Do Elite Colleges Discriminate Against Asian Students?

Applying to College
721

<p>siserune - Yes, in estimating the number of whites you subtracted all URMs and internationals (and I assume Asians as well). However, you then proceed to estimate the number of recruited athletes based on conventional estimates that include ALL recruited athletes–and some of them must be URMs.</p>

<p>
[QUOTE=siserune]

At a typical top 10 school, American whites are around 50-55 percent with URM, Asians and internationals making up the rest. “Hooked” matriculants generally exclude internationals, are about 40 percent of the student body, and include some 15-20 percent (generously) URM. This leaves 20-25 percent of the enrollment as non-URM hooked applicants (athlete, legacy, donor, etc). These are overwhelmingly white. How much of the Asian enrollment do you think is athlete/donor/legacy?

[/quote]

You say that 20-25% of enrolled students are non-URM hooked applicants. However, there are URM athletes and legacies (moreso than Asians in both categories, I believe, though I’m not sure). So if you subtract all URMs, you have to also down-estimate the rest of the hooked applicants.</p>

<p>Wrt East Asian representation in “IQ-taxing fields,” whether or not that is true, it doesn’t matter–we are examining only admitted students, which is based on HS achievement. In admitting X% of students to STEM majors, Asians ought to be overrepresented at the same degree that they would have been overrepresented in the overall pool as predicted by Espenshade using solely academic criteria.</p>

722

<h1>717, ““With these two comments, you turned from being a hero to a villain. I hope you know that.””</h1>

<p>Ummm… is that a threat? :slight_smile: You may want to take a look at this one too.</p>

<p>““The idea that the true rate is around 60 percent is something even a US News journalist should immediately grasp is ludicrous, by comparison with available data… Note well that in Espenshade & Chung’s statistical model of admissions, undoing the odds-ratio of 1/3 that they find for the predictor variable “Asian”, would literally be the same as jacking up the Asian admission rate by a factor of 3… It also remains true that a 3-to-1 Asian disadvantage would (e.g., in Espenshade & Chung’s admissions model) imply nonsense such as a “true” 60 percent Asian share of admissions at Harvard,””</p>

<p>A 1-to-3 Asian disadvantage (to Caucasians in admissions rates) doesn’t mean reversing discrimination would automatically increase the Asian-student population by a factor of 3, because gains in admissions for Asian applicants would simultaneously result in losses in admissions for that of the other ethnic groups, such that the 1-to-3 ratio (between the admissions rates of Asians and Caucasians) would gradually be reversed to about unity when admissions become race-neutral. Table 9.1 of the Espenshade Study <a href=“http://www.nacacnet.org/EventsTraining/NC10/Baltimore/educational/Documents/C313.pdf[/url]”>http://www.nacacnet.org/EventsTraining/NC10/Baltimore/educational/Documents/C313.pdf&lt;/a&gt;, showed that if admissions were race-neutral for the sample elites, the student demographics would change by,</p>

<p>Asian +15.1 % from 23.9 %
White -6.5 % from 59.9 %
Black -5.5 % from 8.3 %
Hispanic -3.2 % from 7.9 %</p>

<p>So according to the Espenshade Study, the “race-neutral” Asian-student population for the sample elites are 30.4 % with URMs and 39.0 % if completely “race-neutral”. Actually this is consistent with “The Duke Test” <a href=“http://talk.collegeconfidential.com/1063406699-post662.html[/url]”>http://talk.collegeconfidential.com/1063406699-post662.html&lt;/a&gt; where the student population is about 22 % Asian undergraduates (and about 5 % “unknowns”) with URMs. By the way, Yale and Princeton look iffy again.</p>

723

<p>■■■■■■ used to amuse me, but I grew tired of them a long time ago.</p>

724

<p>Regarding the viewpoint expressed in #667 and #726, I can’t help but think of the following two quotes from The Boondocks:</p>

<p>“I want young men and young women who are not alive today to know and see that these new privileges and opportunities did not come without somebody suffering and sacrificing for them.” - Dr. King</p>

<p>“Whatever, [insert Protean N-Word here].” - Anonymous</p>

<p>All I can say is, I’m glad that the number of people who were willing to suffer and sacrifice for equality was greater than the number of people who believed that discrimination is healthy and necessary.</p>

725

<p>if i’m asian and black am i “disciminated” againt</p>

726

<p>^ Does it matter? This is not about whether person X or person Y is discriminated against. Rather, it is about the possible existence of systematic discrimination. If anything comes of this discussion or any other discussion about affirmative action, change will not be for 5-10 years, I suspect. It has no impact on any posters here. We, and those who are debating us in a rational fashion, discuss the topic because we care about social justice. (And this may indeed be a case where social justice advocacy is unnecessary–hey, I could be wrong; but you won’t know if you ignore the issue altogether.)</p>

727

<p>^Whoa? Please be sensitive. :slight_smile: Of course, it matters ----- advocating social justice and correcting systemic faults are eventually for the purpose of helping… <em>individuals</em>. And I can tell you that it does impact at least one poster here for additional data points in deciding on where to channel donations. Tiny streams accumulate into rivers. You never know who else is reading. Keep the faith. :-)</p>

728

<p>

</p>

<p>Please learn to do the math properly before posting more of these “corrections”. </p>

<p>The conventional estimate (that I used in my posting) that “40 percent of enrolled students are hooked candidates” does not double-count any students. Black recruited athletes count once, not twice, in the 40 percent figure, although they fall in more than one preference category. The other conventional estimate, 15-20 percent URM, is the assumed size of the set of ALL underrepresented minorities, no matter what additional “hooks” any one of them might have. Subtracting the second number from the first leaves 20-25 percent, which is (assuming the conventional figures apply to Duke) the fraction of the Duke student body who are (a) not URM’s, and (b) received one or another “hook” in admissions. </p>

<p>Let me make the calculation as simple as possible for you.</p>

<p>Ignore the engineering admissions, North Carolina preference, “development” admissions, clustering of Asians within states (Duke admissions limits the number of students from each state), clustering of Asians within high schools (reduced class rank and valedictorian chances per Asian applicant), competitiveness of Asians’ high schools (harder admission, according to Espenshade), and all other factors that might drive up the white/Asian SAT gap.</p>

<p>Limit attention to three types of “hook” : being an early decision matriculant (190 SAT point advantage) ; a recruited athlete (200 pts); or a Duke legacy (160 point advantage). To make the calculation more conservative, reduce the advantage to 160 points for each of these categories.</p>

<p>Let X = percentage of white enrolled students who are ED admits, minus the percentage of Asian students who were ED admits.</p>

<p>Let Y = percentage of white students who are athletes, minus percentage of Asian students who are athletes.</p>

<p>Let Z = percentage of white students who are legacies, minus percentage of Asian students who are legacies. (Express all percentages as numbers from 0.00 to 1.00).</p>

<p>Our estimate is that the Asian SAT average (1600 scale) at Duke will exceed the white SAT average (1600 scale) by at least (X + Y + Z)*160 points.</p>

<p>For example, if whites have a 10 percent advantage in each category, which is extremely plausible and probably a substantial underestimate, then the difference in average SAT’s will be (0.1 + 0.1 + 0.1)*16, or 48 points. The Duke study found 47 points.</p>

<p>Let us know if you understand this method of calculation and have any further comments. For the calculation I just sketched, one can and must sum up all the different preferences for any given white or Asian student (assuming that Espenshade & Chung and Avery/Fairbanks/Zeckhauser admissions regression models are valid).</p>

729

<p>

</p>

<p>The math is pretty simple if you read Espenshade & Chung’s paper.</p>

<p>Espenshade’s simulations are, in essence, adding up the expected number of students admitted in each category. If the probability triples for a given category, the expected admission rate triples. Thus, when Espenshade’s simulations do not result in a tripled Asian acceptance rate (or something very slightly off from that, due to the randomization), at least one of the following is the case:</p>

<p>(1) his regression model is not a good description of the admissions data; or
(2) the statement that “whites are 3 times as likely as Asians [all other factors being equal] to get a fat envelope” is not a correct summary of Espenshade’s findings.</p>

<p>At a school like Harvard, where the yield is about the same for all large subgroups of students (a steady 80 percent for blacks, whites, athletes, New Yorkers, etc), tripling the Asian admission rate relative to non-Asians would, therefore, triple the Asian matriculation share, to 50-65 percent depending on how many Asians you believe were attending in 1997. That is wildly implausible when compared to Espenshade’s data on the Asian share of high SAT scores (40 percent), to Berkeley after prop 209, and to Caltech.</p>

730

<p>

</p>

<p>Keep on flailing. It’s from a published study by one of the US IMO team coaches. Number of data points was in the thousands (olympiad qualifiers), hundreds (IMO selection program qualifiers), and on the order of 150 olympiad winners and IMO qualifiers. </p>

<p>

</p>

<p>Sure there is. This would ordinarily be a sign of discrimination (a glass ceiling, racism, etc), except that those are blind-graded math exams taken within weeks of each other.</p>

<p>

</p>

<p>That was precisely the point. Asians are a large majority (bravo, Chinese immigrants!) but they decline from a supermajority (50-80 percent per state these days) at the USAMO qualifier stage to about 30-40 percent (depending on how you count) at the IMO selection. This implies, of course, that the non-Asian share leaps from around a 35 percent minority to a 60+ percent majority, for a “white/Asian admissions odds ratio” of almost 4-to-1. </p>

<p>I emphasize again that these are historical ratios over many years, not last year’s numbers “with a denominator of 6” as you ridiculously imply.</p>

<p>

</p>

<p>Considering that I have posted that Espenshade’s regressions may (rightly or not) help decide the Jian Li investigation, it’s pretty unlikely that I claimed his results are null. His articles are about what one would expect for social scientists using statistics: a lot of mistakes, and missing the real content of his own data (or refusing to investigate clear patterns), but a large enough data set that it has to say something, no matter how badly it is handled.</p>

<p>You comments about peer review and the lack of specific scholarly refutations of Espenshade’s work are beyond naive. The standards for publishing regression studies in social science (or economics) journals are not particularly high. The refereeing is well below the standard of hard science (and even in those fields it’s far from 100 percent). It’s fairly likely that both the authors and the reviewers would miss a lot of the mistakes, and technical correctness isn’t the standard of judgement. If it says something interesting that spawns new studies and more grant funding to perform those studies, that’s pretty much the gold standard.</p>

731

<p>

</p>

<p>I did, and I appreciate your rationalism. </p>

<p>I too tried my best to be objective. After all, I am a Canuck and in a way, peering into this issue makes me feeling distinctly uncomfortable -like I am a voyeur looking into a inheritance fight.</p>

<p>My interest is more than academic of course. Whether I like it or not, many of the top grads of these elite institutions will end up at the highest level of US government and industry. What these people do will impact not only Americans but all of us who live in other countries, for the better or for the worse.</p>

<p>In recent years, it certainly has been for the worse. For my own “survival”, I have no choice but to learn more about America’s ruling class, and knowing how they are selected and where they are trained is part and parcel of the process. </p>

<p>To make a long story short, and with all due respect, I simply can not believe the ones running America today are the best and brightest the nation has to offer. What I see is a system that is rife with nepotism and evasiveness-the same type of behavour I see in elite college admission. </p>

<p>I am starting to understand.</p>

732

<p>Continuing from #722,</p>

<h1>732, ““The math is pretty simple if you read Espenshade & Chung’s paper. Espenshade’s simulations are, in essence, adding up the expected number of students admitted in each category. If the probability triples for a given category, the expected admission rate triples.””</h1>

<p>The math is pretty simple. What’s wrong is how you erroneously obtain a race-neutral Asian-student population by multiplying the original Asian-student population with a factor of 3, because of a 1-to-3 Asian disadvantage (to Caucasians in admissions rates). Also quoted from some of your earlier messages,</p>

<p>“”“The idea that the true rate is around 60 percent is something even a US News journalist should immediately grasp is ludicrous, by comparison with available data… Note well that in Espenshade & Chung’s statistical model of admissions, undoing the odds-ratio of 1/3 that they find for the predictor variable “Asian”, would literally be the same as jacking up the Asian admission rate by a factor of 3… It also remains true that a 3-to-1 Asian disadvantage would (e.g., in Espenshade & Chung’s admissions model) imply nonsense such as a “true” 60 percent Asian share of admissions at Harvard,”"</p>

<p>You are building a strawman here and it doesn’t work that way, because the 1-to-3 Asian disadvantage (to Caucasians in admissions rates) is a marginal value only true at the original student demographics. Taking the approximation that URMs can be separately treated, when more Asian applicants are admitted and the Asian-admissions rate increases, simultaneously less Caucasian applicants are admitted and the Caucasian-admissions rate decreases. Race-neutral admissions is reached when both of the admissions rates are about the same. I shall put them in symbols for illustration purposes.</p>

<p>Let A be the original admissions rate for Asian applicants,
Let C be the original admissions rate for Caucasian applicants,
Let a be the race-neutral admissions rate for Asian applicants,
Let c be the race-neutral admissions rate for Caucasian applicants.</p>

<p>3A= C, 1-to-3 Asian disadvantage to Caucasians in admissions rates at the original student demographics
a~ c, both of the admissions rates are about the same at race-neutral student demographics</p>

<p>A< a, Asian-admissions rate increases as admissions become race-neutral
C> c, Caucasian-admissions rate decreases as admissions become race-neutral</p>

<p>Combining the above 4 equations, a~ c< C= 3A, or a< 3A= C, i.e. the race-neutral Asian-admissions rate is necessarily <em>less</em> than 3 times the original Asian-admissions rate here and never reaches as high as the original Caucasian-admissions rate. So a race-neutral Asian-student population <em>cannot</em> be obtained by multiplying the original Asian-student population with a factor of 3. The correct approach is to sum the Asian admits one-by-one with their corresponding admissions rates, until the Asian admissions rate is about the same as the Caucasian admissions rate, when race-neutral admissions is reached. Referring to my earlier message #722, the race-neutral Asian-student population only increases by a factor of (30.4/ 23.9)= 1.27 with URMs and (39.0/ 23.9)= 1.63 if completely race-neutral (Yield is approximated to be about the same across different ethnic groups, so that the admit demographics is about the same as the student demographics). And it should be noted that these multiplication factors represent the ratios between the weighted-averaged admissions rates and the original admissions rate, but <em>not</em> the ratios between the marginal admissions rates (when admissions become race-neutral) and the original admissions rate.</p>

<h1>732, ““At a school like Harvard, where the yield is about the same for all large subgroups of students (a steady 80 percent for blacks, whites, athletes, New Yorkers, etc), tripling the Asian admission rate relative to non-Asians would, therefore, triple the Asian matriculation share, to 50-65 percent depending on how many Asians you believe were attending in 1997.””</h1>

<p>As explained above, the correct multiplication factor is only 1.27 with URMs, for race-neutral admissions. So we are talking about an Asian-student population of only about 30 % for race-neutral admissions, if you assume Harvard has discriminated against Asian applicants to the same extent as the sample of elites in the Espenshade Study.</p>

<h1>732, ““That is wildly implausible when compared to Espenshade’s data on the Asian share of high SAT scores (40 percent), to Berkeley after prop 209, and to Caltech.””</h1>

<p>I suppose you meant “Asian-student population” instead of “SAT scores”. :slight_smile: Table 9.1 in the Espenshade Study <em>cannot</em> be applied to Berkeley because the samples are private elites. And it is improper to blindly use the multiplication factors for any school, unless you assume all schools have discriminated against Asian applicants to the same extent. Caltech is generally regarded to run genuinely merit-based admissions, even without URMs. Referring to my earlier message #722, the Espenshade Study actually projected an Asian-student population of 39.0 % for the sample of elites, if completely race-neutral. And this is consistent with the Asian-student population at Caltech.</p>

733

<h1>734, ““I did, and I appreciate your rationalism.””</h1>

<p>CG, the “villian” saying good morning here. Is the Big Brother watching? :slight_smile: But I gotta go. I’ll reply to you later.</p>

734

<p>Some ■■■■■ messages were just being deleted and some message #s were changed.</p>

<p>In the current #732 and #733,</p>

<h1>732 referred in the messages becomes #729,</h1>

<p>and #734 referred in the messages becomes #731.</p>

735

<p>Re #730</p>

<p>Published study…is that so? Please give the authors and name of this study, and while you’re at it, please give the authors and names of the studies you mentioned several pages back (ie. the ones that supposedly offer evidence contrary to Espenshade and Chung [2005] and Kidder).</p>

<p>There is nothing ridiculous about my explanation for what you perceive as odd “progressively increas[ing]” representation. The only truly ridiculous thing is to compare the racial classification breakdown of six individuals against the racial classification breakdown of 470 individuals and then conclude that since the two aren’t the same, some groups must be encouraging their children to pick “low hanging fruit” while others are in it for the long haul.</p>

<p>I repeat: when you change the denominator from 470 to 6, it shouldn’t come as a shocker to anyone that some groups will be “overrepresented” while others will be “underrepresented.” </p>

<p>As for your “white/Asian admissions odds ratio,” is it statistically significant? Don’t forget that of your two supposed counterexamples a few pages back, neither was statistically significant at any conventional significance level, whereas the 0.33 odds ratio was significant at the 1% level. In all likelihood, your statistics background is much stronger than mine, but as I said before, I can be fooled, but not this easily.</p>

<p>Again, I wish to silence no one. I must, however, ask a question I posed earlier: if Espenshade’s research is so bad, why hasn’t anyone refuted it? If it’s so lousy and the standards for publishing are so low, why hasn’t anyone done so already? It’s been four years since it was published!</p>

<p>Indeed, to answer that the standards for publishing are not high is to shoot your own foot; if the requirements are so low, why haven’t you done so? Your posts suggest that you know something that hundreds of other social scientists interested in the subject don’t know. Why don’t you enlighten them? That’s a serious question, not a taunt.</p>

<p>Really, this is what I’ve always found hilarious about the attitude of “Espenshade’s research is crap / badly handled / contains a lot of mistakes / [insert negative descriptions or adjectives here].” If the problems are that numerous and obvious, surely at least one assistant professor somewhere would’ve noticed it already and submitted a paper to improve his publication record.</p>

736

<p>siserune - Not posting corrections at all. I don’t understand your reasoning, which may be your error or mine. Thank you for elaborating on the composition of your figures. As a statistics novice, I personally would be grateful if you explained all future calculations in the style of post #728.</p>

<p>Regarding USAMO, why is it strange that Asian representation would decrease as difficulty increased? Asians are overrepresented at the beginning levels because the pool of qualified talent is deeper–but as you narrow up, even a less qualified group will have outstanding talent at the top, and overrepresented groups will become less overrepresented. The highest level of USAMO is basically measuring mathematical genius–since mathematicians are born, not made–and there’s no reason to think that true math talent is distributed other than equally across the races.</p>

<p>Perhaps social science peer review is bad at checking statistics; I wouldn’t know. But in five years, every single other opportunistic social scientist in the nation is also bad at statistics? That kind of strains plausibility. A damning critique of Espenshade’s work, affecting the validity of his conclusions, would certainly qualify as “[saying] something interesting.”</p>

737

<p>Re: #736 Regarding USAMO</p>

<p>Keilexandra, you can ignore what siserune wrote about this. A cursory look at the lists of MOSP, USAMO qualifiers can tell that s/he just made up those numbers and claims (posts #716 and #730) based on her/his own personal bias . </p>

<p>I will do a rough count of the numbers later when I have time and post the results.</p>

738

<p>As promised, I just wasted more than an hour to compile the number of east Asian participants in USAMO, MOSP, and US IMO teams in the last 4 years using lists ([here](<a href=“http://www.unl.edu/amc/e-exams/e8-usamo/archiveusamo.shtml”>http://www.unl.edu/amc/e-exams/e8-usamo/archiveusamo.shtml&lt;/a&gt;), [here](<a href=“http://www.unl.edu/amc/a-activities/a6-mosp/archivemosp.shtml”>http://www.unl.edu/amc/a-activities/a6-mosp/archivemosp.shtml&lt;/a&gt;), and [here](<a href=“http://www.unl.edu/amc/e-exams/e8-usamo/usamoawards.shtml”>http://www.unl.edu/amc/e-exams/e8-usamo/usamoawards.shtml&lt;/a&gt;)) from the AMC site. I counted only the obvious East Asian last names (Chinese, Korean, Japanese, and maybe a few Vietnamese) and omitted obvious south Asian (and southeast Asian) names. Only US students were counted, Canadian and other international participants were not counted and not factored in the total.</p>

<p>Here are the numbers (all entries are expressed as “east asians/total participants”):</p>

<p>YEAR _____ USAMO _____ MOSP _____ WINNERS(TOP 12) ____ IMO
2006 _____ 204/404 _____ 21/54 _________ 8/12 ____________ 3/6
2007 _____ 226/459 _____ 24/55 _________ 5/12 ____________ 3/6
2008 _____ 258/465 _____ 29/56 _________ 4/12 ____________ 1/6
2009 _____ 245/431 ______ n.a. __________ 6/12 ____________ 3/6</p>

<p>The numbers of east asians in each of the categories (MOSP, winners, and IMO) matched closely with the number of USAMO qualifiers (the result is about exactly what a random choice would predict).</p>

739

<p>Props to NCL for compiling the table. Unless siserune can produe “the” paper, I think this issue has been definitively settled.</p>

740

<h1>731, ““To make a long story short, and with all due respect, I simply can not believe the ones running America today are the best and brightest the nation has to offer. What I see is a system that is rife with nepotism and evasiveness-the same type of behavour I see in elite college admission.””</h1>

<p>CG, I suppose you consider the elite admissions have failed their jobs in supplying qualified new blood to the upper echelons of our society, because their definitions of merit have swayed from predominantly academic. There are obviously major problems surfacing recently in our society. But I don’t yet see any obvious solutions. It is tricky to define what should be merit in elite admissions, for the purpose of selecting productive new blood. Let me give an example, the current Taiwanese President was elected mostly for his principled character against corruption. But this same character, has made him less responsive to public opinion, less adept in political maneuvers, and eventually led to the management disaster of the recent flood. Most people would agree that anti-corruption is virtuous for a government official. But some Taiwanese now said they maybe better off exchanging self-righteousness for some corruption. And is neopotism, (if not excessive) as in legacy status in elite admissions, necessarily unproductive? It is hard to say, because pre-existing connections may facilitate networking, which is essential for career-building. For example, please check out how some business is conducted in China, where neopotism shows you the door, but generally, you still need competency to get through the door. The point is, (except for technical settings) we may not always be able to tell what is merit for any particular real-life situation, until after some trials and errors. Furthermore, the current practices at the corporate and the government sectors, are also filtering platforms where uncompatible participants would be eliminated, potentially capable of neutralizing your desired changes in elite admissions. And as far as how to tinker the system for the better, it seems that our best academics and professionals are still scratching their heads. I don’t suppose you like what I said and I have become a “villian” again. :slight_smile: But there are at least two areas that some elite admissions may need improvement ----- abolish implicit racial discrimination and improve transparency in the selection process.</p>