<p>Evidently, a racist admissions policy is worth paying for. Diversity is good, but you don't realize that things come with costs. If creating diversity means judging people not by their intellects and characters but by their colors, even at the margin, then we'll keep our academic integrity.</p>
<p>I think the most fundamental way of showing respect for somebody is to treat her as an individual, deserving of consideration on her merits, not (even partially) based on her skin color or ability to have babies. Your exalted kind of diversity is crude and dumb. You can have it.</p>
<p>There were only 244 blacks in the entire US who scored above 750 math as of 2005, out of a total of 33,841 for all groups. Our primary and secondary school systems are failing to educate blacks but somehow the universities are expected to take up the slack.</p>
<p>There's no doubt that the SC permitted "holistic" affirmative action to continue in its Michigan ruling but they did not (and never have) required it. There's no reason why a school has to have the maximum amount of AA that the law allows or any at all. It hangs by the narrowest of legal threads - if it came up today with O'connor gone it wouldn't pass - even she said it had to go within 25 years. The voters in several states have shown that they don't like it. Other than the beneficiaries, the groups that seem to like it the most are the existing elites - it assuages their guilt, it gives legacy preference a leg to stand on, it allows us to ignore the failure of the public schools. It's AA that's "conservative" in the sense of keeping the system the same as it is. </p>
<p>I find the "experiencing diversity" argument particularly condescending. URM's are not some kind of zoo animals brought in for the entertainment of your children. I think AA has the opposite effect - when you see a URM sitting next to you in class, you always have it in the back of your mind that they may not be as qualified as the other students. At CalTech you have no such doubts and can treat all students with the respect that they deserve.</p>
<p>What is meant by "diversity", anyway? I can understand wanting your children to experience cultural diversity, but what does that have to do with MIT's AA policy? Most of the ethnic minorities admitted by MIT have lived in the USA their entire lives, and bring no more cultural diversity than Joe Smith the white American male. If you want to bring cultural diversity to MIT, what needs to be done is admit more international students (and therefore decrease the number of American students); I doubt you'd be willing to do that.</p>
<p>Or are you only talking about ethnic diversity? Is it that you want your children to see lots of different skin colors in college? What's the point of that? And as Percy says, it makes URMs into some kind of zoo animals.</p>
As I've stated before in discussions like this, I'm willing to accept this situation, simply because I think the methods that it's possible to use to distinguish between "qualified" and "maximally qualified" are artificial.</p>
<p>Yes, MIT might take one student with a 2200 on the SAT and pass on one with a 2400. In that solely numerical sense, then, MIT is passing on the opportunity to collect the freshman class with the highest possible average SAT score. I just don't happen to believe that small differences in SAT scores are decisive in the future that class is able to achieve.</p>
<p>"I just don't happen to believe that small differences in SAT scores are decisive in the future that class is able to achieve."</p>
<p>Wouldn't you say they are more predictive than say differences in skintone? How much is "small"? Would a 2000 be too low in your view? For a Hispanic student? For a Korean? Why are lower scores OK for some groups but not for others?</p>
<p>If it is of any help, in November 2005, Boston magazine revealed how Harvard admissions worked:
"With so many of its would-be freshmen performing so well on the traditional quantitative measures of grades and standardized tests, those numbers become almost meaningless. So Harvard assembles the jigsaw puzzles that are its incoming classes based on a much more fluid definition of merit."</p>
<p>Mollie, can you see a way to take this discussion from conjectures to data? How would you measure "The future that class is able to achieve"? I have been advocating average grades, a very imperfect meaure of performance, but one that I think is better than percentage who graduate. Do you have any idea of how average grades at MIT relate to SAT scores, holding fixed high school grades? How much of a difference does 2200-vs-2400 make?</p>
<p>I don't know about MIT but I have read studies about college students in general and it's been shown that high school GPAs are a much better predictor of college grades than SAT scores. For whatever it is worth.</p>
<p>But I would like to add, that, COMON, we're tech school students, we argue about numbers all day long, must we obsess over it in our free time, too? Life's best moments don't fall into data sets! </p>
<p>
[quote]
Wouldn't you say they are more predictive than say differences in skintone?
[/quote]
</p>
<p>Yes, they're more predictive than skin-tone. That's why they're obviously more heavily weighted than skin tone. A statement like that just sounds suave. It has no substance.</p>
When I think of outcomes, I tend to think about being able to do what you want to do after graduating from MIT, rather than having to accept a backup plan. I don't know how easily quantifiable that is. It's not entirely related to grades, either (as I'm sure you're aware) -- graduating from MIT with a 4.8 doesn't necessarily get you any more opportunities than graduating with a 4.0, unless what you want to do is something very GPA-dependent, like medical school.</p>
<p>The only SAT-vs-GPA piece of info I have is this: SAT math scores above 700 (which is to say, just about every SAT math score at MIT) don't correlate with grades in a freshman's first math course -- a freshman with a 700 is just as likely to get an A/B/C as a freshman with an 800. This is without controlling for high school grades. I am not sure if it controls for differential course choice.</p>
<p>
[quote]
The only SAT-vs-GPA piece of info I have is this: SAT math scores above 700 (which is to say, just about every SAT math score at MIT) don't correlate with grades in a freshman's first math course -- a freshman with a 700 is just as likely to get an A/B/C as a freshman with an 800. This is without controlling for high school grades. I am not sure if it controls for differential course choice.
[/quote]
</p>
<p>Surely the admissions office has run much better regressions than this, right? What about the Math IIC? I can't imagine 700 being "the same" as 800 -- I've never met a strong high school mathematician who didn't have an 800 there, while a 700 raises some powerful suspicions.</p>
<p>Moreover, unless you're evaluating a crop of students for whom the "ignore all differences in SAT math above 700" policy was used, this lack of correlation --even when things like course choice are rigorously examined -- does not imply that such differences have no meaning. After all, if you have a policy where people with lower-700 scores are at a disadvantage, then anyone admitted with such scores is likely to have strong other credentials. In this situation, a lack of correlation would mean that the relative weights you put on the SAT and on other credentials are pretty good -- not that the SAT has no meaning in that range.</p>
<p>Now, insofar as the math SAT is primarily intended as an indicator of math ability, while "other credentials" can address a lot of other strengths, the optimal weighting of the math SAT in admissions should leave some correlation (although you'd have to control for course selection, which seems enormously relevant here). But I want to make sure that no one infers that a lack of correlation means a lack of value. Unless differences in >700 scores were explicitly ignored while admitting the pool that produced this data set, that conclusion just isn't true.</p>
<p>Scores on the SAT or even the Math IIC are not any strong indicators of how well a student will fare in math at MIT, where everything is calculus based. A good score on the SAT is more indicative of general logical puzzle skills like Sudoku. </p>
<p>MIT does pay a lot more attention to the AMC12 and AIME scores which are listed on the application. </p>
<p>MIT will only provide any advanced standing for scores of 5 on AP Calculus BC. While nearly 60% of incoming freshmen take the credit and skip single variable calculus, they do not perform better on average on multivariable calculus than those who did not take the credit. Calculus (and advanced math a fortiori) at MIT has simply very little to do with high school math. More than half of MIT students have 800 on the Math IIC and many of those will get scores of 30 or less on basic calculus tests.</p>
<p>
[quote]
Scores on the SAT or even the Math IIC are not any strong indicators of how well a student will fare in math at MIT, where everything is calculus based. A good score on the SAT is more indicative of general logical puzzle skills like Sudoku.
[/quote]
Come now. You have no actual evidence that they're "not strong indicators"; your argument appears to consist entirely of the silly inference that since the SAT math tests don't contain calculus, they won't be relevant to performance in a calculus-based curriculum. (If your reasoning here was valid, aptitude testing in general would be doomed!)</p>
<p>This only underscores my general point: we need hard, empirical evidence about how different factors predict student performance. Otherwise we're arguing in a vacuum, without the capacity to rigorously distinguish competing claims about what "really" matters.</p>
<p>One would hope that the admissions office at MIT (of all places!) would have gotten down to business, run a billion regressions, and hashed out an understanding of exactly how admissions factors -- including the vague, subjective evaluations that are so popular -- impact student performance at the university. Maybe it has. But the result cited by Mollie worries me. Presumably she knows about the "zero correlation above 700" because it's been popularized around the admissions department, which suggests that it's been integrated into the practices and philosophy there. It is, however, absolutely worthless without controlling for course selection, and also possibly suspect on the grounds I detailed in my earlier post. If it (and not some more sophisticated analysis) is the justification behind the "we don't care about scores in the 700+ range" claims that come so frequently from the university, count me as genuinely disturbed.</p>
<p>Hopefully Mollie (or someone else) can clear this up.</p>
<p>randomperson has hit the nail absolutely on the head. Mollie's statistic is severely contaminated because the sampling is highly nonrandom. This is like saying that leg length doesn't matter for running speed because marathon winners with all different leg lengths have been observed. !.</p>
<p><a href="from%20Boston%20Magazine%20article,%202005:">quote</a>"With so many of its would-be freshmen performing so well on the traditional quantitative measures of grades and standardized tests, those numbers become almost meaningless."
[/quote]
</p>
<p>The opposite is true. The preceding sentence in the article says that Harvard admitted over 50 percent of perfect SAT scorers that year, and the rate can only go higher if you add perfect SAT-II, several perfect APs in cumulative subjects, and an impeccable high school record. It is true that the admissions rate does not reach 100 percent just by adding more perfect statistics, i.e. you cannot guarantee acceptance by playing only a numbers game. But a very large amount of the variance can be frozen out in this way.
If there is some subjective credential operating in your favor even to a moderate extent (e.g. sports player, photographer for high school newspaper), having a large number of essentially perfect statistical metrics then puts the admissions chances well above 50 percent, even at Harvard.</p>
<p>The median Academic Index at Harvard was reported as 220 by its admissions director. The average at Dartmouth was 212. Most of the AI fluctuation at such schools comes from SAT scores, as almost everybody admitted without special preferences or athletics is in the top 5-10 percent of their class in high school.
The SAT fluctuation or the mid-50 percentiles are not so much a sign that selective schools don't care much about scores and grades once you near the top; it is at least as much a sign that near-perfect academic credentials are much rarer, and much less sneezed at, in applications to the top 10 schools than is generally believed. They cannot fill the class with perfect-SAT I and II and AP valedictorians, as there are not many of those to go around.</p>
<p>re: cellardweller -- MIT gives exemption for at most the first semester of calculus and physics based on the AP exams, so indeed they don't regard the AP as reaching high enough for their purposes.</p>
<p><a href="%5Bb%5Drandomperson%5B/b%5D%20wrote:">quote</a> After all, if you have a policy where people with lower-700 scores are at a disadvantage, then anyone admitted with such scores is likely to have strong other credentials. In this situation, a lack of correlation would mean that the relative weights you put on the SAT and on other credentials are pretty good -- not that the SAT has no meaning in that range
[/quote]
</p>
<p>That is EXACTLY correct. Well said. When people talk about the level of "correlation" between SAT and whatever, they are almost always implicitly referring to some form of (performed or unperformed) regression, and what randomperson says really nails it in that setup.</p>
<p>It seems that some of us agree that we'd need to see more sophisticated regressions to determine the impact of various admissions policies on the classroom performance of the admitees. Can we also agree that until we see such regressions, some of the rhetoric about gender bias ought to be toned down?</p>