Sunday, September 14, 2014

Harvard admissions and meritocracy


Motivated by Steve Pinker's recent article The Trouble With Harvard (see my comments here), Ephblog drills down on Harvard admissions. The question is just how far Harvard deviates from Pinker's ideal of selecting the entire class based on intellectual ability. Others raised similar questions, as evidenced by, e.g., the very first comment that appeared on The New Republic's site:
JakeH 10 days ago

Great article. One quibble: Pinker says, based on "common knowledge," that only ten (or five) percent of Harvard students are selected based on academic merit, and that the rest are selected "holistically." His implication is that "holistic" consideration excludes academic merit as a major factor. But that's surely not the case. Even if Harvard only selects ten percent of its students based on academic factors alone, it seems likely that academic and test score standards are high for the remaining 90 percent. We don't have enough information on this point, because, I suppose, it's not available. (To solve that problem, I join Pinker's call for a more transparent admissions process.)
I don't know exactly how Harvard admissions works -- there are all sorts of mysteries. But let me offer the following observations.

1. Pinker claimed that only 5-10 percent of the class is admitted purely on the basis of academic merit (see more below). The 5-10 percent number was widely reported in the past, including by scholar Jerome Karabel. No one knows what Harvard is up to at the moment and it's possible that, given the high demand for elite education, they have increased their academic focus over the years.

2. IIRC, the current SAT ceiling of 1600 (M+CR) corresponds to about 1 in 1000 ability (someone please tell me if I am mistaken). So there are at least a couple thousand US kids per cohort at this ability level, and several times more who are near it ("within the noise"). A good admissions committee would look at other higher ceiling measures of ability (e.g., performance in math and science competitions) to rank order top applicants. The 800 ceiling on the math is not impressive at all -- a kid who is significantly below this level has almost no chance of mastering the Caltech required curriculum (hence even the 25th percentile math SAT score at Caltech is 770; in my day the attrition rate at Caltech was pretty high -- a lot of people "flamed out"). The reduced SAT ceiling makes it easier for Harvard to hide what it is up to.

4. My guess is that Harvard still has a category, in the past called S ("Scholar"; traditionally 5-10 percent of the class, but perhaps larger now), for the top rank-ordered candidates in academic ability alone. Most of the near-perfect scorers on the SAT will not qualify for S -- it is more impressive to have been a finalist in the Intel science competition, written some widely used/acclaimed code, made (or nearly made) the US IMO or IPhO teams, published some novel research or writing, etc. Harvard sometimes boasts about the number of perfect SAT scorers it rejects each year, so clearly one can't conclude that a 1600 on CR+M alone qualifies for the S category. Along these lines, one even reads occasional stories about Harvard rejecting IMO participants.

5. In remaining categories Harvard almost certainly uses a more holistic approach that also weights athletics, extracurriculars, etc. Some of the people who score high on this weighted measure might not have qualified in S, but nevertheless are near the ceiling in SAT score. It has been reported in the past that Harvard used a 1-5 scoring system in academics, sports, leadership, music, etc. and that to have serious consideration (outside the S category, which is for real superstars), one needed to have two or more "1" scores -- e.g., valedictorian/high SATs + state-level tennis player + ...

From the comments above, it should be clear that one can't simply use the percentage of near-perfect SAT scorers in the class to determine the size of the S category.

See here for discussion of meritocratic test-based systems in other countries. For instance, the Indian IIT, the French Ecole Normale Superieure, and the Taiwan university entrance exams, have in the past explicitly ranked the top scorers each year. (The tests are hard enough that typically no one gets near a perfect score; note things may have changed recently.) I know more than a few theoretical physicists who scored in the top 5 in their entire country on these exams. Mandlebrot writes in his autobiography about receiving the highest ENS score in France.

38 comments:

David Dudley Field said...

Steve,

We might make more progress if you quoted the key evidence in EphBlog and responded to it. Here it is:

-------------
At 75th percentile of the SAT/ACT, Caltech (800/800/790 and 35/36/35) is
indistinguishable from Harvard (790/800/790 and 35/35/35). If
we assume that Caltech admits on the basis of “academic merit” (it
does!), then it must be the case that Harvard uses, more or less, the
same criteria (at least for 25% of its class), otherwise, it would not
have the same extreme distribution of scores. (And even the
tiny advantage to Caltech is probably explained by Harvard putting more
emphasis on the high grades portion of academic merit than Caltech does,
not on the (imaginary!) extra bump that Harvard gives to excellent
sculptors who score 2390 over average sculptors who score 2400.)

...

But, before going further, can we at least agree that I have
demonstrated that, at least for 25% of the class, Harvard is at least as
focused on “academic merit” as Caltech is? In other words, Pinker’s
estimate of 5% or 10% is off by at least a factor of 2.5 to 5. Once we agree on that, we can move on to other portions of the score distribution.
-------------

Do you have a response? If the Caltech and Harvard distributions look the same at the top 25th, then isn't the most likely explanation that they use the same criteria, at least for this portion of the class?

David Dudley Field said...

"IIRC, the current SAT ceiling of 1600 (M+CR) corresponds to about 1 in 1000 ability (someone please tell me if I am mistaken)."

EphBlog discussed this explicitly and provided a link! Here it is again:

http://media.collegeboard.com/digitalServices/pdf/research/SAT-Percentile-Ranks-Composite-CR-M-W-2013.pdf



This is for all three sections, which is too bad since the writing is much less important (in admissions and in evaluating talent). But, for our purposes, it highlights, I think, your tendency to overestimate how many smart people there are out there. Out of 1.6 million test takers, the 1/1000 ability threshold would be at number 1,600 in rank order scores. As the pdf shows, this is at a score of 2360, which is below the 25th percentile at Harvard. (Again, I am making all sorts of simplifications. See the EphBlog post for details.)


Of course, the SAT has trouble distinguishing at the upper levels, A 1600 Math/Verbal is not the special score it was back in the day. But, again, there is no evidence that Harvard admissions standards are any different from Caltech's, at least for the top 25% of the class. Once we agree on that, we can move on to other portions of the distribution.

Shawn said...

While I can see why certain "soft" professions such as consulting would want a Harvard degree because it adds prestige to the firm, it seems to me that many orgs that hire science, tech, and quant types are missing out by not doing more to hire the super smart directly out of high school where the students would have been more willing to accept less money (i.e. the ones who would have otherwise enrolled in Caltech, etc.).

BobSykes said...

If Harvard is like every other school, then the first cull is done by the admissions office's staff of clerks and work-study students. They are given some simple criteria like ACT/SAT scores, affirmative action category, legacy, athletic status and maybe geography. The affirmative action candidates, athletes, children of alumni and favored geographic origin then enter parallel review channels dedicated to them, and the remainder passing the cull will go on to the admissions committee.


Virtually everyone whose application gets to the admissions committee will be able to succeed at that school, and probably many others. At that point, the admissions committee applies whatever criteria the school uses to promote its self-declared mission. At a place like Harvard, they want to prepare the next generation's Ruling Class, and various personality traits become more important than purely academic ability (which all the candidates have anyway). So the fact that that committee does not use purely academic criteria is irrelevant.


The fact that the biggest cull is done by the staff, and not the committee, opens the door to all sorts of abuses. Remember the work-study student at Maryland (?) who bragged on her social media that she trashed all applications from white males in order to fight against the patriarchy and white privilege?


Then there is also the obvious discrimination against East Asians. Harvard's undergraduate student body is said to be 25% Jewish. It would also be 25% East Asian if Harvard did not discriminate.

Butch said...

"For instance, the Indian IIT admission test, and the Taiwan university entrance exam, have in the past explicitly ranked the top scorers each year"
I think it should be stated that India has a caste system. Those at the top(also known as "Brahmins" where bred to be scholars and priests... They are probably why india has a probe orbiting mars and nuclear weapons, the lower castes(and their birth rate) are probably why it is a mostly third world country as of today. Any thoughts on this.
In the U.S, the highest average income per household title is on the Indian Americans, 1% of the population yet 16% of the doctors with higher rates of masters degrees than other Asian American?
What do you think?

steve hsu said...

Most psychometricians (and even many admissions folks) discount the writing portion (score is too noisy, not g loaded) and only look at M+CR.


I'm making the point that you can't just look at SAT scores to see how big the S category is. There are other, better, indicators.

steve hsu said...

Both 75th percentile distributions are "saturated" at the ceiling. You need to look at other indicators.


I gave you an actual algorithm, used by H in the past (and perhaps still today), that could explain the same observation about 75th percentile of scores, without making the S category 1/4 of the entering class.

Al_Li said...

http://www.openthemagazine.com/article/nation/where-are-they-now

Richard Seiter said...

Steve, you've generally been careful to point out you are referring to the "S category" you defined earlier (pure scholastic admits). I think a big reason for the controversy is that Pinker was not so careful and just wrote: "it’s common knowledge that Harvard selects at most 10 percent (some say 5 percent) of its students on the basis of academic merit." IMO inserting the word solely in his quote would have been more correct (changing the meaning significantly, interesting to notice that your paraphrase of Pinker included "sole"). Arguing that academic merit is irrelevant for the rest of the Harvard admits is silly IMO. This is not to say there are not groups for whom these standards are much relaxed (e.g. legacies and athletes), but even those admits are probably a good ways to the right on the bell curve for "academic merit."

I thought the more interesting part of Pinker's essay was the commentary about intellectual engagement on campus. There is a big gray area between not going to class and being a grind, and another between only caring about grades and working hard to learn things you care about (a subset of classes). I wonder how common Pinker's experiences are. Could they be related to the classes he is teaching and/or his "fame" selecting for unengaged students? (seems odd to me since his writing is so interesting and engaging)

I like David's reformulation: "at least for 25% of the class, Harvard is at least as focused on “academic merit” as Caltech is" and think that attempting to answer that question is worthwhile. I disagree with "isn't the most likely explanation that they use the same criteria, at least for this portion of the class?" (same criteria is an overstatement, perhaps similar with an additional qualifier wrt overall academic merit as measured by SAT)

As you observe a relevant comparison would be H's SAT percentiles from past years. Does anyone have that data? How about for Caltech? The low SAT ceiling post-recentering makes meaningful discussions about this issue difficult (sometimes I wonder if that was the point of recentering).

One thing which I think is extremely relevant but largely left out of this conversation is an assessment of what proportion of applicants with exceptional academic merit (relative to the overall admitted class) are not accepted. I would argue systematic bias in that group (e.g. against Asians) is a problem when "selecting for a ruling class." This is especially relevant for H given your observation that they likely have first call on the best (any idea how the absolute number of category S applicants/admits has changed over the years?).

I will add that I think the title of the EphBlog post "Extracurriculars Don’t Matter in Elite Admissions" is an incredible overstatement. Arguments/discussions like this would be easier if folks weren't consumed by the media tendency towards extreme statements/headlines. The typology in http://infoproc.blogspot.com/2009/11/defining-merit.html seems plausible to me and shows the importance of extracurriculars in several categories. Question for David, do you find that typology plausible?

Steve, do you have any idea of what percentage of admits circa 1988 were in category D? That seems relevant to this discussion.

Regarding category S, Scott Aaronson makes some interesting points (e.g. his "professor test") about applicants who appear to be in that category: http://www.scottaaronson.com/blog/?p=2003#comment-138632

steve hsu said...

If you parse Pinker carefully you will see he did not make a mistake. I think he wrote carefully that BEYOND the 5-10% (S category) Harvard STARTS to assign nonzero weight to other factors like athletics or musical ability. He is definitely talking about S, as am I.


"... it’s common knowledge that Harvard selects at most 10 percent (some say 5 percent) of its students on the basis of academic merit. ... The rest are selected “holistically,” based ALSO on participation in athletics, the arts, charity, activism, travel, and, we inferred (Not in front of the children!), race, donations, and legacy status (since anything can be hidden behind the holistic fig leaf)." (emphasis mine)


What David doesn't seem to understand is that starting to assign nonzero weights to other factors could still lead to a top 25% of the class (400 kids) at or near the SAT ceiling. There are a lot of smart kids who 1. hit the SAT ceiling, 2. but are not good enough to qualify for an S category with, e.g., 10% = 160 slots, 3. but who also are good at sports/music/leadership and hence get the nod over a more impressive academic kid who lacks the other factors.


Given its current cachet (probably wins cross-admit battles with every other school at > 70% probability, except possibly Stanford?), Harvard could definitely take more smart kids than it currently does IF IT WANTED TO. That is NOT to say that the current top 25% or even 50% of each Harvard class isn't an impressive group -- they are!

Anonymous said...

Two American IMO gold medalists were declined by Harvard: Sergei Bernstein and Shaunak Kishore. Caltech absolutely would have accepted them.

Aside: the IIT exam has been watered down in terms of exam complexity, but it still functions pretty well as a rank-ordering tool. Just in CS, the last two ACM Doctoral Dissertation award winners did their undergrad studies there, as did this year's Nevanlinna prize winner.

Richard Seiter said...

Thanks for the response and clarification/elaboration. Sorry if I gave the impression I thought Pinker made a mistake (I do think the statement I quoted was insufficiently clear and that contributed to the debates now being had). When I use verbiage like "more correct" I usually mean something like "may have oversimplified a bit" or "is avoiding being explicit" (e.g. for PC reasons), but short of an actual error. Pinker takes great care in his writing and I suspect if it was vague it may have been intentional.

I think people underestimate what S implies (First-rate scholar in Harvard departmental terms) and this contributes to misunderstanding Pinker. The new SAT is not close to resolving that and I think even the pre-recentering SATM may have been ceiling limited in that regard (I have less experience interacting with 750+ SATV from that vintage to judge it).

I'm a bit confused by what seems to be a discrepancy between Pinker's complaints which seem more focused on academic effort/attention than aptitude and the proposed solution which is basing admissions mostly on aptitude. Based on his complaints it seems to me he would be more interested in shifting the academic slacker-grind needle a bit. Perhaps the connection is both imply a lessening of focus on extracurriculars?

Have you had any experiences along the line Aaronson described as his "professor test" for an admissions regime? (Namely: if the professors in some particular department, on being shown an application, would say, “holy crap, admit this student!!,” then the student would reliably be admitted.) This seems like it should correspond to the S category to a fair degree.

It would be fascinating to see numbers for cross-admit battle results sliced by an assortment of demographics (e.g. geography). Are any data like this (publicly) available?

MUltan said...

Back before the '90s reentering many top schools weighted the V twice as much as the M because it was more g-loaded (or perhaps predictive of GPA) and had a higher ceiling than the M. (e.g. 760V was about 1 in 2500 for 1988) By having two "verbal" sections ETS tried to change the SAT to look like it was pre-weighted in the same way, but by removing analogies, having a lower ceiling, and less reliability in the top of the range they sabotaged the test's usefulness for elite admissions.

ETS is not likely to go back to making a test that is good for this purpose, nor are the top schools likely to accept an alternative test that would by necessity seem to be more discriminatory than the current SAT. Perhaps they might be convinced to use current tests with more "top" such as the Miller Analogies Test or the Math, Physics, or Computer Science GRE subject tests. The latter three rely on information that few HS students will have been exposed to, though low scores wouldn't necessarily be meaningful. If they weren't officially required but were given proper weight if submitted, these tests would become quite popular among the best applicants and would lead to having a large pool of students that would come in knowing most of the undergraduate curriculum, which would require a great deal of adaptation on the part of the universities in either offering more advanced classes or granting credit by examination. (Personally, I think that the whole undergraduate degree and credit system should be replaced with computerized adaptive tests - what we have now is not an accurate knowledge, ability or capability measurement system. OTOH the current sytsem suits its real and less mentionable purposes quite well.)

David Dudley Field said...

I agree with your claim about the writing. But the only good data I can find about the raw numbers in the extreme tail includes the writing, so I used it. If you have better data, by all means share it. But I doubt that it would make much difference to the analysis.

With regard to the S category, what, precisely are these "other, better, indicators" that you mention. You might discuss International Math Olympiad and similar (inter)national competitions, but the number of students who participate in this is trivial compared to the population we are examining. And, moreover, these things only exist (counter-examples welcome) in the math/sciences. Assume that Harvard (and Caltech, for that matter) would like to enroll the ten students most likely to become top Ivy League history (or English or sociology or . . .) professors in 20 years, how would you recommend that they select them. Surely, there would qualify as "S Category."

And that is the rub! There is no good data for figuring out who these students are beyond the data that I have already discussed: standardized test scores, high school grades and teacher recommendations. That is all we have.

And, by that data, the population in the top 25% of Caltech and Harvard is indistinguishable.

David Dudley Field said...

To extend my comments above:

> Both 75th percentile distributions are "saturated" within noise at the ceiling. (This ceiling is *below* the 10% S threshold.) You need to look at other indicators.


There are no other indicators, at least in areas outside math and some of the other sciences. You continue to imply that there is some way that Caltech can tell that student X is in the top of the S distribution while student Y is not, even when student X and Y have the same test scores (and grades and teacher recommendations). What is the magic pixy dust that Caltech admissions can see and Harvard (and we) can not? Please be precise.


I agree that the we are saturated within noise at the ceiling. But, if this is the case, then --- unless you/Caltech have some special way of determining that this student with perfect SAT scores, super high grades and amazing teacher recommendations is not really at the tail of the S distribution but this other student, with identical scores/grades/recommendations, is at the tail. How can you tell?


I claim that you (and Caltech) can't tell. That is, at the tail end of the grades/scores/recommendations distribution, we are saturated within noise at the ceiling (not only do our standardized tests not distinguish in the tail, but grades/recommendations are hard to compare across high schools). In that case, it does not matter what mechanism Harvard or Caltech use to select from that pool. You imply that Harvard, in picking within that pool, favors kids with strong extra-curriculars like student government president. But so what? Harvard could just as easily pick randomly from that pool and still end up with the same proportion of extreme S kids as Caltech because there is no way to dinstinguish among them.

Richard Seiter said...

Having applied to Caltech back in the day I would say the interview with a professor (as opposed to the more typical alum interviews, like my sister does for Yale) could be pretty useful "pixie dust" in detecting top tier ability. My understanding is they don't do that now though?


How about 5's in the harder AP tests as another metric you omit? It's not so much about which metrics you can see (though obviously that constrains the problem). It's about how admissions decides to use them.

David Dudley Field said...

> How about 5's in the harder AP tests as another metric you omit?

Nope. The ceiling problem is even worse in APs. For example, half the students who take it get a 5 in the Calculus AP.

http://en.wikipedia.org/wiki/AP_Calculus#Grade_distributions_for_AP_Calculus_BC

That is tens of thousands of students each year.

By whatever measure you want to look at, the top 25% of the Harvard class is indistinguishable from the top 25% of Caltech.

David Dudley Field said...

> There are a lot of smart kids who 1. (nearly) hit the SAT ceiling, 2.
but are not good enough to qualify for an S category with, e.g., 10% =
160 slots, 3. but who also are good at sports/music/leadership and hence
get the nod over a more impressive academic kid who lacks the other
factors.

No. There are not. This whole "S Category" is a figment of your imagination, and dated propaganda from Harvard. (Harvard wants you to think that playing the violin matters so they can conceal how much being black or playing hockey matters.)

Let's say I line up the 500 kids who are in the top 25% of Harvard (400) and Caltech (100) by SAT scores. On what metric (presumably something that gets to the essences of Sness) do the Caltech 100 look better than the Harvard 400?

Richard Seiter said...

Calculus looks like a bit of a cherry picked example (Biology, Chemistry, and Physics C all have a lower % of 5's), but your basic point about the AP score ceilings stands. Thanks for the link.

Let's look at this differently. If Harvard does have something akin to category S (5-10% of class, presumably the top 5-10%) then it seems clear they believe they can tell the difference between that and the top 25%. Do you dispute that?

I think using words like "indistinguishable" undermines your argument. For example, how does the proportion of Asians compare in those two groups? Do you believe the top 25% at Harvard (across the board) has the same math skills as the top 25% at Caltech? Remember there is self selection for who applies to what schools as well as schools selecting for certain characteristics.

Cornelius said...

What fraction of Harvard's applicants are East Asian? That's the fraction of its student body that would be East Asian if it did not discriminate.

Of course, all universities with more applicants than seats in its freshman class must discriminate. The question is, what criteria does it use to discriminate?

steve hsu said...

I thought I answered this in the main post. There are lots of indicators with higher ceiling than the current SAT.

1. Max out SAT at earlier age (like SMPY)
2. Perform well on really hard competition exams like for USAMO or IPhO qualification
3. Science, coding, number theory, etc. projects
4. Explain how Einstein deduced the Lorentz transformation to a Caltech prof interviewing you

Do you know how Intel/Westinghouse pick their winner from ~40 finalists? It's not the project. It's how smart the kid appears to be in interviews.

Just imagine your goal is to pick out the +4 SD kids from a bunch of > +3 SD kids. There are signals that are hard to fake ... Admittedly, ALL the signals we are talking about are a bit noisy. But I'm not sure why you can't believe there are people out there who care about differences in the tail. In my field, choosing graduate students is all about that difference.

How do you think a college coach can tell the difference between a Division I player and an All-Conference HS player who is only Division II or III caliber? They all seem like great athletes to me ...

steve hsu said...

http://en.wikipedia.org/wiki/United_States_of_America_Mathematical_Olympiad

"Qualifying for the USAMO is considered one of the most prestigious awards for high school students in the United States, with only 264 students qualifying in 2013 out of over 350,000 students competing.[1] Top scorers on the USAMO are invited to the Mathematical Olympiad Summer Program and represent the United States at the International Mathematical Olympiad." (Keep in mind even the 350k are a selected population.)

Nope, can't imagine how you'd find a few hundred off-scale kids each year without using the old SAT. Nearly impossible ...

Richard Seiter said...

About the SAT... First, you do realize how mistaken it is to add together the 75th (25th is a typo) percentile individual scores to get an estimate of the 75th percentile aggregate score, right?

http://research.collegeboard.org/content/sat-data-tables has additional data for the SAT.

Of particular note for Steve's point is the table for M+CR at http://media.collegeboard.com/digitalServices/pdf/research/SAT-Percentile-Ranks-Composite-CR-M-2013.pdf which confirms Steve's estimate of 1600 being about 1 in 1000 (actually 1841/1.66M). Does anyone have composite score data for the colleges?

Also see the tables for the individual scores which give ceilings (# scoring 800) of:
CR: 11,138 / 1.66M (about 7/1000)
M; 13,326 / 1.66M (8/1000)
W: 6,835/1.66M (4/1000)

So the writing portion of the SAT has the highest ceiling! This makes the 75th percentile of 790 (9566 / 1.66M) for both Harvard and Caltech especially impressive. It also calls into question the assertion "writing is much less important (in admissions and in evaluating talent)." If writing is less important they are certainly doing a good job of selecting for it by accident! The interesting questions are: "what exactly is the writing section measuring?" and "why are the writing scores so high at H and C?"

On another note, reading the Ephblog college links (thanks!) helps give an idea of what the institutions claim is important. In particular, see

C7 Relative importance of each of the following academic and nonacademic factors in first-time, first-year, degree-seeking (freshman) admission decisions.



Interesting that Harvard only specifies not/considered while Caltech divides into importance buckets.


C9 gives the percentages in 100 point SAT buckets. Interesting to notice the number <600 at H versus C.

Richard Seiter said...

David, you might want to notice that no one here is arguing with you about your importance of race/wealth/athletics assertions (and I agree with you about Harvard being careful to deemphasize that). But I don't think those issues affect the top 25% of the class distribution very much (some though, from the effect Steve noted).


I agree with Steve's point that you quoted (and then attempted to rebut). It matches my personal experience at MIT and what I have seen with friends and relatives at Ivies. If you want to refute that I think you will need a better argument than you have provided so far (the last argument of denial and ridicule was particularly unimpressive).

David Dudley Field said...

> There are lots of indicators with higher ceiling than the current SAT.

Let's go through them one-by-one. But note that this is a side issue. The question we are addressing is not: Are there ways that Harvard/Caltech could, if they wanted to devote the time and resources, to find more students on the tail of the IQ distribution? There are and they could. The question is: Given the data that Harvard/Caltech have access to, is there any evidence that, for the top 25% of the class, Harvard is any less focused on "academic merit" --- and keep in mind that reasonable people think that this is more than just IQ, or that reasonable people can disagree on the definittion --- than Caltech is?

> Max out SAT at earlier age (like SMPY)

There were 34,000 Harvard applicants last year. How many took the SAT (verifiably) at an earlier age? Very, very few. In fact, I bet the number is fewer than 100. Data that only apply to 100 applicants are useless for the debate over whether Harvard/Caltech differ in the 25% of each class.

> Science, coding, number theory, etc. projects

There were 34,000 applicants to Harvard. How many of them even had the chance to do advanced "number theory" projects? Moreover, even if there were hundreds, which I doubt, how could Harvard admissions --- given its current staffing and budget --- possibly evaluate which of them were any good? Answer: They couldn't. And, even if we hired a thousand Steve Hsu's to evaluate these projects, how could you possibly know which ones represent 98% the students own work and 2% good guidance from his high school math teacher and which ones represent the reverse? (And just how many high school math teachers are even qualified to guide a high school students independent work in "number theory?") The whole scenario is absurd. (And, if this ever started to matter, you can bet that the Milton's and Andover's of the world would game it viciously.)

And it is even more absurd when you remember that college decisions are often made in the fall of senior year. What percentage of this work could be done in time to matter for admissions?

> Explain how Einstein deduced the Lorentz transformation to a Caltech prof interviewing you

That is going to scale really well! Again, in theory, there are ways that Harvard (and Caltech for that latter) *might* be able to provide a much finer evaluation of their applicants. But, unless I am mistaken, even Caltech can't do that today. Or do Caltech professor interview all 5,000+ of the applicants each year?

> Perform well on really hard competition exams like for USAMO or IPhO qualification

Agreed. And Harvard cares about this! Doing well on these tests (which, again, are a very small number of people in only one of the academic areas that Harvard cares about) counts in Harvard admissions. It is part of what Harvard considers to be "academic merit." Which is one reason why Harvard has (I think) the best record in the Putnam of any school in the country.

http://www.math.harvard.edu/putnam/


Now, if it was the case that Harvard had, substantially, fewer USMAO geniuses then Caltech, you might have a point. But they don't (otherwise they wouldn't crush it in the Putnam) and, to the extent that they do, it is probably due to Harvard putting more weight on the humanities portion of "academic merit", relative to the math portion, then Caltech does.

David Dudley Field said...

Again, this is not the debate. If your definition of "acdemic merit" is "top finisher in USAMO," then, obviously, it is easy to find 264 names. But Harvard also wants high grades in all your non-math classes, 5's in non-math APs and so on. At least some of those 264 don't have those things, and so Harvard rejects them. (And, of course, Caltech rejects scores of them as well.)

The debate is: On what measures can you distinguish the 400 top kids at Harvard from the 100 top kids at Caltech? You imply that, if they all took the SAT at 13, the Caltech kids would have done better. Maybe! But they didn't, so there is no way to know.

David Dudley Field said...

> But I'm not sure why you can't believe there are people out there who care about differences in the tail.

I agree with this! There are a lot of people who "care about differences in the tail." And a lot of them work in Harvard admissions! My claim is that the top 400 students at Harvard are drawn for the extreme, extreme right tail of "academic merit." They max out on some weighted combination of high grades in the hardest classes, top scores on standardized tests and teacher recommendations. Now, you might prefer a different measure of academic merit, one that weighed tests more and grades less, or whatever. But Harvard's definition is at least reasonable, and it does not include any nonsense extracurriculars, like Pinker's activism and travel.

Now, it is probably fair to say that the top 400 at Harvard and 100 at Caltech are different on some dimensions. After all, choosing Caltech over Harvard or vice versa is a fairly big personality differentiator alone. But, in terms of measures that you or I or Harvard or Caltech admissions can easily see --- without interviewing hundreds or thousands of kids --- the top of the Harvard and Caltech distributions are indistinguishable.

David Dudley Field said...

Let me clarify. I am happy to grant that it is possible that, if Steve Hsu, interviewed these 500 kids, he could determine a rough rank ordering as to their raw intellectual power. Actually, I doubt that, but it is orthogonal to argument. But Steve is not in charge of Caltech admissions. The people in charge of Caltech admissions are, more or less, the same sort of people in charge of Harvard admissions, with the same limitations. They can't interview thousands of applicants and, even if they could, they would have a had time distinguishing among the +3 and +4 sd kids.

So, instead, both sets of people use the same algorithm.

Highest test scores possible.

Hardest classes in your high school.
Highest grades in high school (adjusted for strength of school --- top 5% at Andover is still plenty smart.)
Best teacher recommendations.


They may weight some of these factors differently, but these are the measures that are used because they are the only measures available to the admissions department. (Caltech uses this algorithm for the whole class while Harvard does not. But, in the top 25% (at least) the algorithm is the same.



If it were true that Harvard also put significant (and more than just non-zero!) weight on extra-curriculars like activism and travel, then Harvard students would do less well on the above than Caltech students. Since they do as well, Harvard doesn't.

Pincher Martin said...

David Dudley Field is making some good arguments. I don't know if he's correct, but his arguments are solid.

Pinker would have us believe that it's common knowledge that only 100 to 200 students out of the 2,000 applicants who are admitted to Harvard's freshmen class every year are at the extreme right side of the bell curve. But if Harvard is selecting so heavily against the best applicants with top-flight academic talent in favor of other slighty-less impressive applicants, then it should be apparent when comparing Harvard to other top universities by some statistical measures.

Is Caltech or Stanford or Berkeley or Princeton, for example, getting a larger share of that top-flight academic talent which Harvard rejects? That doesn't appear to be the case. But perhaps all the top U.S. colleges are biased against the top students.

Steve claims that this bias is happening out of sight because the broadest statistical measures (SAT scores, GPA, etc.) we have don't capture it. But DDF is right to ask, how does Steve know for sure?



Also, as far as I can tell, STEM majors make up about one-third of the intended fields of concentration for Harvard applicants, with physical sciences (7.4%), engineering (13.3%), compeer science (4.5%), and math (6.7%) making up most of them. Social sciences (23.7%), biological sciences (22.3%), and the humanities (13.7%) make up the bulk of the rest.

Richard Seiter said...

DDF is making some good arguments accompanied by some overly strong assertions (which appear to be toned down in his later posts) and the occasional error (e.g. adding the individual SAT 75th percentile thresholds to come up with the composite 75th percentile).

I think the major disagreements here are over the definition of "extreme right side of the bell curve" and academic merit as a sole criterion versus as A criterion. One thing to remember is that the admissions differences do show up in the 25th percentile numbers (and below, look up the relative number of below 600 SAT scores at Harvard and Caltech). If there is a bias I think it is against top students who don't have the breadth (extracurriculars etc.) most of the top US colleges want to see and/or fit in an "unfavorable" admissions category.

The one thing this thread has convinced me of is how much difference "recentering" the SAT made for discriminating at the far right in admissions (I already knew that, but hadn't thought hard about how much more difficult it makes things).

Thanks for the Harvard major percentages. I hadn't realized how large the social and biological sciences components are.

ShoeSticktoLast said...

My fingers are tire just looking at all the typing.

Me dad graduated in '66 from the John Harvard Junior Madrasa.

Me too dumb to go Harvard me guess. Only BGI study good enough. http://tinypic.com/view.php?pic=28s67ih&s=8

'Mer'ca is suppurating cunt with yeast infection from Ron Jeremy and James Deen (NOT Dean), that is, from Jewish porno stars.

Steve is gay Chinaman. He no understand when he mounted by Jewish cock.

ShoeSticktoLast said...

Me laugh long time at professional blog commenter Pink Martini.

He think argument joke.

Like 'mer'can comedian Steve Matin say, "Enough with comedy. Let's tell jokes."

He funny man.

Is dere Frynn Effek for penis si'?

ShoeSticktoLast said...

As Herr Professor Dokto Shoe would say:

http://www.hark.com/clips/vlpggcyywq-phd

Pincher Martin said...

Sticky Shoes,

"Is dere Frynn Effek for penis si'?"

If I were you, that's what I'd tell all the old girls I date. When you next drop your trousers in flagrante delicto, quickly cover up with that small sheet of paper you like to flash around showing your GRE scores and maybe it will distract them. Changing the topic can't hurt when a sensitive subject arises.

ShoeSticktoLast said...

Pinkman make me laugh. Ha, ha,...!


What it li'? Hmmm! Me thing for while.


Frashing bona fide' on Shoe brog li' frash penis on porn audition.


Steve ta' li' he John Holmes.


Then I whip ou' an ' he is jus' Chinaman ha ha.

David Dudley Field said...

> the occasional error (e.g. adding the individual SAT 75th percentile thresholds to come up with the composite 75th percentile)

This is a simplification, not an "error." I have mentioned several times that adding up the 75th percentiles to come up with a composite 75th percentile is wrong. But that is all the data that we have, so that is what I use. And, perhaps more importantly, this simplification strengthens my case, I think. (Haven't thought enough about this to be sure.) Recall that, while there is a strong positive correlation among SAT sections, the correlation is less than 1. So, if the 75th percentile at Harvard is at the 99th percentile nationally on each of three separate tests, then the 75th percentile of the Harvard composite is probably even further out in the tails. Hmmm. Or is that backward?

Richard Seiter said...

I provide a link to SAT correlation data in another post if you are interested. (the V/M correlation was < .5, not sure how strong I would call that)

I don't know if your conclusion is correct (the low ceiling makes this hard to reason about and I haven't been motivated to do a simulation yet), but I do know your simplification introduces a consistent error/bias (overstating the composite score at the 75th percentile, which I believe weakens your case) because the same students aren't necessarily the best in all of the tests. I'm not sure how to estimate the magnitude of that error/bias. Call it a simplification if you like (I would, it's the best we have), but at least acknowledge the bias it introduces and be careful about using the estimate to draw conclusions (especially strongly stated conclusions ;-). It might be worth tracking down some other data that has both the components and composites to estimate the error magnitude. Better yet, we just need to find a source for the composite percentiles.

AG said...

"new SAT is not close to resolving that "
Only a system of tests with ever increasing difficulty resembling traditional Chinese empiral exam might solve the problem.
Starting with basic test with low ceiling, only the students passed can apply to next test with higher ceiling. Look at US medical system as example, SAT to MCAT to USMLE, the end products are quite selected.

Blog Archive

Labels