> The general argument of the Bell Curve, that people from different parts of the world have wildly different distributions on many different attributes
It is by no means "the general argument" of the Bell Curve, and saying that pretty transparently shows that you have not read it. You are, of course, excused by the fact that the book has been the target of relentless smear campaign by media and many academics, which resulted in painting completely wrong image of what the book actually is about.
Bell Curve is almost entirely concerning the American society, and only in passing mentions issues and results from different parts of the world. The point of the book is largely that people within the same part of the world have widely varying outcomes, which, to a large degree, are explained by their IQ. Moreover, which will probably be shocking to people who only know Bell Curve through second hand accounts of people very loudly denouncing it, for the most part it explicitly restricts itself to results and data from the population of white Americans only, to show that these correlations are not a result of some sort of insidious ethnic or racial discrimination, as they also exist within white population. Only in last part of the book ethnic and racial disparities are mentioned, mostly to show that these are pretty much what you would expect if you assumed that correlations of intelligence and outcomes within population of white Americans are the same if you extend your analysis to population at large.
> The point of the book is largely that people within the same part of the world have widely varying outcomes, which, to a large degree, are explained by their IQ.
Except the data they use is not IQ (they are using AFQT Score because that's the available data) and they are artificially constructing an equivalent IQ. The original data doesn't even fit a bell curve (because there's no reason a given test should). For some reason, unjustified in the book, they assume that the AFQT score depends only from genetic factors and not from any kind of social determinism.
The fact that the book liberally jumps from AFQT score to IQ to “genetic factor” is an enormous issue which completely destroys the credibility of their work.
Also, in their book it is “to a large degree, explained by their IQ”, because they controlled for almost no social or environmental factors: only the “parental socioeconomic status” (also a made-up metric) is taken into account.
Which means: for the people with a high enough AFQT score, IQ is correlated with AFQT score. This is indeed interesting, but generalizing to the rest of the population is a methodological mistake.
Moreover, there is no reason to assume the AFQT score to depend only on genetic factors alone (even if I don't see reasons to assume genetics has no impact on it either).
And btw, the study you quote explicitly rules out the conversion done in the book:
> no direct one-to-one correspondence AFQT percentile scores and IQ scores can be stated.
> Which means: for the people with a high enough AFQT score, IQ is correlated with AFQT score. This is indeed interesting, but generalizing to the rest of the population is a methodological mistake.
No, because everything else we know about IQ, especially the existence of positive manifold, strongly suggest that it will also correlate at lower IQs too. You can’t just say “the study on X has not included Y, therefore we should have absolutely no expectations about any sort of relationship between X and Y”. That’s not how science works.
Alas, there have been other studies done, and it turns out that indeed, AFQT/ASVAB score and IQ correlation does in fact extend to areas on the left side of the distribution too, exactly as expected. To reiterate: this correlation exists across entire spectrum, not only among high IQ people. If anything, the correlation at high IQ is lower than expected, because of range restriction effect.
> Moreover, there is no reason to assume the AFQT score to depend only on genetic factors alone
Fortunately, nobody is claiming that, so you won’t have to put much effort to convince anyone otherwise.
> no direct one-to-one correspondence AFQT percentile scores and IQ scores can be stated.
It can, though. The problem is that the population of military test takers is not representative of population at large. For one thing, its average intelligence will probably be higher, due to relative scarcity of mentally disabled people. However, if you know the parameters of IQ distribution among military test takers, you can easily convert AFQT/ASVAB to IQ.
> You can’t just say “the study on X has not included Y, therefore we should have absolutely no expectations about any sort of relationship between X and Y”. That’s not how science works.
Oh yes it is. Extrapolating outside of study sample is either a scientific rookie mistake or just plain fraud. Doing so in any hard-science topic would earn you an angry comment from your reviewers.
> Fortunately, nobody is claiming that, so you won’t have to put much effort to convince anyone otherwise.
Oh, and I guess Herrnstein and Murray didn't publish that book? Because they definitely do so in The Bell Curve: They are literally plotting AFQT and SES (their bogus “socioeconomic factors”) against social outcome and since AFQT has a higher correlation than SES, they conclude that genetics must be the decisive factor!
And if you read my previous comment, you'd notice that the AFQT -> IQ is only a part of the issue in the Bell Curve, the bigger part being how they assume AFQT -> genetics because they assume AFQT = IQ = g = genetics. (The last steps being an ideological stance, which Herrnstein have been defending for decades prior to the publication of the Bell Curve so it's not really a surprise…)
> Extrapolating outside of study sample is either a scientific rookie mistake or just plain fraud.
No, extrapolating outside of study sample is the entire point of doing science. Of what use science would be if we could not do that? Imagine, “no, you can’t say that this vaccine is effective, at best you can say that it was effective in the sample of subjects being included in that study, but you can’t extrapolate that outside the sample, that would be a rookie mistake”. This is, of course, absurd. We do science precisely so that we can make useful predictions in day to day life, outside of studies.
> they conclude that genetics must be the decisive factor!
I observe a goalpost being shifted, from “genetic factors alone” to them being decisive. I do not accept that. Please, tell me, where they, or anyone else claims that genetic factors alone determine outcomes.
> they assume AFQT = IQ = g = genetics
I don’t think they do, and it is contradicted by what you wrote in this very comment, where you say they claim genetics is “decisive factor”. This clearly makes no sense under assumption of equality/identity of concepts — you wouldn’t say that A is a “decisive factor” in B, if A and B are the same thing.
> ideological stance, which Herrnstein have been defending for decades prior to the publication of the Bell Curve
Relationship between g and genetics is an empirical, not purely ideological issue. Hernstein has been saying that g is mostly determined by genes precisely because this is the current state of our scientific knowledge, and has been for decades.
> No, extrapolating outside of study sample is the entire point of doing science. Of what use science would be if we could not do that? Imagine, “no, you can’t say that this vaccine is effective, at best you can say that it was effective in the sample of subjects being included in that study
It's not about not extrapolating to individual subjects outside of the test group, it's about not extrapolating to categories which aren't represented. That's why we don't conclude a vaccine is effective on all mammals after a trial on mouses! We experiment on humans, and we even try to get as much diversity as we can (age, gender, preexisting conditions etc.) so the result can be generalized to the entire population.
> I observe a goalpost being shifted, from “genetic factors alone” to them being decisive. I do not accept that. Please, tell me, where they, or anyone else claims that genetic factors alone determine outcomes.
The argument made in The Bell Curve is that genetics is the single most important factor. I wrote “alone” not because there is not other factors, but because according to the authors there is no other factors as important.
> Relationship between g and genetics is an empirical, not purely ideological issue. Hernstein has been saying that g is mostly determined by genes precisely because this is the current state of our scientific knowledge, and has been for decades.
It's not. At least not according to the usual definition of “scientific knowledge” which imply some degree of consensus: this research field is strongly divided on that question, with a clear ideological split. BTW, even the mere existence of g is questioned.
It is by no means "the general argument" of the Bell Curve, and saying that pretty transparently shows that you have not read it. You are, of course, excused by the fact that the book has been the target of relentless smear campaign by media and many academics, which resulted in painting completely wrong image of what the book actually is about.
Bell Curve is almost entirely concerning the American society, and only in passing mentions issues and results from different parts of the world. The point of the book is largely that people within the same part of the world have widely varying outcomes, which, to a large degree, are explained by their IQ. Moreover, which will probably be shocking to people who only know Bell Curve through second hand accounts of people very loudly denouncing it, for the most part it explicitly restricts itself to results and data from the population of white Americans only, to show that these correlations are not a result of some sort of insidious ethnic or racial discrimination, as they also exist within white population. Only in last part of the book ethnic and racial disparities are mentioned, mostly to show that these are pretty much what you would expect if you assumed that correlations of intelligence and outcomes within population of white Americans are the same if you extend your analysis to population at large.