Monday, November 13, 2006

Academic Tabu

I have just been looking over a recent article by Roland Fryer and Steven Levitt in the American Law and Economics Review, "The Black-White Test Score Gap Through Third Grade." It is interesting for what it tells us about race, but more interesting for what it tells us about the state of academic discussion in present-day America.

The authors observe that, while black students enter the school system substantially behind white students as judged by tests of reading and mathematics, the difference vanishes if you control for a small number of environment variables such as socio-economic status of parents and number of books in the home (the authors are not entirely clear about which variables are used where, but those are two of the ones they mention). That result appears to support the conventional belief that racial differences in outcome are due to environment, not to innate differences.

As the children move through the first few grades of school, however, that situation changes. Black students fall behind white students at about a tenth of a standard deviation a year; that remains true even after controlling for the environmental variables that eliminated the initial difference.

The article is in large part about the authors' search for an explanation. In an earlier piece they had conjectured that the difference was in school quality; in this one, with more data, they are able to reject that explanation. The pattern exists for black and white students in the same school, even in the same classroom. It does not exist for hispanic and white students and it reverses for asian and white students. After exploring a variety of alternative explanations, the authors conclude that they cannot explain the data.

Having eliminated all of the possible environmental causes that they can think of, one might expect them to next consider the obvious alternative explanation: Innate differences between the races. That is not the only possibity, of course; I can think of a few more environmental explanations, such as different treatment of blacks and whites by schoolteachers and others or differences in peer group pressures. If, as I think likely, black students tend to socialize with black students and white students with white students, and if black youth culture is less friendly to education, or differs in other relevant ways, from white youth culture, that also might explain the observed pattern.

Nonetheless, insofar as the results are evidence in the controversy over racial differences, they are evidence in favor of such differences, since they are the results one would expect if blacks were on average innately worse, and east asians innately better, at certain things than whites. That explanation is not on the authors' list of conjectures to be tested. The authors try every environmental explanation they can think of a way to test, are unable to explain the data, and, instead of considering a non-environmental explanation, throw up their hands.

Their discussion of the omitted possibility is limited to a single footnote, which reads:

"This theory, if true, also re-introduces the possibility that genetics could play a role. Because we have little evidence on this either way, we choose to exclude it while noting that it is a possibility."

Including that footnote marks the authors, in my view, as more courageous than the typical academic. Nonetheless, it translates as "our results support a view not to be discussed in polite society, so we won't discuss it."

Which tells us something about to what degree the beliefs, at least the stated beliefs, of polite society are based on open enquiry into the evidence, and to what degree on prejudice.

I should probably add that the one discussion of the question of racial differences I have seen that actually provided real evidence against them was by Thomas Sowell in Ethnic America: A History. It was possible only because Sowell, being more courageous still, was willing to seriously consider the possibility that different racial outcomes might reflect genetic differences. He offered the economic performance of West Indian immigrants as evidence that the poor economic performance of American blacks is due to neither genetics not prejudice. West Indians are blacker than Afro-Americans, in terms of both genetics and appearance, yet their family income gets up to the U.S. average in a single generation. Readers curious about his explanation of the difference are invited to read his (very interesting, for many other reasons as well) book.

42 comments:

Anonymous said...

Everything in my experience with the school system (both public and private, and at all levels) suggests that socialization is indeed the problem.

Anonymous said...

Many, many, studies have shown that, say, IQ, or things that are closely correlated with IQ, such as standardized test scores or grades in school, is partially genetic. Blacks as a group do worse on those measures than Whites (who do worse than East Asians, who do worse than Ashkenazi Jews). Given that, it seems unlikely (though I suppose not technically impossible) that none of the difference is due to genetics.

I don't understand the example of the West Indians (which I assume means people from the western part of India, not the "West Indies"). No one's claiming that the genes for skin color are also responsible for economic achievement, right? One could imagine that whatever genes are common among west Africans and their American descendants that would be responsible for their relatively poorer economic achievement, and are correlated with their genes for skin color, are absent among West Indians, despite the fact that they share genes for skin color. Or did I miss something?

Another interesting fact: there is an average difference in SAT test scores among Blacks and Whites, even controlling for household income. However, if you control for household net worth instead of income, the discrepancy goes away. Unfortunately I don't remember exactly where I read that, but probably it was linked on Gene Expression.

David Friedman said...

"I don't understand the example of the West Indians (which I assume means people from the western part of India, not the "West Indies")."

"West Indians" are people from the West Indies, i.e. the Caribbean. They are mostly descended from slaves brought there from Africa. So they should be genetically similar to Afro-Americans, but with less European ancestry in the mix. That fits their appearance; they look more like subsaharan Africans than the average Afro-American does.

The similarity in skin color between people from subsaharan Africa and people from southern India is presumably due to similar adaptations to similar environments--lots of sunlight. The similarity between West Indians and Afro-Americans is due to common ancestry.

Anonymous said...

If we assume a bias towards endogamic reproduction, people tend to chose similar partners when it comes to skin colors for example, it's not only possible that different IQ arise between different groups, it's also extremely likely, although the larger these groups, the smaller the difference.
Since men have a Y chromosome, some amount of separate evolution between the two genders is also likely yielding different IQ averages, aptitudes etc.

Question is: does it really matter?
No.
The bias is too small to make any bayesian assumption on an individual that is not ridiculously irrelevant when compared to the way he dresses, the way he talks etc.

David Friedman said...

Arthur writes:

"Question is: does it really matter?"

That depends for what. As you say, it isn't very important for judging individuals, since we usually have much better information to go on.

But it is important for evaluating statistical results. Suppose blacks, on average, earn ten percent less than whites, or suppose women make up a small fraction of mathematicians. If one assumes there are no relevant genetic differences, that's evidence about the environment, and suggests some sort of irrational discrimination, or possibly a legacy from past discrimination. And that is, of course, how such results are almost always interpreted--except for the difference in lifespan between men and women.

Once one allows for the possibility of genetic differences, the conclusion no longer follows.

Anonymous said...

All modern-day immigrants are self-selected, so perhaps people who actually immigrate from the Caribbean, who come from the same ancestral population pool and American Blacks, lack whatever genetic trait (or social or environmental trait, but I assume that that is Sowell's explanation) there is in the general population of both American Blacks and Black Caribbeans that is responsible for their relatively poorer economic achievement.

Michael Roberts said...

I think something important is missing here: Cultural bias in the evaluation.

If we dressed everyone in identical once-size fits-all uniforms, we might come up with a strong corelation showing that people from China are better looking than people from Zimbabwe because average builds are different, and the clothes fit Chinese folks better. If the uniform were canary yellow, we might find the reverse.


IF: School as we know it plays to the (genetic, cultural, whatever) differences, some groups may be poorly served _because_ they are treated the same as everyone else. I have learning disabilites. Treating me the same as everyone else was BAD(tm). I fell behind.

Also if: all our standardized tests priveledge the same types of things, we might have a really hard time identifing the bias, becasue it tends to be pervasive.

Which starts to sound like an argument for segregation, but I don't like that idea, because it does smack of intrusive governance.

Take it as an argument for Anarchy instead, which can adapt itself to complex situations much more easily, AND can account for non-typical individuals at whatever level of granulation :)

-Unnr

jimbino said...

The debate on genes vs environment will probably never be resolved.

No matter, the real racist crime in our society is the fact that Blacks pay the same taxes that Whites do for public education but do not participate to the same degree in the benefits, particularly those represented by university education.

The same can be said for Social Security, in which the hard-working Black man, in particular, is expected to die at age 66 when his benefits would begin. Many of those benefits go to White widows who have never worked a whit.

If FDR had thought about it, he would have included prize-winning steers, pigs, cats and dogs in Social Security, since their relatively short life-spans would have resulted in even more benefits for the White widows.

The solution to this unfairness is, of course, to privatize all education and retirement accounts.

Anonymous said...

"The Economist", two weeks ago, published a story on poor white youth in England and they noticed that poor white boys did worse than poor asians and poor blacks at school. The study also says that poor whites are in general more neglet than blacks and asian since they're not seen as minority, so they usaully fall out of government social programs.

Mike Huben said...

I should hope that the reason that they don't bring up the "taboo" subject is that the concept of race is pretty much genetically unscientific. There's too much interbreeding to have distinct populations.

I'm a high school math teacher now, and the blatant factor I see in grades is simply whether the students do their homework. Unless the study quantitatively measures that variable, I wouldn't say they're studying the right thing. I can imagine that subtle differences in home environment strongly affect homework: both explicitly assigned homework and the simple thoughtful referencing of the day's schoolwork by the child in conversation.

Anonymous said...

Mike Huben: That's wrong. Race is definitely a scientific concept. The clustering of haplotypes that you get by analyzing DNA corresponds very closely to self-identified race. In addition, there are drugs that are targeted to people of a specific race, which only makes sense if race has some scientific validity. Etc. There is of course admixture between different groups, but that makes race a somewhat blurry concept, not an unscientific one.

David Friedman said...

Mike writes:

"I should hope that the reason that they don't bring up the "taboo" subject is that the concept of race is pretty much genetically unscientific. There's too much interbreeding to have distinct populations."

The relevant question is whether races as currently defined might differ in the distribution of intellectual characteristics. If the amount of interbreeding in current populations is so large that we can be sure they don't, shouldn't the argument apply to physical characteristics as well?

Shouldn't Mike be arguing that there can be no racial difference in average height or other body chracteristics between blacks, whites, and East Asians, as those groups are conventionally defined?

Indeed, shouldn't the argument even apply to skin color, providing we define black and white not by skin color but by the presence or absence of significant African ancestry? Would Mike like to defend the truth of that implication of his argument?

Mike Huben said...

Allow me to simply refer you to the Wikipedia article on race, which at least has the merit of displaying some of the complexity of this issue.

Your point about black vs white, heights, etc. is better explained with the biological concept of cline than race.

But the biggest problem is the tacit "I know one when I see one" definition of race that tends to be used. It's pathetically unscientific, and if you base any reasoning on it, you are basing your reasoning on gross fallacies. Attempts to make it more rigorous, such as "significant African ancestry" overlook the fact that Africans are far more diverse genetically than any other geographic group, and thus would represent MANY races with independent characteristics. It's simply a bigoted eye that lumps them all together because they have dark skin, instead of recognizing their immense diversity.

In short, when I see thinking about racial differences that uses a credible definition of race, then I may consider it reasonable. Otherwise, it's as scientific as creationists talking about biblical "kinds", and should be taboo among thinking people for the practical reasons that it is factually wrong and socially harmful.

BTW, sorry to hear of your father's death, David. He (like you) was a great teacher and promoter of insightful thought. Even I recognize and credit that, though I don't agree with all of his conclusions. I think I would have liked knowing him too.

David Friedman said...

I don't see the point of Mike's argument, because I don't see the relevance of a biological classification of race to the question. We have fairly well defined subgroups. They differ in a variety of ways, including their ancestry. As long as they are not random selections from the human pool, the possibility of genetic differences remains. Whether you want to call them "races" is simply an irrelevant distraction.

Thus Mike writes: "Attempts to make it more rigorous, such as "significant African ancestry" overlook the fact that Africans are far more diverse genetically than any other geographic group, and thus would represent MANY races with independent characteristics. It's simply a bigoted eye that lumps them all together because they have dark skin, instead of recognizing their immense diversity."

It seems to me that the bigotry is on the other side. Given that Africans represent many genetically diverse races, they don't represent the same collection as Europeans. Hence there is no a priori reason to assume the same distribution of heritable characteristics.

In the case of observable heritable characteristics we can see the difference--for example, the greater representation of certain very tall physical types due to a significant contribution from African "races" (in Mike's sense) with that physical structure. To take the obvious point, a population made up half of Hutu and half of Tutsi might end up with the same aveage height as, say, Italians--but a very different distribution.

I notice that Mike has still not answered my challenge. Given that his argument applies just as strongly to physical as to intellectual characteristics, is he willing to claim that there is no difference in the distribution of physical characteristics between blacks, whites, and east asians as those groups are conventionally defined?

A further problem with his argument is that, while sub-saharan Africans are indeed genetically diverse, they lived in environments that had some (not all) features in common, hence faced some similar selective pressure. Dark skin is the obvious example of the result of that, but there is no reason why there should not be some less obvious ones.

And finally, even if sub-saharan Africans were a random draw from the human gene pool, it doesn't follow that Europeans and East Asians are. Since the relevant questions are about differences between groups, you need to have all the groups random, not just one, to conclude that there are no differences.

I should add that, having met Mike, I agree he and my father would have gotten along well.

Mike Huben said...

Let's sort this out so that you can see where I'm coming from.

The idea of race that you've been using seems to be:
continent-sized groupings of people by geography and superficial skin characteristics (your "fairly well-defined subgroups").

And you want to talk about genetic differences between these groups. Then you want to talk about phenotypic (physical) differences between the groups.

It's about as meaningful as talking about the genetic and phenotypic differences between people surnamed Friedman and people surnamed Smith. Both groups are very mixed bags. Would they have differences? In intelligence? Very likely. But unless you are certain that the variation in developmental environment was less than their variation in genetics, you wouldn't be able to say much about whether that variation in intelligence was due to genetics or culture or nutrition. The recent example of height in the Japanese changing due to higher protein diets after WWII is a good example. Nor can we presume that the measurement tools for intelligence are unbiased: if Friedmans made up the test, there might be a bias in test vocabulary due to a preponderance of some ethnic ancestry.

On the other hand, if you were talking about Icelanders, Bushmen, Polynesians, or other relatively isolated populations, you'd end up with fairly sharp, clear differences in genetics from other populations and a more modal distribution of genetics within the population.

I don't accept the idea of races, because they have no good basis in population genetics. Populations usually have fuzzy boundarys, and it is easy to observe clinal variation between and among populations. Self-identified ethnicity is no more scientific, but has the advantage that it doesn't reek of scientism: it allows for cultural explanations rather than just genetic determinism.

And finally, I view the whole thing as putting a statistical veneer over ancient racist stereotypes. Physical and social environment are enormous determinants of intelligence and learning.

David Friedman said...

Mike writes:

"And finally, I view the whole thing as putting a statistical veneer over ancient racist stereotypes."

Whereas I regard your side of the argument as putting a scientific veneer over an environmentalist dogma.

One routinely sees news stories about observed differences in outcomes by race or gender, whether in education, as in the case that started this discussion, income, or whatever. Almost without exception, such stories take it for granted that the cause is entirely environmental.

As the recent flap at Harvard demonstrated, in the case of gender--where the argument for genetic differences is much stronger--merely raising the possibility that differences in outcome are due to a different distribution of heritable abilities sets off a furious attack.

Nothing in your argument tells us whether the observed difference in school performance, or in incomes, between races as conventionally defined is due entirely to differences in the distribution of heritable characteristics, entirely due to differences in the environment, or (as I think most likely) to some combination of the two. Nothing.

The fact that "blacks" and "whites" as conventionally defined are not races in your sense does not tell us whether the distribution of heritable characteristics in the two groups is the same. Hence it provides no justification for taking it as gospel that all differences are environmental. Hence it is merely a way of avoiding the question instead of thinking about it.


"Physical and social environment are enormous determinants of intelligence and learning."

Surely true. But it does not follow that they are the only determinant, and there is by now massive evidence that they are not.

The whole literature on identical twins separated at birth, adoption studies, and the like shows that intelligence as conventionally measured depends on both heritable and environmental factors, and that the former are of comparable importance to the latter.

Mike Huben said...

Yes, let's use gender as our example of why I think measuring by "races" is erroneous thinking.

Gender differences do sharply divide people into two genetic groups, unlike "races". And those genes produce obvious phenotypic differences including anatomical sexual characteristics, size, hair patterns, hormone levels, behavior, etc.

These have been documented for centuries, supposedly in service to science. And to some extent, they were science (I'm not going to detail the unscientific ones.) As science they are defensible. But even so, the political uses (and often the motives) of the research were aimed at confirming popular stereotypes and justifying the social order of the time: women were weak, stupid, suited only for bearing children and keeping the home.

What have we learned since then? First, that the scales used in those measurements were socially constructed to rank highly the dominant sexes, ethnic groups, and classes. Second, that differences don't necessarily imply superiority of one trait over another: you don't need strength to pull a trigger or to drive a bulldozer. Third, that measures like intelligence are currently ill-defined but must consist of many separate components. Emphasis on one component over another can be a specialization, and may well be an adaptive polymorphism in equilibrium. And fourth, that no matter what is measured, obstacles to women's equal participation are either imaginary, or culturally imposed, or can be culturally alleviated.

The recent flap at Harvard caused a furious attack because an otherwise intelligent man ignored this modern, hard-won knowledge.

This sort of thing was covered at much greater length in Gould's The Mismeasure of Man.

Now, science is defensible even if it says things that are socially unpleasant, even if it will be put to uses that are unjust, unfair, criminal, genocidal, whatever. However, performance on some test measuring a particular culturally determined task and "races" both fail the test of being real science.

And finally, I spotted the big fallacy inyour original post. You write "insofar as the results are evidence in the controversy over racial differences, they are evidence in favor of such differences..." That's wrong because corellation is not cause, and corellation is not evidence of cause. It indicates only a possibility of cause. And you even cited the footnote to that effect.

The reason "That explanation is not on the authors' list of conjectures to be tested" is not a matter of "polite society": it is a matter of the author's recognition of the stringent evidence required to support such a claim.

So am I "putting a scientific veneer over an environmentalist dogma"? I don't think so: the record of science on these subjects is that genetic claims have generally fallen to environmental claims with the exception of a few cosmetic and anatomically sexual characteristics. Where the genetic claims have won, it's been where there are clear-cut causal explanations in biochemistry: for fuzzy claims like intelligence or test scores, environment seems to always be incredibly important. We're not talking dogma here: we're talking about a distinct pattern in science.

Anonymous said...

I think looking back at the history of similar discussions on race or gender differences explains a lot of the political or emotional reaction to these arguments now, but doesn't do much to tell us about whether the arguments are any good.

I took David's original post to be about the bounds of polite discussion. Given evidence that equally supports two statements:

a. The black/white IQ difference is mostly caused by differences in upbringing and physical environment.

b. The black/white IQ difference is mostly caused by differences in genes.

we can expect (and observe) that scientists will actively discuss (a), but probably will minimize discussion of (b).

That doesn't really tell us whether race has physical meaning (though of course it does, albeit an imperfect one). It's about the bounds of polite discussion, and how this can constrain scientific discussion.

Mike Huben said...

Albatross, the evidence is not equal.

If I were to say that the labor theory of value and the marginal theory of value are equally supported, David and just about every economist would snigger. At this point, labor theory of value is considered an ideological relic of Marxism.

There is ENORMOUS evidence that upbringing and physical environment cause differences in measured IQ. There's no question that these vary according to "race" in the colloquial sense.

There is MINOR evidence that genetic differences cause differences in IQ: one of the more obvious ones is trisomy 21 (Down syndrome, aka mongolism). Since brain development is based upon genes (interacting with environment), and there is variation in genes, we'd expect there to be some variation in IQ due to genes. But there is as yet no evidence at all that combines "race" with allelic frequencies as a cause of racial IQ differences. It's possible, but not validated.

Nor can we conclude that variation unexplained by current knowledge of environmental effects must be genetic. The Flynn effect points to an enormous difference (large enough to explain the gap in whites and blacks) which has not yet been attributable to any particular environmental factor. Yet it cannot be genetic unless we can show some very strong genetic selection going on.

You may want to look at the wikipedia article race and intelligence. It covers all this and lots more.

David Friedman said...

Mike writes:

"There is MINOR evidence that genetic differences cause differences in IQ: one of the more obvious ones is trisomy 21 (Down syndrome, aka mongolism). Since brain development is based upon genes (interacting with environment), and there is variation in genes, we'd expect there to be some variation in IQ due to genes. But there is as yet no evidence at all that combines "race" with allelic frequencies as a cause of racial IQ differences. It's possible, but not validated."

You are mixing two different claims. There is overwhelming evidence that genetic differences cause differences in IQ, from identical twin studies, adoption studies, and the like. They give reasonably consistent results, showing that, in the U.S. and similar societies, something like half the variance of IQ is genetic.

The question of whether racial differences are due to genetics or environment is much less clear, both a priori—there is no strong reason to expect them to be—and empirically. But the fact that IQ is largely heritable opens up the possibility.

I agree with your "it is possible, but not validated." My complaint is that the conventional wisdom, enforced by strong social sanctions, is that it is impossible--that merely considering the possibility is wicked. Every article that reports statistical results on differences by race and attributes them to discrimination is implicitly assuming it is impossible.

Mike writes:

"And fourth, that no matter what is measured, obstacles to women's equal participation are either imaginary, or culturally imposed, or can be culturally alleviated.

The recent flap at Harvard caused a furious attack because an otherwise intelligent man ignored this modern, hard-won knowledge."

The first half of that is true on in the sense that "can be culturally alleviated" means "can be reduced." We have certainly not shown that the distribution of intellectual abilities is the same by gender, which is what your claim would otherwise require; such evidence as we have is that it is not the same. And, given that males and females are subject to different selective pressures, it would be surprising if it were.

But once we recognize that the "knowledge" in question is limited to "differences in outcome can be reduced by changes in culture," your statement about the Harvard flap becomes indefensible. What the flap was over was someone raising the possibility that observed differences in outcomes might be due to different distributions of innate abilities. Since we have no evidence against that possibility, attacking him for raising it was pure bigotry and I am surprised to find you defending it.

This is becoming long for comments, but there was one more point I wanted to make. You cite a book by Gould. Have you tried looking at works disputing that book? I haven't read that one, but my experience of Gould, writing about things I know something about, is that he routinely misrepresents the views of people he is attacking, taking advantage of the fact that most of his readers will never see the other side of the argument.

You might want to take a look at the response of the authors of _The Adapted Mind_ to his comments; I think it's webbed somewhere. Also see if Krugman's comment on Gould is still up somewhere; it is not complementary, and Krugman is someone whose views I would expect you to respect.

David Friedman said...

A second response to Mike, on what seems to me a methodological issue. He writes:

"And finally, I spotted the big fallacy inyour original post. You write "insofar as the results are evidence in the controversy over racial differences, they are evidence in favor of such differences..." That's wrong because corellation is not cause, and corellation is not evidence of cause. It indicates only a possibility of cause."

Evidence is not proof. Suppose we have some fact for which there are three possible explanations, all three a priori possible. I claim that if we show one of the possible explanations to be false, we have provided evidence for both of the others. On straightforward Bayesian grounds we have increased the subjective probability of both of them—although we have not proved either.

That particular paper succeeded in disproving, to the satisfaction of the authors, all of the environmental explanations the authors were able to come up with and test. Hence it is evidence for all of the remaining explanations, of which genetics is one.

David Friedman said...

I mentioned Krugman's comment on Gould, and a little googling found a sample, at:

http://www.pkarchive.org/theory/evolute.html

Krugman describes Gould as the Galbraith of evolutionary biology, and he doesn't mean it as a complement.

Anonymous said...

There's no clearer example of the academic left's abandoment of reason that the argument that race does not exist, an idea so blindingly obviously false that only a very highly educated person (and intellectually gullible) could possibly find it compelling.

The first problem is the idea that scientific truth is different from everyday truth. Sorry, but scientific truth, unlike Biblical truth can't be different from the plain old regular truth. Dropping the term scientific into the discussion is just a distraction from the descent into sophistry.

And at this level of sophistry almost every concept that exists, scientific or otherwise has a fuzzy boundary problem. Is color a meaningful scientific concept? Certainly not. When does yellow become green? What category does mauve belong to?

Even gender has fuzzy boundaries. There are people born with female genitalia but XY chromosomes and vice versa. Based on Mike's reasoning, this fuzzyness must also torpedo any discussion of genetically based gender differences. Are men taller than women, on average? Can't say, since I don't know, scientifically, what precisely a woman is. Do Africans on average, have darker skin than Europeans? Scientifically, I don't know. La la la la la!

Perhaps the most absurd aspect of this whole Orwellian exercise is the assumption that if race could be defined away that racial differences and racial problems would go away. I think a race-blind society is an admirable goal, but the way to get there is by discounting race not by pretending that it doesn't exist.

Ulitimately this sort of nonsense just makes academics look like irrelevent clowns and discourages the brightest and most curious undergraduates students from even joining this silly frilly Mad Hatter tea party.

-Mercy

Mike Huben said...

I think you are misinterpreting the meaning of heritability in twins tests and the like. Even if there was 100% heritability of intelligence, environment could still change intelligence of one twin an arbitrary amount. Twin studies and the like generally attempt to control for the environment. However, real populations will have different envronments, whose effects could have pretty much any ratio to those of the genes. For example, exposure to lead paint in impoverished children, but not in middle class children.

So when I was talking of minor evidence, I meant very much the same thing you say: "the fact that IQ is largely heritable opens up the possibility." But there is a vast and conclusive pile of evidence that environment greatly effects intelligence.

As for your complaint against "conventional wisdom, enforced by strong social sanctions", I don't think much of it. Egalitarian research was able to defeat the previous racist "conventional wisdom, enforced by strong social sanctions": why shouldn't new racist science be able to overcome such hurdles if it really has scientific merit? But I don't think those are the real hurdles: I think the real problem is that race is ill-defined and that it's too early for science to understand the genetics involved and too difficult to measure all the environmental influences.

"What the flap was over was someone raising the possibility that observed differences in outcomes might be due to different distributions of innate abilities. Since we have no evidence against that possibility, attacking him for raising it was pure bigotry and I am surprised to find you defending it."

Of course there are different distributions of innate abilities between men and women. The apalling thing was that he was suggesting a scientifically unsupported mechanism to explain what is already clearly due to sexual prejudice and an old boy's network. The literature on the subject is overwhelming, and Harvard has repeatedly lost civil suits in court over this. He was defending simple bigotry with a spurious explanation. It's not bigotry to oppose bigotry: you're blaming the victims.

As for Gould, I've hardly ever liked his work. I've had that Krugman quote at my web site for years. I side with Ed Wilson on Sociobiology and Evolutionary Psychology in disagreement with Gould. His last work about "separate magisteria" for relgion and science was accomodationist twaddle. I've met him, and observed him in action, and not liked his attitude. I read Mismeasure Of Man more than 20 years ago, and might re-evaluate it differently, yet still I think it is an important antidote to genetic determinists.

"I claim that if we show one of the possible explanations to be false, we have provided evidence for both of the others." This is the same as the creationist fallacy that if they disprove evolution, you should believe in creationism. It is based on a false dichotomy, that one of the suggested theories must be correct. But of course, there could be yet another unmentioned theory that is actually the correct one, in which case showing the original theory to be false does not provide any evidence that other suggested false theories are true. In this case, some environmental causes were ruled out, but you yourself claimed there were others which were not considered. "Bayesian grounds" don't rescue you from this problem: see Probabilities of probabilities. In particular, you cannot model this probabilistically because of the non-random selection of the theories that were tested.

David Friedman said...

1. I'm glad to see that your view of Gould is similar to mine. Nonetheless, in your previous post, you took his side of the argument as gospel when it supported a position you wanted to believe.

2. You write: "why shouldn't new racist science be able to overcome such hurdles if it really has scientific merit?"

The word "racist" is not a value-free descriptive term. Yet you use it to describe scientific research—hypothetically correct scientific research—which (hypothetically) supports a conclusion you dislike. Think about what that implies about the degree to which you are thinking with your guts rather than your head.

Mike writes:

"Even if there was 100% heritability of intelligence, environment could still change intelligence of one twin an arbitrary amount. Twin studies and the like generally attempt to control for the environment."

What such studies estimate is how much of the variance of measured IQ in an actual population is due to environment and how much to genetics. If the conclusion were that 100% was genetic (it isn't), that would mean that no environmental variation in that population was large enough to have any effect on measured IQ.

"The apalling thing was that he was suggesting a scientifically unsupported mechanism to explain what is already clearly due to sexual prejudice and an old boy's network."

It's amazing how your standards of evidence change according to whether you like the result. Precisely how does one do an experiment, or statistical analysis, to demonstrate what fraction of the difference between the fraction of (say) mathematicians who are male and the fraction who are female is due to sexual prejudice and an old boy's network? Yet you are asserting, with absolute certainty, that 100% of it is. You are so certain that you regard it as outrageous for someone to raise the possibility that some of it might have another cause.

"This is the same as the creationist fallacy that if they disprove evolution, you should believe in creationism. It is based on a false dichotomy, that one of the suggested theories must be correct."

I think you are missing the distinction I already made between proof and evidence. I suggest a simple Bayesian claculation. Start with three alternatives of which you know one and only one must be true, and each of which has positive prior probability. Now add one additional fact—that alternative 1 is false. calculate the posterior probabilities. If you wish, repeat for N alternatives.

A fact that raises the probability of a proposition is evidence for that proposition—even if evidence short of proof.

Anonymous said...

Any model that explains the gender differences in mathematical fields in terms of discrimination needs to explain the pattern of discrimination. Plenty of women graduate in law and medicine, relatively few in engineering or computer science. Similarly, plenty of women graduate in biology and psychology, far fewer in physics and economics.

Now, there's not any obvious reason I can see why medical and law schools would be *less* subject to discrimination and old-boy networks than engineering and computer-science departments. I can't see an inherent reason why biologists would be less inclined to discriminate against women than chemists or physicists, or why mathematicians would be especially into discriminating against women.

There's a pretty obvious model here, which says that men tend to be a bit better at math than women, and that the more mathematical the discipline, the more likely men are to dominate it. I'm not claiming that's right--it may be completely wrong--but it tracks with some available data (men tend to be better at some kinds of geometric reasoning), and it matches the observations you can make in any university or office.

Andrew said...

How do they control for racism in the study? I haven't read it, but I doubt they have a good way to do so. Even if you compare students in all-black neighborhoods who go to all-black schools to all white ones, there is still exposure to the rest of the world. What do you think it is like to grow up watching television where everyone who looks similar to you and your parents is either a sports star or an idiot?

Throughout America, blacks are treated differently than whites (and asians, and hispanics), purely because of race. The defining characteristic of being "black" is not any particular genes, but just an appearance that people recognize as "black" and thus getting treated as "black" by most of American society which consciously or subconsciously treats black people differently.

Asians face a different sort of racism. As they are stereotyped to be smart and studious, even those who come from poorer backgrounds (rather than the wave of asian immigrants from the 60s/70s who came for PhDs) may be lifted by this stereotyped. Teachers and society may expect more from them simply because they look like other asians who have indeed been smart and studious.

My estimate is that the environmental impact of racism and race culture is several orders of magnitude higher than that effect of genetic differences between races.

Anonymous said...

andrew:

Er, are you really saying that being black has nothing to do with genes? Or that seeing someone who is obviously black in the US tells you no statistical information about their genetic makeup?

It would be nice to rule out the possibility that the black/white IQ difference is largely genetic, but this needs to be done based on evidence, not hope. And the evidence needs to be evaluated in the same way when it points toward the result you want as the result you don't.

Mike Huben said...

"... you took [Gould's] side of the argument as gospel when it supported a position you wanted to believe."

Ah, an ad-hominem accusation of confirmation bias. Backed by the libertarian power of mind-reading: my mention that a book covers an issue shows not just that I give it some weight, but that I consider it gospel.

"The word "racist" is not a value-free descriptive term." Fine. I'll give up racist as soon as you give up "free market", because that's not value-free either. And once again, you're relying on your imaginary libertarian mind-reading power: I did consider whether the word was appropriate because of historical associations. Racist has the exact right dictionary meaning: why don't you see if you can find a synonym with the same recognition.

"What such studies estimate is how much of the variance of measured IQ in an actual population is due to environment and how much to genetics."

I think you misunderstand. First, these are not actual populations (in the sense of race). Second, actual populations have different environments whereas, in twin studies, environments are controlled. When environments are controlled, varaince due to environement is minimized. In actual populations, variance due to environment BETWEEN POPULATIONS would not be minimized, and thus could be MUCH greater and directional.

"It's amazing how your standards of evidence change according to whether you like the result."

Ad-hominem accusation of confirmation bias yet again. I could easily claim the same thing for your position, but the pickings are easier when you don't seem to understand the import of the studies.

"Yet you are asserting, with absolute certainty, that 100% of it is."

I never fail to be dazzled by libertarian mind-reading. It's like you're inside my head or something! Not.

"Precisely how does one do an experiment, or statistical analysis, to demonstrate what fraction of the difference between the fraction of (say) mathematicians who are male and the fraction who are female is due to sexual prejudice and an old boy's network?"

Why, that's a fine question, and applies to many other fields as well. Let me appeal to your confirmation bias and cite Bryan Caplan (another libertarian economist) as he explains that certain hypotheses are better than others even though both are possible:

"An economist who attributes hyper-inflations to radically and continuing declines in the demand for money contradicts no economic theory. He is
however still a bad economist, because he analysis of which factors are quantitatively significant is so far off."

This is an incredibly important explanation of why so many libertarian arguments are just plain wrong. It can be rephrased: "In a rich theoretical environment, there can be many explanations of real situations all of which are theoretically correct. But chances are that very few are quantitatively correct, ie. reflect sufficient real effects to constitute an accurate model."

We know that there are MANY environmental influences on intelligence of populations which have LARGE measured causal effects. We know that genetic influences are possible, but have no measured differences between populations controlling for environment, let alone measuring genetic differences.

As for your idea that falsifying a competing hypothesis is evidence for your own, I still say it is wrong. On several grounds.

First, it fails the giggle test: by your standard, if I think of any other competing hypothesis, no matter how ridiculous, then I would have to decrease the probability of the other hypotheses.

Second, you're relying on subjective choices of how to interpret it as evidence. You suggest a simple Bayesian test: but say we use a slightly more sophisticated model. For example, we group hypotheses according to environment vs. heredity, and assign probabilities to the groups. Then, elimination or addition of a hypothesis to one group will not change the probabilities of any hypotheses in the other group. And of course the original assignment of probabilities and the assumption that elimination of a hypothesis affects ALL of the other probabilities in one direction are also subjective choices.

Third, I can't think of any scientific field where Bayesianism is used to rate hypotheses by all competing sides. Perhaps one side uses it as a scientismic rhetorical club. I don't think there's any good way to assign stable enough probabilities. Where do you assign Bayesian probabilities to rival theories in your published academic work, David?

M.C. said...

Yes, taboos absolutely do affect the direction of academic inquiry.

Anonymous said...

Mike Huben,

Claiming that you are wrong for reasons about you, rather than for reason about your reasoning, would be an ad hominem attack. Claiming that you've erred in your reasoning is not an ad hominem attack.

Concerning Bayesian methods, you don't even need them to get the conclusion David reaches. Let A = Pr(Race differences in academic performance are caused by prejudice.) and let B = Pr(Race differences in academic performance are caused by heritable factors associated with racial groups). Let C = Pr(Race differences in acedemic performance are cause by something else other than prejudice or racially linked heritable attributes.)

By construction, A + B + C = 1. The study David cites is evidence for revising C downward, whether by Bayes rule or by any other methodology which recognizes that the probabilities must sum to 1. A and B then have to increase in such a way that the sum of the increases is equal to the decrease in C.

Mike Huben said...

James:

"... you took Friedman's side of the argument as gospel when it supported a position you wanted to believe."

So now, I'm only criticizing your thinking, right? How convenient! We can dismiss all libertarians arguing simply because they exhibit the grossest sorts of confirmation bias. Heck, we can then go on to dismiss all science, religion, politics, and other beliefs of anybody. Anything to avoid actually examining the argument to see if it makes sense.

Look, if David wanted to address a position, he'd say something about the position, not about the person who presented it. He doesn't refute Gould's position: at most he points out that he and I often disagree with Gould and often don't respect Gould's positions. Yet he and I probably agree that Gould's belief in evolution is justified: we disagree in details at most. So we cannot rely on Friedman's ad-hominem dismissal of Gould.

"A and B then have to increase in such a way that the sum of the increases is equal to the decrease in C."

Only one of them has to increase: the other could well decrease or stay the same.

Anonymous said...

Huben,

"... you took Friedman's side of the argument as gospel when it supported a position you wanted to believe."

"So now, I'm only criticizing your thinking, right?"

Sure. In your own words, it would be an "accusation of confirmation bias," but not an ad hominem attack, as the content of such an accusation would be concerning my thinking, not my person.

"Only one of them has to increase: the other could well decrease or stay the same."

Sure, but this would only be warranted if the evidence against some part of C implied a revision concerning the relative risk of A with respect to B.

Mike Huben said...

No, James, if the hypotheses were grouped as I suggested to David, then it would not require additional information.

If you think criticizing thinking in a non-specific manner (such as claiming confirmation bias) is not ad-hominem, then you're crazy. Get it?

Anonymous said...

Mike,

"...if the hypotheses were grouped as I suggested to David, then it would not require additional information."

How, according to your proposed system, would you determine or revise group level probabilities? Only if you hold the group level probabilities constant, can you get the result you describe, but holding probabilities constant at any level is equivalent to selectively ignoring evidence.

"If you think criticizing thinking in a non-specific manner (such as claiming confirmation bias) is not ad-hominem, then you're crazy. Get it?"

Criticizing thinking by specifying the type of bias perceived doesn't strike me as non-specific. As far as whether or not this is an ad hominem, I recommend visiting this description which you link from your own website.

Look, it's ok. David claimed that you committed a fallacy and you mistakenly thought that he was making an ad hominem attack. You can come up with theories about my mental health all you like, but no one has made an ad hominem attack against you. You should be pleased by such civil treatment, maybe even reciprocate.

Mike Huben said...

You know, James, if you actually READ the source you link, you'd see that accusations of confirmation bias are an example of ad-hominem called "poisoning the well". The same as calling you crazy, since you didn't get the joke.

Your infantile requests for methodology are no better than a two year old repeatedly asking "why?" David started with an overly simplistic model: I shouldn't be held to any higher standard of precision than he is.

As I've said in the past, I'll take advice on discussions from you when you finally show evidence that you understand them well enough to contribute.

Why don't you just let David stand up for himself?

Anonymous said...

Mike Huben said:

"We know that there are MANY environmental influences on intelligence of populations which have LARGE measured causal effects. We know that genetic influences are possible, but have no measured differences between populations controlling for environment, let alone measuring genetic differences."

We also know of many large genetic influences on intelligence, in terms of genetic disorders that lead to either mental retardation or to more specific mental problems.

Twin studies give us really strong evidence that some large fraction of intelligence (as well as personality and other related stuff) is based on genes. Identical twins separated at birth have much more strongly correlated IQs than fraternal twins separated at birth; unrelated kids raised in the same homes have little or no correlation in IQs as adults.

That doesn't tell us whether the black/white IQ difference is genetic, but it does give us strong evidence that there is a genetic component to IQ scores. Now, we face two groups with relatively large genetic differences and an IQ score gap that's persisted with small decreases since the first IQ tests were developed, and strong evidence that the differences in IQ scores are not much affected by cultural bias in the tests (tests with no apparent cultural component give about the same IQ difference as culturally-loaded tests, IQ scores predict school outcomes for blacks about as well as for whites, and when you rank the difficulty of specific questions on a IQ test for both black and white test takers, they find the same questions harder and the same ones easier).

With all that, the test score difference may be entirely environmental, with the obvious guesses about differences being internal culture (how you're raised) and external culture (how you're treated), both of which are very different for blacks and whites. But it's not like there's no reason to entertain the question of whether the difference could be caused mostly by genetics.

Anonymous said...

"You know, James, if you actually READ the source you link, you'd see that accusations of confirmation bias are an example of ad-hominem called "poisoning the well". The same as calling you crazy, since you didn't get the joke."

No. Claiming that the other person has committed a fallacy, such as falling prey to confirmation bias, is not an ad hominem attack, nor is it poisoning the well.

"Your infantile requests for methodology are no better than a two year old repeatedly asking "why?"

Instead of embarrassing yourself by resorting to pejoratives, why don't you defend your position? If you're right, this stuff is entirely superfluous.

"David started with an overly simplistic model:"

So you want to claim that it was overly simplistic and then refuse to explain why the parsimony of DF's approach is a problem. Ok. Have at it.

"I shouldn't be held to any higher standard of precision than he is."

Look, if you offer a more complex methodology, there's more to specify. I'm unsure why you are not more enthusiastic about clarifying your methodology for joint hypothesis testing.

DF's description of his approach, while admittedly parsimonious, was such that I could find a charitable interpretation. With yours, I got the impression that you were suggesting that group level probabilities should be held constant even as the probabilities of theories within groups are revised. Since this strikes me as incompatible with what it means to assign a probability to a group of hypotheses, e.g. Pr(theory A is true OR theory B is true OR ... OR theory N is true), I thought I must have misunderstood you. Given your present occupation, I couldn't imagine that you'd make such an error, so I thought I'd let you clarify.

"As I've said in the past, I'll take advice on discussions from you when you finally show evidence that you understand them well enough to contribute."

Let's forget about you taking my advice on discussion for now. What I really want to know is when you'll clarify and defend your novel approach to the testing of joint hypotheses.

"Why don't you just let David stand up for himself?"

Exactly what are you on about now? David can post and comment whenever he pleases, whether or not I "let" him do so.

Mike Huben said...

James wrote:
"No. Claiming that the other person has committed a fallacy, such as falling prey to confirmation bias, is not an ad hominem attack, nor is it poisoning the well."

Unfortunately, James has zero reading comprehension. The site he referes to says:
"This particular form of Argumentum ad Hominem, when you allege that someone is rationalizing a conclusion for selfish reasons, is also known as "poisoning the well.""

Confirmation bias is an obvious selfish reason.

But I'm not going to bother to answer all your ridiculous arguments. A few months ago, I bookmarked a blog post titled Rhetorical Strategy which matches very closely your style of defocusing discussions from their original intent.

I view such rhetorical strategems as injurious to real discussion. Which is exactly why I prefer to discuss this subject with David rather than you.

Now, I suppose you'll play innocent here, or ask me how that applies to you. Grow up.

Anonymous said...

Huben,

Claiming that you've fallen prey to some bias is not a personal attack. If you want to believe someone has made a personal attack against you, go right ahead. I doubt that David would; he is, if nothing else, polite. But believe what you please.

"Now, I suppose you'll play innocent here, or ask me how that applies to you."

Interesting forecast, but I couldn't play innocent if I wanted to and I'm rather sure you'd rather embarrass yourself with pejoratives than defend the idea that your link pertains to me.

Besides, I'm far more interested in seeing a defense of your novel approach to assigning constant probabilities to joint hypotheses even as evidence is discovered to undermine the individual constituent hopotheses.

John Lott said...

I don't know if it matters, but the QJE abortion and crime paper by Donohue and Levitt is about the only economics paper on crime that has been published in the last thirty years that doesn't account for racial demographics. While one would normally account for racial demographics in crime, it would seem especially useful in this case because one of the alternative explanations is eugenics. That African-Americans had abortions at twice the rate of whites and that they also tend to have higher crime rates. One would seemingly want to control for this just to make sure that it is a change in human capital per se and not the changing racial composition of the population that might explain any changes in crime rates. I believe that there paper falls apart for other reasons and that the relationship between abortion and crime is the opposite of what they claim, but, while I could be wrong, I assume that they didn't want to have to deal with these issues of race.

Anonymous said...

Mike Huben wrote, "I don't accept the idea of races, because they have no good basis in population genetics."

But H. Tang, T. Quertermous, B. Rodriguez et al. [Genetic Structure, Self-Identified Race/Ethnicity, and Confounding in Case-Control Association Studies (2005)] have documented "a nearly perfect correspondence between genetic cluster and self-identified race/ethnicity for major ethnic groups living in the United States, with a discrepancy rate of only 0.14%."

Referring to that study (which is one among many), Charles Murray wrote in The Inequality Taboo (2005):

"When a statistical procedure, blind to physical characteristics and working exclusively with genetic information, classifies 99.9 percent of the individuals in a large sample in the same way they classify themselves, it is hard to argue that race is imaginary."