So are you saying regression to the mean can't happen over many generations since "regression to the mean doesn’t happen twice"? Gregory Clark's work shows that regression to the mean can occur incrementally over many generations (even with assortative mating). How does one reconcile this apparent inconsistency?
Yes, Clark is referring to regression in social status not in intelligence per se (but surely intelligence is a major contributor to social status).
Maybe I would understand your original scenario better if we also looked backwards to the grandparents' parents and their forebears. On average, what would you expect their IQs to have been each preceding generation? On average, would they have all been 120 as well? It seems their pedigree would be an important determinative factor. (I guess that's what Ives Parr is bringing up below.)
In other words, what exactly is the mean in question? There is a population mean of 100 but there also might be subpopulation means that are different. How do these different means interact and affect regression(s)? I assume assortative mating plays a large role in this issue.
I know some time ago, Clark noted that there isn't much evidence that American Blacks and Jews have shown measurable regression to the mean over recent decades. But I do recall him noting that other groups could be expected to regress to the mean but it would happen slowly, likely taking centuries. (Maybe this multi-generational regression only occurs when assortative mating isn't perfect enough to sustain ongoing high status/intelligence levels in offspring?)
- Regarding Blacks and Jews, I found this paper with a google search:
"Regression to Mediocrity? Surnames and Social Mobility in England, 1200-2009"
"...Complete social mobility and the absence of persistent social classes is not true of the modern USA where at least two groups – Blacks and Jews – have, as table 1 shows, not shown regression to the mean."
- Regarding other groups taking centuries to regress, I don't recall if it was in a video interview, an article, or 'Farewell'. But I'm pretty sure he did say it because it stuck out in mind mind when I heard it. He may discuss this issue in the article above but I haven't had time to read much of it yet.
"[...] the expected value of b in estimating the connection between grandfather and grandson, will now be just the same as for the connection between father and son. Similarly for a father and any more distant descendant, the regression coefficient will be no greater. After one generation there will be no further regression to the mean."
That's one possible theoretical model in which there is inherited social class. That paper (which claims that England had an effectively classless society of social mobility) also seems to have a different perspective from his more recent work. More recently he's written about how wealth being directly inherited shows different patterns from genetic traits, and claimed that basically every society (including egalitarian Sweden and communist China) has little social mobility.
You can keep extending the example back up the family tree. If all 4 grandparents have 120 IQs, does it make a difference if their eight parents averaged 100 or 120 or 140? I can't see why it wouldn't.
So if I understand correctly it would be more accurate to use the aggregate mean of the two families for regression? Say the Smith’s avg IQ was 117 and the Doe’s 100 — the combined familial mean would be 108.5. So say John Smith (130) and Jane Doe (102) had a son. Using general pop mean of 100: ((130+102)/2)-100) x .6 + 100 = 109.6 but using the combined familial mean ((130+102)/2-108.5) x .6 +108.5 = 113
The Johnson’s avg IQ 125 the White’s avg IQ 105 — combined familial mean 115. If Jack Johnson (130) and Sue White (98) had a daughter. Using general pop mean 100: ((130+98)/2-100) x .6 + 100 = 108.4 but using their combined familial mean ((130+98)/2-115) x .6 +115 = 114.4.
So if Mark Smith (113) and Sally Johnson (114.4)(combined familial mean: Smith’s (117) + Johnson’s (125)=121 have a child we’d expect ((113+114.4/2-121) x .6 +121=116.62 using familial mean whereas we’d expect ((113+114.4)/2-100) x .6 +100=108.22?
I'm not sure if that's how the math would work, but the familial mean would have incremental predictive validity beyond the mean of the general population and the parents.
I see. This would have interesting implications — namely, that family pedigree is quite important, if not more important than a couple’s individual IQs.
You cite the formula for 40% regression as if it were common knowledge. I have seen it detailed only by Stephen Hsu in a 2011 paper, and indirectly by James Flynn in "Does your family make you smarter."
Statistically, it has to be true for the median intelligence and standard deviation (nominally 100 and 15) to remain constant from generation to generation. I posted a video to that effect a few years back.
Searching my archives, I found the 2011 paper entitled "Investigating the Genomic Basis for Intelligence" by Steve Hsu of Caltech. The full text of a short section entitled: "Your Kids and Regression" is as follows:
Assuming parental midpoint of n SD above the population average, the kids’ IQ will be normally distributed about a mean which is around +.6n with residual SD of about 13 points. (The.6 could actually be anywhere in the range (.5, .7), but the SD doesn’t vary much from choice of empirical inputs.)
So, e.g., for n = 4 (parental midpoint of 160 – very smart parents!), the mean for the kids would be 136 with only a few percent chance of any kid to surpass 160 (requires about a 2 SD fluctuation) For n = 3 (parental midpoint of 145) the mean for the kids would be 127 and the probability of exceeding 145 less than 10 percent.
>But given the high additive heritability of IQ and assortative mating for IQ, isn't it more likely that the grandparents actually come from a high genotypic IQ family line?
Well, yes. The question is whether they are coming from a line that is relatively more genetically endowed than the average person with an IQ of 120 (and I assume that is a no).
Right. If you could also find out the IQs of siblings, uncles/aunts etc. you could probably better guesstimate if the four grandparents each with 120 IQs came from castes where 120 IQ is standard, or if they happened to be higher IQ than most of their relatives.
Reading up on some English intellectual dynasties like the Huxleys (Julian, Aldous, and the least well-known, Andrew who won the Nobel in chemistry), I could imagine people for whom there wouldn't be much regression toward the mean past their grandparents' 120s. The Huxley brothers were descended from TH Huxley on one side and Thomas Arnold, the most famous educator of Victorian times, on the other (well, it's more complicated than that because Andrew was a half-brother of Julian and Aldous).
The 19th Century British intelligentsia clearly did more assortative mating than Richard Herrnstein would have anticipated.
But, most of the time, if your grandparents averaged 120 IQ, they probably enjoyed some genetic good luck relative to their grandparents' average IQ.
Thanks! Everytime someone said RTM I always wondered what "mean" should mean. Family mean, ethnic mean, social class mean, and so on. How does a mean form and how do we know?
It's weird how many people invoke the phrase "regression to the mean" without considering which mean they're referring to. HBD'ers mock the left for believing evolution stops at the neck, then turn around and pretend evolution stops at the racial level.
So are you saying regression to the mean can't happen over many generations since "regression to the mean doesn’t happen twice"? Gregory Clark's work shows that regression to the mean can occur incrementally over many generations (even with assortative mating). How does one reconcile this apparent inconsistency?
Yes, Clark is referring to regression in social status not in intelligence per se (but surely intelligence is a major contributor to social status).
If people continue breeding into the surname, then RTM will happen more times because it's happening for the first time in several generations.
Maybe I would understand your original scenario better if we also looked backwards to the grandparents' parents and their forebears. On average, what would you expect their IQs to have been each preceding generation? On average, would they have all been 120 as well? It seems their pedigree would be an important determinative factor. (I guess that's what Ives Parr is bringing up below.)
In other words, what exactly is the mean in question? There is a population mean of 100 but there also might be subpopulation means that are different. How do these different means interact and affect regression(s)? I assume assortative mating plays a large role in this issue.
Funny, since that misunderstanding of regression to the mean was used to argue against Clark's Farewell to Alms. https://entitledtoanopinion.wordpress.com/2009/01/18/greg-clark-responds-to-critics/
I know some time ago, Clark noted that there isn't much evidence that American Blacks and Jews have shown measurable regression to the mean over recent decades. But I do recall him noting that other groups could be expected to regress to the mean but it would happen slowly, likely taking centuries. (Maybe this multi-generational regression only occurs when assortative mating isn't perfect enough to sustain ongoing high status/intelligence levels in offspring?)
Where did he say that?
- Regarding Blacks and Jews, I found this paper with a google search:
"Regression to Mediocrity? Surnames and Social Mobility in England, 1200-2009"
"...Complete social mobility and the absence of persistent social classes is not true of the modern USA where at least two groups – Blacks and Jews – have, as table 1 shows, not shown regression to the mean."
https://www.econ.ucdavis.edu/faculty/gclark/papers/Ruling%20Class%20-%20EJS%20version.pdf
- Regarding other groups taking centuries to regress, I don't recall if it was in a video interview, an article, or 'Farewell'. But I'm pretty sure he did say it because it stuck out in mind mind when I heard it. He may discuss this issue in the article above but I haven't had time to read much of it yet.
Ethnic groups with endogamy won't regress anywhere, these are just their population means.
From that paper:
"[...] the expected value of b in estimating the connection between grandfather and grandson, will now be just the same as for the connection between father and son. Similarly for a father and any more distant descendant, the regression coefficient will be no greater. After one generation there will be no further regression to the mean."
That's one possible theoretical model in which there is inherited social class. That paper (which claims that England had an effectively classless society of social mobility) also seems to have a different perspective from his more recent work. More recently he's written about how wealth being directly inherited shows different patterns from genetic traits, and claimed that basically every society (including egalitarian Sweden and communist China) has little social mobility.
You can keep extending the example back up the family tree. If all 4 grandparents have 120 IQs, does it make a difference if their eight parents averaged 100 or 120 or 140? I can't see why it wouldn't.
Seb, my IQ is 200 (FunIQtest.uwu). If I smash your mom, how clever will our child be? 😈
.
https://x.com/waitbutwhy/status/1805489828705841438
So if I understand correctly it would be more accurate to use the aggregate mean of the two families for regression? Say the Smith’s avg IQ was 117 and the Doe’s 100 — the combined familial mean would be 108.5. So say John Smith (130) and Jane Doe (102) had a son. Using general pop mean of 100: ((130+102)/2)-100) x .6 + 100 = 109.6 but using the combined familial mean ((130+102)/2-108.5) x .6 +108.5 = 113
The Johnson’s avg IQ 125 the White’s avg IQ 105 — combined familial mean 115. If Jack Johnson (130) and Sue White (98) had a daughter. Using general pop mean 100: ((130+98)/2-100) x .6 + 100 = 108.4 but using their combined familial mean ((130+98)/2-115) x .6 +115 = 114.4.
So if Mark Smith (113) and Sally Johnson (114.4)(combined familial mean: Smith’s (117) + Johnson’s (125)=121 have a child we’d expect ((113+114.4/2-121) x .6 +121=116.62 using familial mean whereas we’d expect ((113+114.4)/2-100) x .6 +100=108.22?
I'm not sure if that's how the math would work, but the familial mean would have incremental predictive validity beyond the mean of the general population and the parents.
I see. This would have interesting implications — namely, that family pedigree is quite important, if not more important than a couple’s individual IQs.
You cite the formula for 40% regression as if it were common knowledge. I have seen it detailed only by Stephen Hsu in a 2011 paper, and indirectly by James Flynn in "Does your family make you smarter."
Statistically, it has to be true for the median intelligence and standard deviation (nominally 100 and 15) to remain constant from generation to generation. I posted a video to that effect a few years back.
https://www.youtube.com/watch?v=jXDGdr4m5qI&t=422s
But who am I? I would love to be able to point to an authoritative source. Can you tell me your authority?
emil kirkegaard has a proof somewhere
>, it has to be true for the median intelligence and standard deviation (nominally 100 and 15) to remain constant from generation to generation
no???
Searching my archives, I found the 2011 paper entitled "Investigating the Genomic Basis for Intelligence" by Steve Hsu of Caltech. The full text of a short section entitled: "Your Kids and Regression" is as follows:
Assuming parental midpoint of n SD above the population average, the kids’ IQ will be normally distributed about a mean which is around +.6n with residual SD of about 13 points. (The.6 could actually be anywhere in the range (.5, .7), but the SD doesn’t vary much from choice of empirical inputs.)
So, e.g., for n = 4 (parental midpoint of 160 – very smart parents!), the mean for the kids would be 136 with only a few percent chance of any kid to surpass 160 (requires about a 2 SD fluctuation) For n = 3 (parental midpoint of 145) the mean for the kids would be 127 and the probability of exceeding 145 less than 10 percent.
I can't prove it mathematically, but spreadsheet simulations show that it must be so.
I am too stupid to be able to read this poast.
>But given the high additive heritability of IQ and assortative mating for IQ, isn't it more likely that the grandparents actually come from a high genotypic IQ family line?
Well, yes. The question is whether they are coming from a line that is relatively more genetically endowed than the average person with an IQ of 120 (and I assume that is a no).
Right. If you could also find out the IQs of siblings, uncles/aunts etc. you could probably better guesstimate if the four grandparents each with 120 IQs came from castes where 120 IQ is standard, or if they happened to be higher IQ than most of their relatives.
Reading up on some English intellectual dynasties like the Huxleys (Julian, Aldous, and the least well-known, Andrew who won the Nobel in chemistry), I could imagine people for whom there wouldn't be much regression toward the mean past their grandparents' 120s. The Huxley brothers were descended from TH Huxley on one side and Thomas Arnold, the most famous educator of Victorian times, on the other (well, it's more complicated than that because Andrew was a half-brother of Julian and Aldous).
The 19th Century British intelligentsia clearly did more assortative mating than Richard Herrnstein would have anticipated.
But, most of the time, if your grandparents averaged 120 IQ, they probably enjoyed some genetic good luck relative to their grandparents' average IQ.
Thanks! Everytime someone said RTM I always wondered what "mean" should mean. Family mean, ethnic mean, social class mean, and so on. How does a mean form and how do we know?
It's weird how many people invoke the phrase "regression to the mean" without considering which mean they're referring to. HBD'ers mock the left for believing evolution stops at the neck, then turn around and pretend evolution stops at the racial level.