Thursday, February 17, 2005

Summers firing

OK, you can read the entire transcript of Larry Summers' controversial remarks here (NBER Conference on Diversifying the Science & Engineering Workforce). Remember, he was invited to be a provocateur and explicitly stated he wasn't speaking as Harvard President, but as an economist. Do these remarks warrant his removal as president of Harvard? (The issue is that a small gender-related difference in standard deviation leads to huge asymmetries in the tails of the ability distribution - many more really dumb men than women and many more math genius men than women. Who knows whether it's true, but does it justify his firing?)

"...I'm focusing on something that would seek to answer the question of why is the pattern different in science and engineering, and why is the representation even lower and more problematic in science and engineering than it is in other fields. And here, you can get a fair distance, it seems to me, looking at a relatively simple hypothesis. It does appear that on many, many different human attributes-height, weight, propensity for criminality, overall IQ, mathematical ability, scientific ability-there is relatively clear evidence that whatever the difference in means-which can be debated-there is a difference in the standard deviation, and variability of a male and a female population. And that is true with respect to attributes that are and are not plausibly, culturally determined.

If one supposes, as I think is reasonable, that if one is talking about physicists at a top twenty-five research university, one is not talking about people who are two standard deviations above the mean. And perhaps it's not even talking about somebody who is three standard deviations above the mean. But it's talking about people who are three and a half, four standard deviations above the mean in the one in 5,000, one in 10,000 class. Even small differences in the standard deviation will translate into very large differences in the available pool substantially out.

I did a very crude calculation, which I'm sure was wrong and certainly was unsubtle, twenty different ways. I looked at the Xie and Shauman paper-looked at the book, rather-looked at the evidence on the sex ratios in the top 5% of twelfth graders. If you look at those-they're all over the map, depends on which test, whether it's math, or science, and so forth-but 50% women, one woman for every two men, would be a high-end estimate from their estimates. From that, you can back out a difference in the implied standard deviations that works out to be about 20%. And from that, you can work out the difference out several standard deviations. If you do that calculation-and I have no reason to think that it couldn't be refined in a hundred ways-you get five to one, at the high end.

Now, it's pointed out by one of the papers at this conference that these tests are not a very good measure and are not highly predictive with respect to people's ability to do that. And that's absolutely right. But I don't think that resolves the issue at all. Because if my reading of the data is right-it's something people can argue about-that there are some systematic differences in variability in different populations, then whatever the set of attributes are that are precisely defined to correlate with being an aeronautical engineer at MIT or being a chemist at Berkeley, those are probably different in their standard deviations as well. So my sense is that the unfortunate truth-I would far prefer to believe something else, because it would be easier to address what is surely a serious social problem if something else were true-is that the combination of the high-powered job hypothesis and the differing variances probably explains a fair amount of this problem."

Q: You know, in the spirit of speaking truth to power, I'm not an expert in this area but a lot of people in the room are, and they've written a lot of papers in here that address ....

LHS: I've read a lot of them.

Q: And, you know, a lot of us would disagree with your hypotheses and your premises...

LHS: Fair enough.

Q: So it's not so clear.

LHS: It's not clear at all. I think I said it wasn't clear. I was giving you my best guess but I hope we could argue on the basis of as much evidence as we can marshal.

Q: It's here. (Referring to papers?)

LHS: No, no, no. Let me say. I have actually read that and I'm not saying there aren't rooms to debate this in, but if somebody, but with the greatest respect-I think there's an enormous amount one can learn from the papers in this conference and from those two books-but if somebody thinks that there is proof in these two books, that these phenomenon are caused by something else, I guess I would very respectfully have to disagree very very strongly with that. I don't presume to have proved any view that I expressed here, but if you think there is proof for an alternative theory, I'd want you to be hesitant about that."

22 comments:

Anonymous said...

Agreed!

Anne

Anonymous said...

Comments are sluggish for several days, Steve. Possibly here, possibly there?

Anne

Anonymous said...

When exactly are lynchings justified in your opinion?

Steve Hsu said...

Anne: Blogger has been congested for a few days. Seems to be better now...

I guess since lynching implies violation of due process, it is never really justified :^)

Probably not the best choice of words on my part, although the concept of mob justice does seem appropriate.

Anonymous said...

Agreed 100%. I don't think he should be (or will be) fired over this.

However, I am glad to see his feet being put under fire. I am unimpressed with his 'economic brilliance', as in this discussion or his brilliant insight about "the insufficient use of the third world's capacity to absorb pollution": I think many people (of all stripes) can entertain several such hyotheses or sources of 'economic inefficiencies'. And he is a bit too quick to pronounce people as anti-Semitic (which clearly does not look good on anyone's resume).

varun

Anonymous said...

I'm astounded and ashamed at the heat Summers is getting for proposing a perfectly plausible statistical hypothesis (and clearly stating it as such). We MIT grads thought that's how science was done: hypothesize, test, refine. It's another example of political correctness trumping the scientific method.

Did you know Summers is the nephew of two Nobel laureates in Economics (Paul Samuelson and Kenneth Arrow)?
http://www.atlanticblog.com/archives/001083.html

Steve Hsu said...

DB,

See the previous post on Summers and Asperger syndrome for more on his genealogy - dad was an econ prof, too!

Judging by the transcript Summers was *far* from aggressive in the NBER talk, at least by physics standards :-) Things may be very different in other fields...

BTW, for those who don't read the whole transcript, he says right at the beginning that societal factors and discrimination probably play important roles as well - but why is the situation worse in the hard sciences, as compared to law, medicine or the humanities?

Anonymous said...

The speech was finely and sensitively written, and might be argued with at length to much avail. There was nothing wrong with the speech; the lack of extensive friendly argument however is a loss.

Anne

Anonymous said...

Interesting to find how Ernst Mayr would have conducted the argument.

Anne

Anonymous said...

There is a most difficult but important article from Malawi in the New York Times, but this tread was not the proper place for this posting. There will be a place.

Anne

Anonymous said...

Your “defense“ of Summers deserves a few strong criticisms.

“Lynching” - Summers is under attack for statements he made, but this is a far cry from lynching. Clarence Thomas certainly abused this phrase, but his mis-use or ignorance need not be propagated any further. There are many historical documents the address the practice of lynching, and here is one that I found using a quick Google search,
http://www.liu.edu/cwis/cwp/library/african/2000/lynching.htm

Suffice it to say, that Summers’ current situation is by no means a lynching. He is in a dilemma of his own making, and he deserves harsh treatment.

Summers, as the president of a prominent university, cannot expect the latitude to exercise the unbridled freedom of an academic researcher. Steve’s recent post regarding Carly Fiorina and the attributes required of a successful CEO have bearing on this issue. Summers leads a large and powerful corporation and while he may encourage unbridled creativity, his job is to lead (…and of course fundraise). He cannot simultaneously play freely in the academic sandbox and fulfill the responsibilities of the Harvard presidency.

In brief, Summers made a stupid mistake and it certainly warrants the scrutiny he is now facing.

WJEV

Steve Hsu said...

WJEV,

I agree, lynching is a bad term to have used.

I also agree that Summers should have anticipated the PC furor his remarks elicited, and hence as a CEO should probably have not made them at all.

The scrutiny he is under now is warranted not only because of his comments, but because he has exhibited an abrasive management style.

But, I don't think the comments alone justify his removal as president.

Carson C. Chow said...

I have to agree with WJEV on this one. As a university president, the message that comes across is "It's not my fault that there are huge gender disparaties in some departments because it could be due to intrinsic differences." Even if we could test and verify his hypothesis, what would it solve? Would everyone suddenly feel comfortable with the lack of women? Are we going to recruit based on the expected statistical distributions? Who is to say how far out on the tail you need to be to be on the faculty? What does it even mean to be out on the tail? What measure are you using? It seems to me that the only measure that is meaninful is whether or not you are successful in science and that can't rule out anything. He saw what happened ten years ago with the Bell Curve. Now he brings this up? I think Summers will be forced to step down.

Anonymous said...

The speech was no doubt written with fine intent, not to offend, and apologies were made, and a proper chance for more discussion and change has arisen. Then we can criticize the ideas, sharply, and expect change, but I do not find any need to set aside Larry Summers.

For me, the most telling point is the extent to which there are women in biology. Ah, we hear, how often do I hear, biology is a lesser science. Real scientists opt for chemistry, no applied physics, no theoretical physics and math. Could it be that biologists know too much of how distinct the individual is, how abstract the mean?

We get trapped in our fields. What does Summers know of biology, of Darwin and population study in biology? Where is the least evidence that would separate genetic ability from cultural influences. Means were abstractions to Darwin and Ernst, while reality resided in the individual. Change the culture; really change the culture slowly I suppose but persistently and watch patterns of behavior change. Have they not done so slowly through the years? But do not expect cultural typings to stay as they have been, for there is no necessity to the patterns.

Anne

Steve Hsu said...

Carson wrote:

Who is to say how far out on the tail you need to be to be on the faculty? What does it even mean to be out on the tail? What measure are you using? It seems to me that the only measure that is meaninful is whether or not you are successful in science and that can't rule out anything.Uhh... are you saying there is NO correlation between traditional metrics (GRE or SAT scores, success on the Putnam, whatever) and success in science (say, particularly, in theoretical physics)?

Unless you are, the implications are pretty clear. In case you haven't looked into the data that Summers referenced, it shows that already at a young age (like 10 or 12) there are significantly more "gifted" (say, top 5% on these (not useless) tests) males than females (he says 2x, but I think that is actually charitable relative to the numbers). Assuming a normal distribution (central limit theorem and all that, or just looking at the scores for large samples), you can quickly get to Summers' conclusion.

Now, of course you could say that already by age 10-12 the infernal effects of male genes/hormones - oops, I mean societal pressures and sexism - have caused this discrepancy. Fine - Summers never had to claim that any of this was genetic in basis. But once the effect is there at age 12, you no longer have to invoke sexism at later stages (e.g., in the faculty hiring process) to explain the outcome: more males in jobs requiring +4 sigma deviation.

I certainly believe there are discriminatory barriers in place - I'm not claiming there are not. But it is amazing how easy it is for the PC armies to simply dismiss the effect that Summers was talking about - "Can't be true. Of course it's wrong. Meaningless tests. What? Oh yes, we do use GRE scores to filter our grad admissions, yes, well we do see correlations between GRE scores and success in our program..."

BTW, the difference in mean scores is small in the studies, and less sophisticated critics of Summers keep repeating the mantra that there are no significant differences in M/F math performance, without really understanding his point.

Anonymous said...

Steve I agree with you. Finding on ability differences are not inherently threatening provided institutional supports of the differences are being actively worked against.

Anne

Anonymous said...

There is no need to be angry with Larry Summers, just to use the dialogue to foster institutional integration or the conditions for institutional integration.

Anne

Anonymous said...

This debate seems to have left the issue of Summers (who made a big mistake and may be forced out of his position) and moved on to the subject of his comments. I think he posed some interesting hypotheses, but the system (gender participation differences in science) is far too complex to be explained by simply a single parameter, intelligence. Beyond some lower threshold, it is not a good metric to determine success. And this ignores the issue that there is no reliable way to quantify intelligence, especially in the tails of the distribution. Intelligence is not a scalar, but a multifaceted parameter that has many different components (mathematical, physical, geometrical,…).

I know a highly regarded theoretical physicist who could not repair a hole in a bicycle inner-tube. Fortunately he lives in a highly developed country, the US, and could buy a new tube or pay someone to fix it (or hood-wink and experimentalist into doing it for free). However, if he lived in a subsistence farming community in the Australian outback, I am pretty sure he would be considered stupid. So beyond some general set of criteria, evaluating intelligence depends on the application (e.g., theoretical physics vs neuroscience).

Carson’s suggestion that standardized tests are not good indicators of success is probably correct. These tests establish the suitability of one’s abilities and educational background to the higher education system (i.e., evaluates some necessary conditions), but many other factors influence one’s career path. I do not mean to say that there is no way to evaluate candidates and make some predictions about future success, but intelligence alone is not a good indicator, and there are many external influences that must also be addressed.

I think this debate is converging toward a question of nature versus nurture. I am decidedly on the side of nurture. The ability to comprehend and solve complicated problems is not sufficient to assure one’s success. The ability to grasp, utilize and communicate the importance of ideas is an equal or probably greater indicator of future success.

WJEV

Carson C. Chow said...

Steve,

Actually, I understand his point exactly. We could even try to test his hypothesis by seeing whether or not male brains have more variation in some measure like synaptic density or size of prefrontal cortex, etc. I wasn't arguing the science per se.

However, I don't think we're ready for such a study. The argument is that if there are small discrepencies in the mean or variance they will exhibit large differences in the tail. Fine, suppose we find such a thing. Then what do we do? Do we say we don't need to do anything because the reason that the physics faculty have so few women is biological. What about the promising female student. Does she now have to carry the burden that she may be inherently inferior? I think we need to wait at least one more generation and then see what the ratios are. If, after we have seriously tried to equal the playing field and then there are still discrepencies then we can do the research to examine it more carefully. To me it's like doing a test for a genetic defect that can't be passed on and one that we can't fix. What is the ethical thing to do? I personally feel that we should not do the test. So call me a PC soldier if you like but I don't think a university president should bring up such a topic unless he can guarantee that the results will not alter policy anywhere in the world to actively promote gender equality.

Steve Hsu said...

I didn't mean to suggest that success in science could be reduced to a single exam score - of course factors like work ethic, creativity, dedication, etc. are not easily captured by an exam. But on the other hand we wouldn't use such exams if we didn't find them predictive in some way.

It is clear to me that women face a lot of hurdles at all stages in their careers that men don't have to. Certainly, it is in our interest to fix this problem, as humanity benefits by utilizing everyone's talents to the fullest.

But I think you have to admit that if the data on these M/F differences had been collected in some other species, say in dogs or monkeys, we would be much more willing to accept the conclusions than we have been in the case of humans, precisely because it is such a touchy subject. I've seen ideas accepted as reasonable "working hypothesis" based on much, much weaker data. (Naked mole rats, anyone?) The book by Simon Baron-Cohen that I reviewed has a lot to say on this subject.

Carson C. Chow said...

If Summers had said - we seem to see gender differences in test scores as early as 12 years of age in males and females and a small discrepency here can lead to much larger differences later, then I would have no problem with the statement. However, what he said was - hmm, even though there is discriminatory bias, there could also be an additional factor because in 12 year olds... The message is entirely different. In the first, the message is - perhaps if we intervene at an earlier age the discrepency could be alleviated, the second is more of - it's not my fault there are no woman at Harvard. Even if he did not mean the latter, as a university president it is his responsibility to not be misinterpreted on such a touchy topic. Look what the blogosphere did to Eason Jordan. Was that a fate he deserved? I think the two situations have parallels.

Anonymous said...

Summers is a hypocrite.

Here's what he said about those who propose that colleges divest from their investments in Israel: that they're "anti-Semitic in effect, if not intent."

If he wants to fling slurs like that around, he shouldn't wince when people call him a bigot.

Blog Archive

Labels