Ben Golub

siserune --

What I noticed is that you just don't see as many drone applicants among women. There were definitely a lot of students, mostly guys, that I saw who clearly were set up to be engineers or scientists by parents and had no passion or particular aptitude for the subjects. Girls applying to Caltech tended to have better reasons for wanting to go and more of an understanding about what a scientific education was good for. Their essays were, on average, much better.

The usual story behind this is that few girls are pushed from their childhoods to be engineers and go to MIT or Caltech, whereas lots of boys are even if they don't have much intrinsic interest. Plus, girls on average are just better at writing. So the girls are better at presenting themselves and few are applying because someone else wants them to. That is what people mean, I think, by "self selection".

siserune

How do the distributions compare on the objective credentials? It is clear that at the extreme top men outnumber the women. Does self-selection produce a more solid non-extreme top range among the women, or are men superior by those measures for a long way down from the maximum?

Also, how do the mostly-objective credentials look, such as science projects (Intel etc) rather than the purely-objective such as math or physics olympiads? I assume females are better represented among papers and lab projects, especially in biology and chemistry, than in exam competitions, but do they reach or even exceed 50 percent of the credential-holders in those categories?

electron5

SOrry if this was already posted but I couldn't keep up with you guys reading this.

What about the fact that most MIT students were deciding between schools like CalTech, Princeton, Harvard, Yale, and Stanford (not to mention Olin)? I guess all these admissions processes have become jokes haven't they then?

CAdream

There are whole books that have been published recently about some of the more absurd aspects of Ivy league admissions. Some of the worst aspects of Ivy admissions (like pandering to fame, fortune, and political connections) have been avoided at MIT. Great! There's still plenty of issues to talk about, though.

Ben Golub

siserune -- as you know, women aren't well represented among top scorers in math and science exam competitions. They do much better in the lab science projects (but to be honest I never counted and I don't know if anybody keeps statistics -- though one could count Intel and Siemens finalists).

I would say more of the extreme outliers sat both the top and the bottom are largely men (something that checks out against the standard demographic fact), while there is a solid middle range in which there is plenty of each gender. In that range, men beat women (on average) in math/science scores and women beat men (on average) in the subjective parts of the application. So a lot can be done to the gender distribution of the admitted class depending on how you weight these factors.

Aedar

Has Caltech done any follow up studies that relate the factors considered in evaluating applications to some measure of success in college, such as average grades? Given Mollie's posts about women having higher grades than men at MIT, it seems at least possible that the relationship between SAT scores and college grades varies by gender.
If so, and if (as I think appropriate) SAT scores are valued for their predictive role and not in and of themselves, there could be interesting questions of just how to define an "unbiased" admission policy- would be fairer to set the marginal admits of each gender have the same predicted "level of sucess" and different scores. or the same scores and different predicted success?

Ben Golub

Yes, that's a good idea. Caltech used to do extensive statistical analysis of admits back in the 1980's, and that has sort of stopped. I'd be a big fan of starting again.

From what Gary Lorden (statistics professor and a member of the committee back then) told me, the model was GPA as a function of SAT M, SAT V, SAT II Math, high school GPA, and perhaps a few other things. The R^2 was about .5, and the committee did pay attention to which factors were significant in deciding what to pay attention to in admissions. You're essentially suggesting adding gender*SAT cross terms, which would be an interesting exercise.

I really wish I had it in me to take a year off before grad school and do this stuff seriously for a year.

cghen

When Professor Lorden showed those statistics, I remember there were some funny correlations: like GPA was inversely proportional to verbal SAT or something along those lines, but yeah, the fit wasn't so good.

Aedar

Ben, given that Athey, Avery et al have done related statistical exercises as part of their research, could you maybe do some sort of project on this as part of your grad school program? The hard part is getting hold of data, but it seems that you'd have a good shot at getting data from Caltech?

Ben Golub

Yeah, Athey's stuff on grad school admissions is great (if bracing) reading, and the Avery et al studies are interesting, too. I think that I could certainly do this and call it research. The problem is that it is not easy to learn to do good empirical work with slippery data like these, and becoming a decent applied econometrician would compete with other things I could be doing. On the other hand, these issues are truly important and deserve serious analysis, so if I ever think of a really good angle and convince someone to give me data, it would be worthwhile.

epiphany

electron, I posted a similar message to your post 483, possibly on a parallel thread in PF or Admissions. As with here, no one has seemed to respond to the obvious "problem" of "objective credentials" of cross-admittances to schools who have certainly not given any advantage to females and are similar reaches.

Aedar

Suppose that the pool of admitted women at MIT has lower SAT scores than the men. This might look like "affirmative action" for them, but the higher GPA of female students points the other way. So don't we need to see the results of the sort of regression-with-interaction-terms that Ben described in order to decide whether women are being discriminated for or discriminated against?

sakky

Well, basically, yes. I know it's harsh, but it's the truth. Again, we're not talking about individual year-to-year anomalies, but rather a consistent trend , which I believe has been shown to be the case. Something has to account for a consistent trend. I would point my finger to maturity. While there are several possibilities to explain a consistent trend, maturity seems (at least to me) to be the most likely of them.

Some women at MIT drop out/transfer elsewhere also. Hence, it's a wash. What we are talking about is a consistent difference in graduation rates between men and women. Something has to explain it.


Because it wouldn't be politically correct to say so. After all, if it were true (that women dropped out of MIT at higher rates than men), that would probably be used as ammunition for people who want to get rid of gender-based AA, and that's why people would shy away from saying it. D'Souza once defined political correctness to be publicly espousing a position that you privately believe to be untrue.

However, I'll be politically incorrect and surmise that I suspect that URM's drop out of MIT at a higher rate than whites, and that Asians probably drop out the least. However, I am not going to take a stance on race-based AA.

mythmom

I find this thread fascinating. I have onnly a few things to contribute.

1) James Watson was explicit in his autobiography that his standardized test scores (I.Q. 115) would disqualify him for science. Initially he was discouraged at every turn. Too bad he didn't accomplish much!

2) When Richard Feyman's son tested with a 125 I.Q. the teacher was apologetic. However, his wife just laughed because that was exactly R. Feynman's I.Q. Too bad he didn't accomplish much.

3) Both Kurt Vonnegut, Thomas Pynchon and Richard Powers all trained in science as undergraduates. They all left science for literature because they felt scientific disciplines are too specialized and confining. Imagine the humanity science lost!

4) I have no doubt that all you guys on this thread would score higher than I would on a standardized math exam (back in the day I scored a measly 700 M and 790 V.) However, I did get the only 100 on the NYS Chem Regents(in the entire state) the year I took it as a sophomore. As a girl in very advanced math classes I did feel stupid (two girls to 20+ boys), but now I wonder if this was lack of confidence. In those days no girl was encouraged to go into math/science. (My great interest was in chemistry at the atomic level). Did it really serve society for me to now have a Ph.D. in literature, a discipline in which I was not continually demeaned? Even philosophy was incongenial. I love literature, and especially scientifically based literature (I have taught Darwin and Einstein in a Humanities setting) but I think I could have done something as a scientist and in MIT's current climate I probably could.

I don't think the issues you raise completely invalid; everyth5ng bears examination, but the relationship of standardized test scores to innovation is very complex.

InquilineKea

Wasn't it 124ish?

Anyways, Watson didn't really do anything after he found the double helix structure of DNA. (source, John Simmons biography of him). In fact, considering that the double helix structure would have been discovered in a couple of years without the Watson-Crick team anyways - his accomplishments to science were virtually nil. At least he was good at acting like an *******.

==
Anyways, I feel that IQ tests are overrated (considering that their only validity lies in the correlation matrix - and the correlations between IQ and various individual aptitude factors only max out at 0.5 - leaving plenty of room for variation). Achievement tests are much better, but yet there aren't any good achievement tests that can DISTINGUISH between students at the top. They would be more respected if equally capable individuals spent a good amount of effort on them - the problem is that they don't - and aren't expected to try their hardest on achievement tests like the AIME and Olympiads. Moreover, not everyone takes them. If people took AIME and Olympiad exams as seriously as they took the SATs, then achievement tests would gain much more respect (in fact, they have stronger correlations with undergrad GPA than with "aptitude" tests - although the correlations are weak notwithstanding).

Achievement tests that are at a high enough level can help distinguish between students of different aptitudes, rendering an aptitude test completely unnecessary. Moreover, achievement tests pretty much form at least half of a college's grading system anyways. Of course, there's then the concern that they're more preppable than IQ tests. But the Internet is the ultimate leveler of the playing field. And materials can be distributed over the Internet for free rather than through expensive test prep companies.