<p>I just did. See above, now in boldface. Professor Thomas Espenshade of Princeton has published a new book, from which the “3-to-1 Asian discrimination” figure you quoted (inaccurately) originates. If you read that book, you will find that he has statistical models of the college class rank data for students in his sample. He finds that Asians have lower class ranks at graduation, in the raw data set and also in his statistical regression models after controlling for choice of major and a host of other factors. The effect is worse than being an athlete or a legacy.</p>
<p>It would be really funny if this leads to attempts to discredit Espenshade, who until now had been the personal stats god of those claiming discrimination. Unlike the “admissions discrimination against Asians” theory, the “US East Asians are decimated under a meritocracy” hypothesis has the advantage of being consistent both with his models of admission (where Asians suffer) and his models of class rank (where Asians underperform), in addition to the data on math competitions, prizes, professional awards, etc etc.</p>
<p>It would take you about a minute to perform your own, presumably fudge-free count. Please explain, from the US-only list in post #2 (students whose names appear on USAMO, US MOSP, or US IMO/IOI/IPhO team lists), what percentage in the top 5, 15, 25 and 81, are East Asian by your exact count or estimate. We can then compare with my statement and see whether you have any numerical disagreement, i.e., whether your tally of the US East Asian percentage exceeds 30 percent. </p>
<p>My classification of the US and non-US students was not based on names. I searched the USAMO, MOSP, US IMO, and foreign IMO lists to determine national origin. Within the US list, anyone whose name could potentially make them East Asian was counted as such, e.g., anyone with last name Lee is presumed to be Chinese or Korean, although in theory he could be of Mayflower ancestry.</p>
<p>Siserune: Again, you are providing links to CC discussion and name annotation by you. This really puts serious doubt on the validity of your claims, since you even fail to find unmolested data to support your claim. Where is your Harvard data link? In one post, you claim that the high GPA of Asians in MIT is meaningless, but in another post you claim that East Asian class rank (GPA) is lower than athletes’, and again fail to provide a link to the data to support your claims. Such contradiction is quite laughable. In this forum, we are discussing whether white or Asians are discriminated in MIT admission because MIT publishes race category data as such. There are no East Asians, Middle East Asians or South Asians category. In Espanshade’s book, there is no separation of Asians into East Asians or South Asians either. All your claims appear phony (with a lot of makeup number) and unsupported. You are doing propaganda here. Unless you can provide raw data furnished by third party unmolested, I don’t think that I can be serious with you. </p>
<p>By the way, the Chinese and Korean lineages in the Putnam top scorers comprise 32/81=39.5%. Did you make the MOSP or USAMO yourself? From your previous posting, you appear to think that your ancestor’s heritage (high IQ Jews) is superior to East Asians’. I wonder why you so care about East Asians’ inferiority.</p>
<p>As the words “I just posted it in the [CC] Harvard forum” indicated, you have access to the data by pushing your mouse button exactly three times. First to Ivy League, then the Harvard forum, then find the thread, which I opened, on the first page. </p>
<p>When you have examples of Data Molestation that are more profound than a failure to provide one-click convenience, let us know. </p>
<p>Until then, here is the list of Harvard valedictorians. Of the 18 that could be identified since 1992, only two are Asian and only one, David Liu, belongs to the population under discussion: US (domestic admissions path) East Asians. This under-representation is heavier than it looks because one has to measure it not against the whole Harvard population, but the academically elite part of the population that has a realistic chance at being valedictorian. If you believe that Asians are over-represented in the group of the strongest 200 students, then although 14-20 percent of the undergrad population they could be a much higher fraction of the competitors for the Freund prize. Alternatively, if you don’t believe that Asians are over-represented among the top students, the whole “discrimination” theory starts to look questionable. </p>
<ol>
<li><p>Membership in the potential-valedictorian pool (the academic upper tier of applicants) at the time of admission, and some imprecise but presumably difficult academic selection from within that pool into a still more elite group. </p></li>
<li><p>The decimation of (US, East) Asians under that academic selection, i.e., the sharp decrease of the US East Asian to US non-Asian ratio when going from the “high-credential potential US valedictorian pool” to the “actual US valedictorian list”.</p></li>
</ol>
<p>Note that the relevant Asian share in the starting pool of possible valedictorians is a lot higher than its share in the Ivy League student population. According to Espenshade’s data from 1997 at three schools “representing the top tier of US higher education”, Asians were 40 percent of the high SAT scorers in the applicant pool (in those days, high verbal scores from international Asian students would have been rare, so this is close to being the number of Asian-Americans) . More precisely, he published the statement that if the elite school(s), and he had the complete data on all applicants in a mid-1990’s admissions cycle, had sent admission letters first to those with SAT 1600, then the 1590’s, and so on down the list, Asians would have been 40 percent of the admissions, with the admission cutoff being close to 1500. </p>
<p>URM’s are almost nonexistent in the group of potential valedictorians, as we can also see from the Phi Beta Kappa lists at MIT. Another way to look at it is that if you exclude the students for whom URM, athlete or legacy status was needed for admission, you are left with about 60 percent of the student body, and in this group Asians are not 15-25 percent but more like 35-45 percent, and having the highest representation in the academic upper tail of the distribution. </p>
<p>You can vary the assumptions a bit as to how large the “potential valedictorian” group is in each entering class of Harvard or Princeton students. It is public information, as recently as MITChris’ posting in this forum of the 2014 MIT admission statistics, that all the elite universities (other than Caltech) admit approximately the top 10-15 percent of the class purely on academic criteria, using terms like “academic stars” or “potential summa cum laude graduates”. Taking into account that not all the students in this group, and a few students outside this group, have the grades and other indicators of potential valedictorian status, it is reasonable to assume about 100-200 US students are candidates for the #1 spot four years later. Looking at the valedictorian list, we don’t really care what it measures except that it’s some kind of difficult selection from an underlying pool whose demographics we know are highly enriched in Asian-Americans, compared both to the US population and the elite university student population. </p>
<p>The point, of course, is that the USnonAsian-to-US-East-Asian ratio grows by factors like 4 or 10 under these valedictorian selections (the number will vary depending on the specific assumptions, but what will not vary is the number being significantly larger than 1.00). The statistical significance will also be pretty high, and certainly high when considering multiple experiments such as the combination of the Harvard and Princeton valedictorian lists together with the decimation seen in the MIT PBK data. The valedictorian data, however, are noisy in that one or two Asian valedictorians would lower the ratios (somewhat) and the statistical significance (a lot) for those individual experiments. The overall statistical pattern is quite robust as it also encompasses math competitions, research prizes, elite graduate fellowships, and other selections past the college admissions stage. </p>
<p>Here is the Princeton valedictorian list for 1997 to 2009. The list is dominated by science, engineering and economics majors, so if there is any bias in this metric it should be in favor of increased Asian numbers. Of 12 US valedictorians, only one is East Asian.</p>
<p>I’ll define more precisely what I mean by “starting pool of possible valedictorians”. A better expression of the meaning would have been “valedictorian candidates based purely on objective or impersonal qualifications”.</p>
<p>The idea is to imagine some sort of ranking formula, or ranking process performed by a hypothetical version of the admissions office, based on applicants’ academic performance data or any other material submitted, but EXCLUDING (and blinded to) all personal data such as race, religion, or information about family background. Ranking by SAT score is one example of such a process and of course other information could be taken into account. The output would then be a ranking of the applicants or at least a list of the high-ranking candidates deemed (based on these completely impersonal metrics with identifying data removed) to be possible valedictorians.</p>
<p>Based on all sorts of published and anecdotal data I believe that Asians would be represented at a higher rate, call it X percent, on such a list than in the current student body at elite schools.</p>
<p>It does not follow, however, that Asians represent X percent of the best students, or that admission of Asians from that pool at a rate substantially lower than X is evidence of discrimination. The point of my postings in this thread is that if you allow the other, personal data as “predictor variables” in the statistical models that are overt or implicit in such ranking processes, the effect of being (US, East) Asian is strongly negative, as Prof. Espenshade has found in his class rank regressions. This would imply that pure merit admissions, such as one that tries to maximize the number of academic prizes earned by graduates 20 years later, or alumni donations, or other metric defined without regard to race, would either (a) select East Asians at a rate lower than their credentials would suggest when personal information is excluded, or (b) award what will show up in statistical models as a racial affirmative action for Asians, by ignoring the background data and admitting them at the “objectively required” rate and having them underperform the credentials after admission.</p>
<p>The data seen in this thread (and there is more of it, of course) suggest that universities are doing both. As detected in Espenshade’s regression models of admission, universities discount the Asians credentials. As detected in his class rank regressions, the valedictorian data, the math competition results, etc — the discounting may actually not be strong enough if pure academic ability of the graduating class is the goal.</p>
<p>Siserune: Again, you are providing links to CC discussion and name annotation by you.
[/quote]
</p>
<p>No, I performed (and posted the result of) a simple mathematical operation of “intersecting two data sets” – data sets that you have cited in your own postings as being authoritative. Specifically, I listed, in the other thread, the intersection of the following two sets:</p>
<p>A : USAMO qualifiers from the 50 US states, in 2006-2009 (lists are available at the AMC web site that you linked)</p>
<p>B : Winners of the 2009 Putnam competition; top 25 + Honorable Mention (top 81; list available at the AMC Putnam site that you also linked).</p>
<p>The result was a list of 50 students that one might call, simply, “the US winners of the Putnam competition in 2009”. No subjective process of name annotation (such as classifying some of the names as Asian), is involved in defining or determining that list of fifty. There are different methods, some more efficient than others, of finding the list, i.e., of computing the intersection of sets A and B. I explained my method on the other thread (linked above), and anyone interested can replicate the results in a matter of 10 minutes. The resulting list, which I posted, performs a very simple task: it shows the entire progression of 2006-9 American USAMO qualifiers to Putnam competition high scorers in the December 2009 contest. </p>
<p>If you know of any specific reason to disbelieve that the list performs the function advertised, what is it? Hint: additional suggestions that one or more “high-IQ Jews” (sic) are waging a propaganda campaign to defame the Chinese people, or continuing to say that data “appear” to be fudged without displaying examples of errors in the data, are not specific reasons. Displaying a Putnam winner on my list of 50 who did not quality for USAMO in 2006-9 would be a lot more convincing.</p>
<p>Since you apparently don’t have any specific dispute concerning the list (of which 50-states USAMO qualifiers made it into the group of Putnam winners), but do maintain that my count of East Asians from the 50-American-Putnam-winners-2009 list is subject to mysterious and obscure errors of “name annotation”, could you simply tell us what is your own count of the apparent number of East Asian names in the top 5, 15 , 25 and 81 (honorable mention) categories of that list? A list of four integers would suffice to answer the question.</p>
<p>No evidence was posted of high Asian GPA at MIT. You simply assume that Asian performance is higher (why?), but the data point in the opposite direction. These facts may be shocking, especially to Chinese-Americans for whom it is a nearly universal dogma that Berkeley numbers show that MIT or Harvard discriminates. Nevertheless, the data are quite consistent in contradicting that dogma, especially when considering credential-adjusted performance, and this has been evident for many years to anyone who cares to look.</p>
<p>At any rate, your “high Asian GPA” is solely a reference to the MIT biology department PBK numbers (i.e., 7 you counted as Asian) in 2007, and what I actually claimed concerning that oh-so-convincing data point is:</p>
<ol>
<li><p>your MIT Phi Beta Kappa data were carefully pre-selected, and then inflated by adding international students and US Indians (comprising four of your “seven” PBK winners, and more than doubling the number) when the question was about the underperformance of US East Asians. This is ignoring the additional problems of tiny sample size and the peculiarities of Course 7, such as the preponderance of Asians, females and pre-meds skewing the data compared to MIT as a whole. </p></li>
<li><p>The complete annual PBK list for MIT is published every year. This data is a larger sample, and lists the hometown (in particular, the US or foreign residency) for each student elected, so one can easily separate the international students. I posted the 2006-9 figures, which again show the (US, East) Asian numbers being decimated, with the US nonAsian to US.E.Asian ratios rising by factors as high as 2-to-1. Either MIT is discriminating in awarding the grades to US East Asians (while not discriminating against foreign Asians, US Indians, or non-US students!), or the US East Asian share of the top students is, under a pure academic selection, dropping hugely from the time of matriculation to the time of graduation. Here, once again, are the 2006-9 numbers, this time with percentages and Asian to non-Asian ratios:</p></li>
</ol>
<p>
</p></li>
<li><p>The PBK numbers are not meaningless, and I did not claim that. They are noisy, because grades are a noisy and manipulable measure, and they are so noisy as to be worthless when you consider single years from a single department pre-selected for its high population of (a) Asian (b) female (c) pre-meds, and (d) add in foreign Asians and US Indians, both of which are overperforming populations. Even so, the multi-year, US East Asian data is available from the Course 7 web page you cited, after subtracting non-US students identified from the MIT PBK lists. After doing this exercise, one sees that the US East Asian population at MIT is not represented at the UC Berkeley or US Bio Olympiad or US Chemistry Olympiad levels (“50+ percent” according to your postings), or even at the 1997 Asian share of high SAT scores (40 percent). I didn’t run the numbers for every year, but it’s not even clear that US East Asians are reaching their share of the Course 7 undergrads. If you think Asians are doing OK in the bio department Phi Beta Kappa results, you should explain what you think the US.E.A fraction, and the approximate USwhite-to-US.E.Asian share, is in the population of (1) MIT Course 7 majors, and (2) MIT Course 7 majors whose high school academic credentials are high enough to make them contenders for a Phi Beta Kappa selection (50-80 students per year, 7-20 in the bio department) at graduation. If USEA’s are, in fact, holding their own relative to their number of Course 7 majors, in category (2) they appear to be getting crushed by the time of graduation, because the ratios at the time of matriculation are similar to that in the science competitions.</p></li>
</ol>
<p>
</p>
<p>What I actually claimed is that Espenshade found:
– Asians had lower class ranks than whites in his data (his book has figures for the number in the top 20, 40, 50, 60, and 80 percent of ranks, with the white share being higher in every range). He did not directly compare Asians to legacies or athletes, or if he did, the information does not appear in the book. Median Asian class rank was at the 52nd percentile, about 5 points lower than the 57th percentile median for whites. The usual excuse for this is that Asians are concentrated in harder majors, but Espenshade accounts for this, and still finds underperformance: </p>
<p>– in Espenshade’s statistical model of his class rank data, controlling for field of study, high school GPA, SAT scores and many other variables, the statistical “effect” of changing a given student from white to Asian (holding all the other variables equal) is to lower the class rank at the time of graduation by 5-10 percentiles. This is a stronger effect, in Espenshade’s regression models of classrank, than changing a non-athlete to an athlete or a non-legacy to a legacy. That’s what was meant by saying that the negative effect of being Asian on class rank was stronger than being an athlete or a legacy. </p>
<p>Espenshade’s finding is not the same statement as “Asians have lower class ranks than athletes”. One expects (and predicts statistically from the available variables) a higher class rank for Asians than athletes because, as a group of matriculants, Asians have higher SAT and grades than athletes. The question addressed by Espenshade’s models, and my comments in this thread, is whether Asians over- or under-perform relative to their credentials. This is the standard question to ask in this context and anyone who has read studies on this will be familiar with statements like “test scores underpredict female performance and overpredict URM performance” (effects also found in Espenshade’s statistical models). </p>
<p>All this can be read in </p>
<p>No Longer Separate, Not Yet Equal
by Thomas Espenshade, Alexandra Radford, and Chang Young Chung
Princeton University Press, 2009</p>
<p>You can read most of the material I described online by searching for Espenshade/Radford/Chung together with terms like “52nd percentile”, or “OLS regression for percentile class rank at graduation”.</p>