<p>It is not our policy that there is an absolute threshold of academic qualification, and I apologize if I have misrepresented that fact. </p>
<p>Let me state it more precisely: there is a point past which any additional increases in your score accrue diminishing returns. In other words, getting to a certain point is VERY important, but getting beyond that point is less and less important as your score gets higher and higher. </p>
<p>What I’m trying to push back against is the idea that SAT scores add value to your application in and of themselves. They don’t. What an SAT/ACT score does is give us predictive data that tell us whether you would struggle or succeed at MIT. </p>
<p>Here is what Matt McGann, MIT '00 and Associate Director of Admissions, has to say. It’s a bit dated - back from the 1600 days - but still accurate: </p>
<p>I’m sure MIT’s admissions staff is more sensitive to the vagaries of applicant behavior because they appreciate the stress candidates are feeling, but I must admit that it turned me off to see postings such as SAT Subject Test x 790 (retest 800) or SAT 780/800 (retest 800/800). It gives the impression that the candidate is hung up to an extreme degree on external measures of performance to establish differentiation. That might lead those that haven’t studied admissions patterns for the last twenty years to question how the student would fit in an MIT environment that is both extremely challenging yet emphasizes cooperation. Or it could mean nothing at all…</p>
<p>Nowhere have I said that our women or URM applicants have lower scores than other applicants. I have not said this for two important reasons.</p>
<h1>1 - It isn’t true.
[/quote]
</h1>
<p>Is it true or untrue that, as one moves higher on the math-SAT-and-science-olympiads scale (that is, the SAT scale thought of as extending beyond 800 by including national and international math/physics/CS olympiad results, chess ratings, or other SAT-style objective data of the kind the Caltech admissions guy said are gender-disparate at Caltech), two things happen:</p>
<p>(a) admission rate goes up;
(b) the ratio of US females to US males (or URM to non-URM) at that level in MIT’s applicant pool goes down?</p>
<p>Please note that in discussing SAT data I am referring only to the math SAT and things that could be considered as substantially an extension of the math SAT scale but are nationally uniform, comparative, and based on objective exams. Possibly one should add other science competitions such as the chemistry & biology olympiads or Intel/Siemens talent searches, but it’s not clear how much these are cognitive selectors rather than contests of scholarship. Also, in MIT’s pool, using total (math plus verbal) SAT scores will mask the gender differences because MIT’s distribution is very compressed on the math scale (due to the 800 cutoff) but further from the top on the verbal, where women are more competitive.</p>
<p>
</p>
<p>The admission probability goes up with SAT (and further olympiad results, etc) in a nonlinear and steepening fashion. This was graphed nicely for MIT in the College Admissions Project study data from ten years ago (the “Revealed Preferences Ranking” paper downloadable from SSRN and frequently cited in these parts). If the SAT were merely a cutoff, meaning that admissions results would be identical if in each candidate’s file the SAT data were suppressed and replaced with the words “GOOD ENOUGH” if above the break-points, the curve might rise a little bit, but basically would be flat. The only way that would not be the case (under a break-point-only usage of SAT) is if the rest of the selection is looking at factors that correlate so well with the math SAT that they effectively are the math SAT, or measuring the same thing, such as physics olympiads, TopCoder results, USAMO qualification etc.</p>
<p>Although self-selection can certainly raise the female admission rate, there is no form of self-selection that could equalize the male and female SAT (math) score distributions in the applicant pool. For instance, the females would need to be especially shy about applying when their SATs are low but become overconfident (compared to men) about applying when their SATs are high, and it’s not likely that both are true. If females require a higher SAT minimum than men to be convinced to apply, this would affect the number of applicants but it would not influence the 2-to-1 ratio of men to women at the higher scores. If females set higher thresholds than men on other factors, such as grades or level of professional interest, when deciding whether to apply to MIT, these barely correlate with SAT at all once you set a high cutoff, and cannot overcome the escalating gender disparity toward the top.</p>
<p>At the levels beyond SAT 800, you would need to impose “radical self-selection” by preventing most of the male USAMO qualifiers from applying to MIT in order for some sort of statistical parity of applicants to be theoretically possible.</p>
<p>All this is of course even more true of the credential distribution of URM versus non-URM applicants, as one can infer from national data.</p>
<p>That really surprised me. I am an international EC, and I think I deal with a half-dozen grading systems in my region alone. Not just A-levels, and IB, but a wide variety of intriguing variants including one (actually quite good) school in my region that does not believe in giving grades at all. That is one of the reasons why context is so important.</p>
<p>Who cares (for purposes of this topic, anyway) what people from Caltech admissions said about their applicant pool? The Caltech applicant pool and the MIT applicant pool are not the same thing.
[/quote]
</p>
<p>I think it would be a hard slog trying to convince people that:</p>
<p>a. Caltech differs drastically from MIT in its applicant demographics;</p>
<p>b. It is a fluke that Caltech’s experience accords with the familiar national trends; and</p>
<p>c. There are self-selection factors that shape MIT’s female applicant pool (in particular, factors that might reverse the sharp escalation in gender disparity that occurs at the high end of the distribution of mathy analytical credentials) but do not apply to Caltech’s applicant pool. It is already difficult enough to formulate a self-selection process that could have that effect at MIT considered alone, without also asking that it not have an equivalent effect at Caltech. </p>
<p>The only relevant difference I can imagine is that Caltech is so notoriously difficult that the male applicants also have a high degree of self-selection, even compared to MIT applicants, and thus the weaker SAT scores don’t appear in their pool at all. At MIT, which also has business and art and music majors, perhaps some low-scoring men apply more indiscriminately than women, and drive down the male average. At a school with an SAT math average above 750, it takes three perfect scores to balance out a single applicant who thinks he can get in with a 600, so any excess of low scores would be particularly effective in lowering the male average. However, the calculations in this thread are addressing the high scores and the upper tail of the credential distribution, and there men predominate in the national pool. What if anything could alter this in MIT’s pool and not Caltech’s is rather mysterious. </p>
<p>
</p>
<p>The question has been about the applicant SAT-score distributions, not the accepted students. MIT has very little control on who the applicants are, so between the Caltech statements and the known national patterns, yes, I would say there is some evidence. Not conclusive evidence, but enough that speculating that “it might not be true at MIT” (an appeal to ignorance) or an admissions officer asserting but not quantifying the claim that “it’s not true” (an appeal to authority) are not sufficient to refute the suspicion.</p>