levels.

I will extract data for the following ethnic groups:

W = White (n1)
W/BC = Mixed White and Black Caribbean
W/BA = Mixed White and Black African
W/A = Mixed White and Asian
A = Asian (n2)
I = Indian
P = Pakistani
Ba = Bangladeshi
Bl = Black (n3)
BC = Black Caribbean
BA = Black African
C = Chinese (n4)
All = average for all children

n1: in the full data ’White’ is subdivided into White British and several other smaller groups.
n2: ’Asian’ covers Indian, Pakistani, Bangladeshi and ’other Asian’. ’Other Asian’ covers miscellaneous groups (Vietnamese, etc), but not Chinese. Middle Easterners (Arabs, Turks, etc) may be included in ’other Asian’, but in the 2001 Census most of them seem to classify themselves as ’White’. People of Asian origin via the Caribbean or east Africa usually give ‘Indian’ as their self-identified ethnicity.
n3: ’Black’ includes Black African, Black Caribbean, and ’Black Other’. ’Black Other’ may include some mixed-race children, but those who are known to have a White parent should be counted under W/BC or W/BA.
n4: ethnic Chinese people in Britain are mainly of Hong Kong origin. They are traditionally counted separately from other Asians in British statistics.

Now, at last, some statistics. For Key Stage 1 I have averaged the pass rates for the three tests (Reading, Writing and Mathematics) covered in the full data. For GCSE, the figures give the percentages of children at age 15/16 getting passes in at least 5 subjects at Grade C or higher. For GCE A-level, the figures give average aggregate points per student who attempts A-levels. As performance is so different for boys and girls, I have given data for boys, girls, and the average for both sexes. Percentages are given to one decimal place.



*I have averaged the pass rates for the three subjects using percentages for each subject rounded to the nearest point. The average therefore necessarily terminates in .0, .3 (i.e. .333... rounded down), or .7 (i.e. .666... rounded up). Anyone who wants to calculate the averages more precisely from the raw data is welcome to do so!





It may be of interest to put these results in rank order of ethnic groups. For simplicity I will only use the combined boy-girl data.


1. C________91.0
2. I_________89.0
3. W/A______88.7
4. W_______86.3
5. W/BA____84.0
6. W/BC____83.3
7. A________81.7
8. BC______80.3
9. Bl_______79.0
10. BA______78.3
11(=) P_____77.7
11(=) Ba____77.7


1. C________74.2
2. I_________66.6
3. W/A______65.7
4. A________55.3
5. W_______52.3
6. Ba_______48.4
7. W/BA_____47.1
8. P________45.2
9. BA_______43.3
10. W/BC____39.7
11. Bl_______39.0
12. BC______35.7


1. C________315
2. W/A______301
3. W________277
4. W/BA_____268
5. I_________257
6. W/BC_____247
7. A_________244
8. BA________223
9. P_________218
10. Bl________217
11. Ba_______216
12. BC_______203

Ignoring the mixed-race data for the moment, the other ethnic groups show few surprises as compared with the earlier data. The rank order is not very different at the different stages. The Chinese are consistently at the top, Indians are high, and Black Caribbeans are low. But note that Pakistanis and Bangladeshis start at the very bottom in Key Stage 1, and improve somewhat in later stages. Black Africans also improve from a low starting position. I suspect that in both cases this is due to an initial English language deficit, which they later overcome. The other slight anomaly is that Indians slip back somewhat from their high position when they get to A-level. Bearing in mind that the A-level rankings are based only on those children who take A-levels, it is possible that Indian children are under greater social and parental pressure than other ethnic groups to take A-levels, as a prerequisite for university entry, even when they are less academically able. This would depress the average score of Indian candidates. Of course, this suggestion would need to be explored more fully.

Turning to the mixed-race results, the performance of White-Black Caribbean and White-Black African children is intermediate between that of children from the relevant unmixed groups. This is what we would expect if the factors underlying academic performance have a large genetic component. No doubt one could also find a purely environmental explanation, though it is not obvious (to me, anyway) that children of White and Black parents would have a quality of environment somewhere in between that of ‘pure’ White and Black children. But in any case the neat pattern is spoiled by the mixed White-Asian group, who have performance better than that of either of the parent groups (unless we suppose that the Asian parent is nearly always Indian, which is not the case). I suppose it might be suggested that this is due to ‘hybrid vigour’, but I think this hypothesis is ’more ingenious than sound’. It is necessary to remember that mixed-race partnerships are not a random sample of the parent populations, so there could be bias either upward or downward in the genetic and environmental endowment of mixed-race children, relative to the average of the parent ethnic groups. I didn’t expect to find any useful facts on the subject, but I was pleased to find some interesting recent research by Reya Muttarak of the University of Oxford. Indeed it is so interesting that I will devote a future post to it! For the moment, the main point to note is that in Britain mixed-race partnerships are on average associated with higher-than-average educational attainment and social class (relative to the relevant group averages) of both partners. The correlation is particularly strong for Asians, fairly strong for Black women, but less clear-cut for Black men. The high performance of White-Asian children therefore correlates with the relatively high average educational and social level of both their parents. Obviously this correlation could be given either a genetic or an environmental explanation.

I had intended also to discuss the links, if any, between poverty, ethnic group, and educational performance, but I will save that for another post.

Posted by David B at 07:35 AM