"Hereditarian, (Frost, 2021, 2019) has proposed some explanations. His first explanation is that blacks are likely to cheat more."
I fell off my chair when I read that sentence. Did I actually write that? Well, no. I wrote that Nigeria is a low-trust society and that Nigerian students feel it is legitimate to cheat on university exams. It doesn't follow that they are more willing than white Europeans to cheat friends and family members. (I suspect the reverse is true). It's an even greater semantic leap to say that "blacks" in general are likely to cheat more.
"His second argument is that because blacks faced immigration restrictions in coming to the UK, we have a selected sample of especially intelligent blacks."
I considered that argument and rejected it. I did make a second argument, i.e., higher cognitive ability does exist among some African groups, like the Igbo. Why didn't you mention it?
"Nigerian academic achievement may be genuine in some cases. This is particularly so with respect to the Igbo, who have a longstanding record of achievement within and outside school."
Finally, I have never identified myself as a "hereditarian." The term is meaningless and is most often used by people who want to shut down the nature-nurture debate (like "ultra-Darwinian" and "genetic determinist"). Why are you using it? Please don't use me as a sock puppet for views I don't actually hold.
Apologies if I got the wrong impression of your arguments, although I think 'sock puppet' Peter's positions are perfectly reasonable. Of course black people are probably more likely to cheat in exams. That seemed to be what you were saying in your blog posts and I agreed. My impression was that you thought selective immigration had some effect even if you didn't go to the extreme of Cochran's position, this also seemed to be very much implied by your discussion of the high IQ African groups that moved at a disproportionate rate.
I used the label hereditarian for you and I do so for myself because I think it is more respectable than saying race scientist.
Your impressions are wrong, in some cases terribly wrong. I wasn't saying that black people are more likely to cheat on exams. I was talking in general about people who live in low-trust societies. In such societies, it's legitimate to cheat as long as you aren't cheating family or friends.
I don't believe in the argument of selective immigration. Let me quote my post (which you cite): "To make that argument work, however, Nigerian immigrants to the UK would have to be much smarter than the average Nigerian, with an IQ more than one standard deviation higher and probably two." That isn't the profile of Nigerian immigrants in the UK. If Igbo immigrants do well in the UK, it's because they do well in Nigeria. They do well everywhere.
Sorry, but I am not a hereditarian:
"Hereditarian: a person who believes that differences between individuals or groups, including moral and intellectual attributes, are predominantly determined by genetic factors".
That is not at all what I believe. If you wish to be "respectable," please respect my beliefs.
Why not simply write: "One explanation is that blacks are likely to cheat more." Why are you afraid to own your ideas? Why are you forcing me to become your sock puppet?
This is not a matter of honest disagreement. I didn't say that "blacks are likely to cheat more." I said that Nigeria is a low-trust society and that students in a low-trust society are more likely to cheat on a university exam. That isn't a racial thing. Most societies on our planet are low-trust, including many white ones.
Your only justification is that you had the impression that I was thinking about black people in general. Well, I wasn't. That impression is yours, and yours alone.
You now seem to be arguing that your impressions of reality are more important than reality itself. Your power of imagination gives you the right to reimagine reality as you see fit, even to the point of misrepresenting reality. I used to think that only the "woke" were infected by that kind of thinking. Apparently not.
"58% of the variance in maths GCSE results was explained by intelligence in 2002, yet by 2010-2013 this halved to 31.4%. By comparison, 73.4% of the variance in SATs in America could be explained by g before the 1994 reforms (Frey and Detterman, 2004). GCSEs are just not a good measure of intelligence anymore."
I believe you should use the g-loading rather than the variance explained by g, as the expected gap would be closer to proportional to g-loading than to variance explained by g? I.e. a g gap of 1 standard deviation would not lead to a test gap of 0.58 and 0.31 standard deviations, but instead closer to 0.76 and 0.56 standard deviations? (Not exactly these numbers because Cohen's d is kind of weird, but approximately.)
On that note, if "The shared environment accounts for 14–21%", a shared environment gap of 1 SD would equate to a test gap of 0.37-0.46 standard deviations, correct?
So to cancel out that expected 0.56 SD gap in GCSE math scores, one would need a 1.22 to 1.52 SD advantage in shared environment, right? Or in other words, the average of the group with the lower g would need to be at about the 90th percentile in terms of shared environment(strict education obsessed parents/community presumably).
I'm guessing Nigerian parents pull up the average in that regard a decent amount. Looks like black Africans outperform whites on Maths&English GCSEs, while black Caribbeans are 0.4 SD behind whites*, pretty close to the 0.56 SD expected math gap and 0.52 SD expected English gap based on an assumption of a 1 SD g-gap.
Making the test more "shared environment"-loaded doesn't just disadvantage high-g groups, it also advantages "high studiousness" groups.
Very similar to your numbers for variance explained by g: 31% for math and 27% for English. In other words, shared environment may be as important as g for the new GCSEs.
Something that makes me question the low g variance explained for GCSEs though is that they correlate highly with CAT4 scores, 0.78 for GCSE maths in 2019 for example:
This only holds true if you assume that under equal environments Maths GCSEs and g will show the same racial gaps. There is no reason to believe this, given the test is less g-loaded than, well, g you would expect the Maths gap to be smaller. Consistent with Spearman's hypothesis.
You might also be confused about g-loading then? g loading is how well something correlated with the general factor of intelligence not how heritable a trait is. So something can be very heritable but not correlated with g.
It's unclear exactly how much of the gap was reduced by the lower g-loading. If I were to do a paper I'd probably provide some calibraed estimates based on whatever estimated g-loadings we could get.
It isn't the whole part of the story - whites perform uniquely poorly in the GCSEs compared to their IQs. So there's something going wrong with teaching or culture for whites as well. But the issue is somewhat speculative and the blog was already too long.
I would argue that in this specific instance variance explained is better. We care about how well g fits to Maths GCSE (RSS/TSS), not the average effect size. Mostly I just stuck to how Deary reported his results though (share of variance explained).
I don't understand what you mean by "We care about how well g fits to Maths GCSE (RSS/TSS)". Surely what we care about when considering the racial gap in g is how strongly Maths GCSE is influenced by g?
Rushton's application of R-K selection theory to races would suggest that blacks would mature earlier than whites, so perhaps they would be competitive with whites at 16 years old, but not at 18. This wouldn't explain GSCE result changes over time, but might still explain much of the data.
You're saying that GCSEs have become easier and less g-loaded (and thereby more trainable). And therefore the white-black difference has narrowed over time. So either:
1) White scores have stayed constant while black scores have gone up, OR
2) Scores for everyone have gone up, but scores for blacks have gone up higher than for whites.
Which of those statements is true? I'm guessing that (2) is true since the test is now easier. But how do you explain why White scores are not going up higher than blacks?
In the US, the SAT exams for universities have also gotten easier and less g-loaded. The exam is now quite trainable and East Asians are training the hell out of it. Any comment this chart here?
Why aren't whites going up as much as blacks? Reduced g-loading does this directly. The more tests reflect noise, rather than g the more racial gaps should tend to 0. Then something we only eluded to briefly is the increased environmental component which may also be playing a role. If part of what is causing the g loading to fall is an increased environmental component, then that will also change race gaps depending on which groups work hardest/have the best environment.
So let's look at the American context. A reduced g-loading will tend to reduce the white-asian gap and the white-black gap. But at the same time, the environment or training is becoming more important. This will tend to increase the white-asian gap (because asians work hard), but not have so much of an effect on the white-black gap (assuming both groups work equally as hard in the USA).
Because black culture is generally better in the UK, an increased role for the environment will give blacks an advantage over whites, whilst this is not the case in America.
"Hereditarian, (Frost, 2021, 2019) has proposed some explanations. His first explanation is that blacks are likely to cheat more."
I fell off my chair when I read that sentence. Did I actually write that? Well, no. I wrote that Nigeria is a low-trust society and that Nigerian students feel it is legitimate to cheat on university exams. It doesn't follow that they are more willing than white Europeans to cheat friends and family members. (I suspect the reverse is true). It's an even greater semantic leap to say that "blacks" in general are likely to cheat more.
"His second argument is that because blacks faced immigration restrictions in coming to the UK, we have a selected sample of especially intelligent blacks."
I considered that argument and rejected it. I did make a second argument, i.e., higher cognitive ability does exist among some African groups, like the Igbo. Why didn't you mention it?
"Nigerian academic achievement may be genuine in some cases. This is particularly so with respect to the Igbo, who have a longstanding record of achievement within and outside school."
Finally, I have never identified myself as a "hereditarian." The term is meaningless and is most often used by people who want to shut down the nature-nurture debate (like "ultra-Darwinian" and "genetic determinist"). Why are you using it? Please don't use me as a sock puppet for views I don't actually hold.
Apologies if I got the wrong impression of your arguments, although I think 'sock puppet' Peter's positions are perfectly reasonable. Of course black people are probably more likely to cheat in exams. That seemed to be what you were saying in your blog posts and I agreed. My impression was that you thought selective immigration had some effect even if you didn't go to the extreme of Cochran's position, this also seemed to be very much implied by your discussion of the high IQ African groups that moved at a disproportionate rate.
I used the label hereditarian for you and I do so for myself because I think it is more respectable than saying race scientist.
Your impressions are wrong, in some cases terribly wrong. I wasn't saying that black people are more likely to cheat on exams. I was talking in general about people who live in low-trust societies. In such societies, it's legitimate to cheat as long as you aren't cheating family or friends.
I don't believe in the argument of selective immigration. Let me quote my post (which you cite): "To make that argument work, however, Nigerian immigrants to the UK would have to be much smarter than the average Nigerian, with an IQ more than one standard deviation higher and probably two." That isn't the profile of Nigerian immigrants in the UK. If Igbo immigrants do well in the UK, it's because they do well in Nigeria. They do well everywhere.
Sorry, but I am not a hereditarian:
"Hereditarian: a person who believes that differences between individuals or groups, including moral and intellectual attributes, are predominantly determined by genetic factors".
That is not at all what I believe. If you wish to be "respectable," please respect my beliefs.
I've removed the word hereditarian and added a pointer to the comments so readers are aware that you disagree with our characterisation.
Why not simply write: "One explanation is that blacks are likely to cheat more." Why are you afraid to own your ideas? Why are you forcing me to become your sock puppet?
This is not a matter of honest disagreement. I didn't say that "blacks are likely to cheat more." I said that Nigeria is a low-trust society and that students in a low-trust society are more likely to cheat on a university exam. That isn't a racial thing. Most societies on our planet are low-trust, including many white ones.
Your only justification is that you had the impression that I was thinking about black people in general. Well, I wasn't. That impression is yours, and yours alone.
You now seem to be arguing that your impressions of reality are more important than reality itself. Your power of imagination gives you the right to reimagine reality as you see fit, even to the point of misrepresenting reality. I used to think that only the "woke" were infected by that kind of thinking. Apparently not.
"58% of the variance in maths GCSE results was explained by intelligence in 2002, yet by 2010-2013 this halved to 31.4%. By comparison, 73.4% of the variance in SATs in America could be explained by g before the 1994 reforms (Frey and Detterman, 2004). GCSEs are just not a good measure of intelligence anymore."
I believe you should use the g-loading rather than the variance explained by g, as the expected gap would be closer to proportional to g-loading than to variance explained by g? I.e. a g gap of 1 standard deviation would not lead to a test gap of 0.58 and 0.31 standard deviations, but instead closer to 0.76 and 0.56 standard deviations? (Not exactly these numbers because Cohen's d is kind of weird, but approximately.)
On that note, if "The shared environment accounts for 14–21%", a shared environment gap of 1 SD would equate to a test gap of 0.37-0.46 standard deviations, correct?
So to cancel out that expected 0.56 SD gap in GCSE math scores, one would need a 1.22 to 1.52 SD advantage in shared environment, right? Or in other words, the average of the group with the lower g would need to be at about the 90th percentile in terms of shared environment(strict education obsessed parents/community presumably).
I'm guessing Nigerian parents pull up the average in that regard a decent amount. Looks like black Africans outperform whites on Maths&English GCSEs, while black Caribbeans are 0.4 SD behind whites*, pretty close to the 0.56 SD expected math gap and 0.52 SD expected English gap based on an assumption of a 1 SD g-gap.
Making the test more "shared environment"-loaded doesn't just disadvantage high-g groups, it also advantages "high studiousness" groups.
* https://www.ethnicity-facts-figures.service.gov.uk/education-skills-and-training/11-to-16-years-old/a-to-c-in-english-and-maths-gcse-attainment-for-children-aged-14-to-16-key-stage-4/latest#by-ethnicity
This old table may illustrate the change in the GCSEs: https://preview.redd.it/m7pjubthjasy.png?auto=webp&s=c9993e64351aefd51bbd7be89e4674445d740694
The stereotypically education obsessed groups increased their relative scores greatly after 2003.
Another study of GCSEs taken around 2011 finds even higher shared environment effects:
"In addition, estimates of shared environment are also similar across subjects: English (31%), mathematics (26%)..."
- https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3859476/
Very similar to your numbers for variance explained by g: 31% for math and 27% for English. In other words, shared environment may be as important as g for the new GCSEs.
Something that makes me question the low g variance explained for GCSEs though is that they correlate highly with CAT4 scores, 0.78 for GCSE maths in 2019 for example:
https://support.gl-education.com/media/2785/cat4-international-technical-report.pdf (page 10)
Wow, I'll need to look into this. Thanks.
Thanks for sharing. Looks like we have got this wrong. I've added a correction on the post and will share on Twitter.
"Since 2010, the g-loading of GCSE has increased substantially."
Sorry to be a nag, but I'm not sure that's true either. GCSE scores seem to have been highly correlated with the CAT3 back then:
https://lesacreduprintemps19.files.wordpress.com/2012/02/glassessment.pdf (page 7)
And fwiw back in 2002:
https://www.researchgate.net/figure/Correlations-of-Year-7-CAT-and-KS2-points-scores-with-GCSE-outcomes-five-years-later_tbl1_232987451
This only holds true if you assume that under equal environments Maths GCSEs and g will show the same racial gaps. There is no reason to believe this, given the test is less g-loaded than, well, g you would expect the Maths gap to be smaller. Consistent with Spearman's hypothesis.
Which part did of my comment did I not account for the g-loading of Math GCSEs?
You might also be confused about g-loading then? g loading is how well something correlated with the general factor of intelligence not how heritable a trait is. So something can be very heritable but not correlated with g.
Guy's analysis looks correct to me!
I didn't mention anything about heritability? I mentioned strict parents as part of one's shared environment.
Yeah, perhaps the correlation is better than variance explained is better in science communication. Awkward this post came out only yesterday...
https://kirkegaard.substack.com/p/variance-explained-is-mostly-bad?utm_source=profile&utm_medium=reader2
It's unclear exactly how much of the gap was reduced by the lower g-loading. If I were to do a paper I'd probably provide some calibraed estimates based on whatever estimated g-loadings we could get.
It isn't the whole part of the story - whites perform uniquely poorly in the GCSEs compared to their IQs. So there's something going wrong with teaching or culture for whites as well. But the issue is somewhat speculative and the blog was already too long.
I would argue that in this specific instance variance explained is better. We care about how well g fits to Maths GCSE (RSS/TSS), not the average effect size. Mostly I just stuck to how Deary reported his results though (share of variance explained).
I don't understand what you mean by "We care about how well g fits to Maths GCSE (RSS/TSS)". Surely what we care about when considering the racial gap in g is how strongly Maths GCSE is influenced by g?
Immigration selects for IQ but also conscientiousness. Blacks care more about GCSE's than the white working class
Rushton's application of R-K selection theory to races would suggest that blacks would mature earlier than whites, so perhaps they would be competitive with whites at 16 years old, but not at 18. This wouldn't explain GSCE result changes over time, but might still explain much of the data.
You're saying that GCSEs have become easier and less g-loaded (and thereby more trainable). And therefore the white-black difference has narrowed over time. So either:
1) White scores have stayed constant while black scores have gone up, OR
2) Scores for everyone have gone up, but scores for blacks have gone up higher than for whites.
Which of those statements is true? I'm guessing that (2) is true since the test is now easier. But how do you explain why White scores are not going up higher than blacks?
In the US, the SAT exams for universities have also gotten easier and less g-loaded. The exam is now quite trainable and East Asians are training the hell out of it. Any comment this chart here?
https://twitter.com/UnsilencedSci/status/1578000255895105537
2 Is what is happening.
Why aren't whites going up as much as blacks? Reduced g-loading does this directly. The more tests reflect noise, rather than g the more racial gaps should tend to 0. Then something we only eluded to briefly is the increased environmental component which may also be playing a role. If part of what is causing the g loading to fall is an increased environmental component, then that will also change race gaps depending on which groups work hardest/have the best environment.
So let's look at the American context. A reduced g-loading will tend to reduce the white-asian gap and the white-black gap. But at the same time, the environment or training is becoming more important. This will tend to increase the white-asian gap (because asians work hard), but not have so much of an effect on the white-black gap (assuming both groups work equally as hard in the USA).
Because black culture is generally better in the UK, an increased role for the environment will give blacks an advantage over whites, whilst this is not the case in America.