"Race" in College Admission FAQ & Discussion 7

<p>

</p>

<p>re: SAT coaching, Espenshade/Radford/Chung found that, in their sample:</p>

<ul>
<li><p>Asians are more prone to pay for private SAT tutoring, </p></li>
<li><p>for those Asians who do get tutoring, the number of hours is higher than for whites</p></li>
</ul>

<p>E-R-C also applied the same type of statistical regression model that they used to investigate admissions, to analyze which factors influence a student to pay for private SAT tutoring. According to their regression, all else being equal, Asian ethnicity and first- or second-generation immigrant status both increase one’s chances of hiring a private SAT coach. The effect was substantial.</p>

<p>The admissions implication of this and other forms of over-preparation found by Espenshade would be that Asian SAT scores (and Asians’ academic credentials in a more general sense) are expected to be, on average, inflated: Asians’ scores and credentials would, to some small or large extent, tend to overpredict Asians’ academic performance. This is precisely what Espenshade found in the chapter on academic performance at college.</p>

<p>Some might find it amusing or appalling that the research apparatus of quantitative social science, with all due ceremony, has apparently validated some of the hoary stereotypes about academic preparation among Asian applicants to selective colleges.</p>

<p>

</p>

<p>I’m not sure why you think that’s relevant. </p>

<p>First, blacks and Hispanics (and international students) are on different admissions tracks than US white and US Asian applicants, and we were discussing only the academic and admission situation of US whites compared to that of US Asians, particularly US East Asians.</p>

<p>Second, as far as it concerns discussions of “summer schools and academic camps” in elite college admissions, black/Hispanic students might as well be living on other planets. The programs they attend (or attended in the mid-1990’s) are largely intended to raise performance up toward the white/Asian levels, from below, and are often associated with words like remedial, minority outreach, youth enrichment and corporate donations. The programs Asians and whites like to attend are for those wanting to go far beyond the mean, and are associated with words like gifted program, acceleration, and Sputnik.</p>

<p>It’s the same with SAT coaching for minorities, by the way. Other studies have shown that blacks and Hispanics don’t necessarily use SAT tutoring at lower rates than whites, even when taking income into account. It’s low scores (or feared low scores), whatever that means for an individual, that drive the consumption of SAT tutoring. But given the national score distributions, the motivation must surely be different than for the high-range Asian college applicants. Everyone wants to raise their scores, but URM are trying to keep afloat and Asians are trying to “rock the curve”.</p>

<p>

</p>

<p>It is well known that SAT scores overpredict performance for blacks and underpredict it for females. We weren’t discussing the first case. For females, yes, the meritocratic admissions procedure would be to cut them some slack on test scores in admission, which is what actually happens. Whether there is too much or too little SAT bonus (so to speak) for females, I can’t say. Espenshade’s data show overperformance by females, but he uses grades which are a metric that is biased upward for females.</p>

<p>Edit: I should point out that Espenshade’s academic performance models, consistent with all other studies, did manifest the over/under prediction effects for blacks and females. This is another indicator that he also got it right when picking up an underperformance effect for Asians. He also found overprediction for Hispanics, and I gather from your comment that this, too, is consistent with SAT studies, but I am familiar mostly with the black and female data.</p>

<p>Re: #927

</p>

<p>By excluding college dropouts, this is a comparison between the top 89% white,top 78% black, top 88% Hispanic, and the top 93% asian students. If we are to compare the performance of two schools in a hypothetical standard test, would it be a fair comparison if one school submits its top 50% scores while the other school submits only top 10% of its scores? Maybe a good principal can make a convincing argument that all those failed the test shouldn’t be counted. </p>

<p>Comparing only graduates, the median class rank for whites and asians are 57 and 52. If the dropouts are included at the bottom of class, using simple linear regression and assuming uniform distribution between given sample points in figure 6.3 (p246), the median for whites drops slightly to 56.7 and median for asians moves up to 54.2. The inclusion of dropouts in the analysis makes the supposed white-asian gap shrink by half to 2.5%, I don’t have the data to know if it is statistically significant or not, and this is before we consider the effect of uneven distribution of students in different majors.</p>

<p>

</li>
</ol>

<p>blah, blah, blah, …
Your are making a bunch of claims that simply cannot be found in the book.</p>

<p>It is fairly simple to show that re-subgrouping and re-ranking of a pre-ranked sequence will almost always lower the average rank of a highest pre-ranked subgroup. For a non-uniform re-subgrouping, this effect can be very large. I have given [one</a> such example](<a href=“http://talk.collegeconfidential.com/1064809628-post923.html]one”>http://talk.collegeconfidential.com/1064809628-post923.html).</p>

<p>So instead of all your nonsensical gobbledegook and hand waving, why don’t you give us some example or calculation to show that the un-even aggregation into different majors will not depress the rank of a highest pre-ranked subgroup?</p>

<p>

</p>

<p>Let me get this straight. If Asians have “higher…utilization of summer schools and academic camps” than whites, then the “upward bias” that results from their “inflated” preparation needs to be “meritocratically discounted,” but if blacks and Hispanics also show “higher utilization” of said programs, there’s no “upward bias” at all: they’re just preparing adequately, there’s no “inflation,” and so there’s no need for any “meritocratic discounting”? O…K…and I suppose your point, if true, that blacks and Hispanics do not attend the same “summer schools and academic camps” as Asians rationalizes away your double standard, then. Or is blacks’ SAT performance already “meritocratically discounted,” since it is “well known that SAT scores overpredict performance for blacks”?</p>

<p>

</p>

<p>Wait, if everyone wants to raise his scores, then aren’t the scores of all students “inflated”? Should every applicant’s scores be “meritocratically discounted” to reflect what it would have been without the “upward bias” that surely resulted from their efforts to raise their scores (ie. “inflate” them)?</p>

<p>siserune, if you’re training to be an academic social scientist, I really wish you the best of luck in getting published. Your posts show a clear history of picking and choosing evidence that suits your hypotheses and discarding anything that doesn’t fit, even if the evidence was created by you. I’m fair: I don’t dispute ERC’s findings of Asian underperformance. I simply dismiss your “less sophisticated educational strategy” and “low hanging fruit” explanations.</p>

<p><a href=“fab:”>quote</a></p>

<p>

Since when does “slowly but surely” turn into “decimated”?

[/quote]
</p>

<ol>
<li><p>Multiple gradual selections take place over time. Clearly you don’t expect radical changes at each and every selection.</p></li>
<li><p>You should learn what “decimated” means. It is a precise and apt descriptor for the declines under discussion: some gradual, some drastic, and some of the latter as a result of several of the former. </p></li>
</ol>

<p>

</p>

<p>If you want to dispute #782, please (a) read it carefully, and (b) quote in this thread any parts you claim are wrong. Some of your objections were answered in #782, with specific calculations. (The rest of your objections are wrong or irrelevant, but we can come to that in due time as soon as you actually indicate a statement you disagree with.) </p>

<p>NCL’s objections are apparently dead, by the way, as he does not dispute the refutation of his remarks posted in the current discussion. (He read and replied to the posting containing the refutations.) If you think you can do better, by all means, let us know what errors I have made. We await your insights. </p>

<p>

</p>

<p>That’s not valid. MOSP –> Top24 is not a selection, and while Top24 to Top12 could be considered a selection, it’s not independent from the preceding one. </p>

<p>There are exactly two selection paths in the table:</p>

<p>USAMO –> MOSP –> IMO and USAMO —> Top 24 —> IMO</p>

<p>For the first, the data is a mixture of different MOSP groups and the data was missing for 2009, so I posted a discussion of the first transition only, which did display a decline (overall and within each year).</p>

<p>For the second, full data was available for 2006-9 and I posted calculations in #782. Again, a decline was visible.</p>

<p>If you have further questions, please refer to and quote #782.</p>

<p>

</p>

<p>Er…in comparison to which races?</p>

<p>siserune, which definition of “decimated” would you like for me the learn? The “to select by lot and kill every tenth person of” definition? Or maybe the “to destroy a great number or proportion of” one? Oh, my bad, maybe you meant the “to reduce markedly in amount” one? For the sake of good-faith discussion, I’ll assume you meant the “reduce markedly in amount” one.</p>

<p>So apparently siserune didn’t bother to adjust NCL’s table to reflect the true selection paths, but that is fine; he has done so now, and I will follow his new correction. Referring to the table in post 782 of [the</a> older but related](<a href=“http://talk.collegeconfidential.com/college-admissions/790609-do-elite-colleges-discriminate-against-asian-students-53.html]the”>http://talk.collegeconfidential.com/college-admissions/790609-do-elite-colleges-discriminate-against-asian-students-53.html) thread, let’s start with USAMO -> MOSP -> IMO for the years 2006-2009.</p>

<p>2006
USAMO (50%)
MOSP (39%, decrease)
IMO (50%, increase)</p>

<p>2007
USAMO (49%)
MOSP (44%, decrease)
IMO (50%, increase)</p>

<p>2008
USAMO (55%)
MOSP (52%, decrease)
IMO (17%, decrease [note: the 1/6 refers to U.S. competitor Alex Zhai, whose apparent “less sophisticated educational strategy” and chasing of “low hanging fruit” nevertheless earned him a full score of 42/42 alongside two other people whose groups according to siserune also practice a “less sophisticated educational strategy”])</p>

<p>2009
USAMO (57%)
MOSP (n/a)
IMO (50%, unknown MOSP percentage)</p>

<p>Of course, it appears that siserune chose to look only at the “Total” percentages, where indeed one can see a decline from 53% to 44.9% to 41.6%. And as usual, siserune only tells you half the story; he claims that this is definitive proof of East Asians’ being “reduced markedly in amount” even though two years show an increase in East Asian percentage from MOSP to IMO (2006-2007), the only ostensible year that backs siserune’s “decimation” assessment in fact saw the “low hanging fruit” and “less sophisticated educational strategy” guy score 42/42 on the IMO alongside two PRC nationals (2008), and one year doesn’t even have full data.</p>

<p>In the world of siserune, a group that makes up 53% of all beginning participants and 41.6% of all finalists in a four-year period was “decimated” in the selection process. Somebody please contact Dr. Espenshade. I think siserune has an explanation for ERC’s documented Asian underperformance–“less sophisticated educational strategies” and too much focus on “low hanging fruit”!</p>

<p>Re: #930

</p>

<p>You should go back and re-read [my</a> response](<a href=“http://talk.collegeconfidential.com/1063438681-post802.html]my”>http://talk.collegeconfidential.com/1063438681-post802.html) to you in [an</a> old thread](<a href=“http://talk.collegeconfidential.com/college-admissions/790609-do-elite-colleges-discriminate-against-asian-students.html]an”>Do Elite Colleges Discriminate Against Asian Students? - Applying to College - College Confidential Forums). </p>

<p>What the numbers in USAMO qualifier lists show clearly is a trend that each year, East Asian students perform better than the preceding classes, a trend that is more dramatic in recent years. This is reflected as a higher percentage of E. Asian 9th or 10th graders in the USAMO than 12th graders. You claim that this is the proof of asian underperformance.</p>

<p>However, to prove you claim, you cannot simply compare the percentile difference between 9th and 12th graders in the same year, but to follow each freshmen class through its senior year and see if there is performance drop-off along the way. The AMC site has records that go all the way back to 1985. Let me tell you, the evidence of asian underperformance in USAMO is simply not there.</p>

<p>Let’s go over some numbers. I did a count myself and the numbers of identified of E. Asian surnames may be slightly different from yours. (I only counted USAMO, not USAJMO; I will use your USJMO number).</p>

<p>One year data, percentage of USAMO qualifiers in 2010:
…% E. Asians
9th graders (Classs’13): … 16/22 (72.7%)
10th graders (Class’12): … 30/42(71.4%)
11th graders (Class’11): … 50/87(57.5%)
12th graders (Class’10): … 45/100(45%)</p>

<p>The 9th and 10th graders did better than their 11th and 12th grade counterparts. And the proportions of 9th and 10th graders qualified for USAMO (71-72%) and USAJMO (67.2%, your number) are very similar, no under-performance here. (Of the 12 USAMO qualifiers in grades 8th and lower, 8 are East Asians (75%); I will discount this number here, because it’s an opportunity available to only very few middle school students.)</p>

<p>We can go back a few more years and follow a few recent classes from 9th through 12th grade and see how they performed in USAMO qualification through all 4 high school years. </p>

<p>… Class’04 … Class’05 … Class’06 … Class’07 … Class’08
9th: … 16/48(33%) … 16/47(34%) … 18/32(56%) … 9/25(36%) … 26/35(74%)
10th: … 42/111(38%) … 25/70(36%) … 34/68(50%) … 27/65(42%) … 77/126(61%)
11th: … 18/49(37%) … 22/57(39%) … 27/61(44%) … 41/82(50%) … 74/124(60%)
12th: … 27/82(33%) … 23/76(30%) … 48/112(43%) … 46/130(35%) … 68/129(53%)</p>

<p>… Class’09 ……… Class’10 ………… Class’11 ……… Class’12 ….…… Class’13
9th: … 28/61(46%) …… 31/62(50%) …… 42/71(59%) … 40/52(77%) … 16/22 (72.7%)
10th:… 54/110(49%) … 61/110(55%) … 63/104(61%) … 30/42(71%)
11th: …64/119(54%) … 60/114(53%) … 50/87(57%)
12th: … 62/124(50%) …45/100(45%)
Data presented as number of East Asians/numbers of all qualifiers in the same grade, percentage in parenthesis. </p>

<p>I don’t know if Class 2012 and Class 2013 will keep their pace. But I don’t see any pattern of underperformance here. And you won’t find it even if you go all the way back to 1985. What the data do show, however, is a improving trend of performance in qualifying for USAMO by E. Asian students year over year, which is really a reflection of the underlining demographic shift among some subgroups of high school students of E. Asian heritage. I have all the USAMO data (1985-2010) imported into a database, if you want data for a specific year or class, I will post them. It gets very tiring to type and format them out in that little post message box.</p>

<p><a href=“NCL:”>quote</a></p>

<p>I have all the USAMO data (1985-2010) imported into a database, if you want data for a specific year or class, I will post them

[/quote]
</p>

<p>Excellent!</p>

<p>I’d like to verify some of the numbers in your posting and make further analyses, including additional years. That is, I would like to see the whole database. I can receive email at <my cc=“” username=“”> at gmail.com. Thanks.</my></p>

<p>^All data in my database are imported from records on the AMC site. It is very easy to check my counting and tabulation using those records. </p>

<p>Here are data for 5 more classes (class '99-'03). I will post some more later.</p>

<p>… Class 1999 … Class 2000 … Class 2001 … Class 2002 … Class 2003
9th: … 1/9(11%) … 5/14(36%) … 4/19(21%) … 4/19(21%) … 19/49(39%)
10th: … 2/25(8%) … 8/31(26%) … 8/31(26%) … 14/60(23%) … 31/90(34%)
11th: … 13/47(28%) … 13/48(27%) … 16/49(33%) … 8/50(16%) … 19/65(29%)
12th: … 11/64(17%) … 16/64(25%) … 28/69(41%) … 20/84(24%) … 30/83(36%)</p>

<p>Data presented as number of East Asians/numbers of all qualifiers in the same grade, percentage in parenthesis.</p>

<p>… </p>

<p>5 more classes.</p>

<p>….…… Class 1994 … Class 1995 … Class 1996 … Class 1997 … Class 1998
9th: …… 3/8(38%) … 2/6(33%) … 2/11(18%) … 3/10(30%) … 2/4(50%)
10th: … 7/24(29%) … 5/26(19%) … 7/26(27%) … 6/19(32%) … 5/23(23%)
11th: … 5/43(12%) … 9/50(18%) … 15/66(27%) … 10/53(19%) … 10/56(18%)
12th: … 7/57(12%) … 13/63(21%) … 13/76(17%) … 14/66(21%) … 12/64(19%)</p>

<p>…
all the rest</p>

<p>…Class 1985 … Class 1986 … Class 1987 … Class 1988 … Class 1989
9th: … … … … … … … … … … … … … … … … 1/3(33%) … 0/1(0%)
10th: … … … … … … … … … … 1/3(33%) … 1/5(20%) … 0/17(0%)
11th: … … … … … . 2/20(10%) … 1/24(4%) … 10/38(26%) … 4/40(10%)
12th: … 7/37(19%) … 7/43(16%) … 5/28(18%) … 15/63(24%) … 11/83(13%)</p>

<p>……… Class 1990 … Class 1991 … Class 1992 … Class 1993
9th: …… 0/1(0%) … . 1/3(33%) … 1/1(100%) … 1/3(33%)
10th: … 2/13(15%) … 5/14(36%) … 4/22(18%) … 4/8(50%)
11th: … 9/39(23%) … 13/40(33%) … 16/50(32%) … 7/36(19%)
12th: … 7/54(13%) … 16/57(28%) … 17/55(31%) … 13/58(22%)</p>

<p><a href=“NCL:”>quote</a></p>

<p>^All data in my database are imported from records on the AMC site. It is very easy to check my counting and tabulation using those records.

[/quote]
</p>

<p>Your data aren’t replicable, and of course the real point is not only to verify your counts, but to use statistical software to perform more complicated calculations than is possible from the counts alone. That requires the database. One would have to reconstruct all your tabulations from scratch, contestant by contestant. Even for the counts alone, there are questions:</p>

<p>Your number of qualifiers from the 50 US states for 2000-2009 are as follows.</p>

<p>year 2000–2001–2002–2003–2004–2005–2006–2007–2008–2009
NCL 222----257—307—234—232—237—381—426----429—394</p>

<p>Including the numbers of qualifiers listed at wikipedia (which includes non-US) one sees that the discrepancy with your numbers is substantial.</p>

<p>year 2000–2001–2002–2003–2004–2005–2006–2007–2008–2009
NCL 222----257—307—234—232—237—381—426----429—394
wiki 239----268—326—250—261—259—432—505----503—514
diff 017----011—019—016—029—022—051—079----074—120</p>

<p>Some of these differences are huge, e.g., I don’t believe that 120 competitors from 2009 were non-US or below grade 9 in 2009. One can check this at the AMC site, and I get the following counts:</p>

<p>442 US grades 9-12 student = Total 514 (wiki was correct) - 32 non-US (30 Canada, 2 Korea, 1 Puerto Rico, 1 Guam) - 40 US youngsters (grades 4 to 8).</p>

<p>Your count was 394, which is off by 48, or more than 10 percent:
394 = 52 (9th graders, class of 2012) + 104 (10th graders, class of 2011) + 114 (11th graders, class of 2010) + 124 (12th graders, class of 2009).</p>

<p>Correct me if I am misreading your table, which I reproduce here:</p>

<p><a href=“NCL%20table%20extract:”>quote</a>
… Class’09 ……… Class’10 ………… Class’11 ……… Class’12 ….…… Class’13
9th: … 28/61(46%) …… 31/62(50%) …… 42/71(59%) … 40/52(77%) … 16/22 (72.7%)
10th:… 54/110(49%) … 61/110(55%) … 63/104(61%) … 30/42(71%)
11th: …64/119(54%) … 60/114(53%) … 50/87(57%)
12th: … 62/124(50%) …45/100(45%)

[/quote]
</p>

<p>I assume you don’t expect me to serve as a personal fact-checker going through all the other years. If you open-sourced your data it could quickly be standardized including a consensus on the classification of the qualifiers, and statistical modeling can be performed. I assume you’re not afraid of some harmless R code (which I can post) being run on the database to sift out the patterns in the data, right?</p>

<p>Historical data are at:
[USAMO</a> Archive](<a href=“American Mathematics Competitions | Mathematical Association of America”>American Mathematics Competitions | Mathematical Association of America)
and
[United</a> States of America Mathematical Olympiad - Wikipedia, the free encyclopedia](<a href=“http://en.wikipedia.org/wiki/United_States_of_America_Mathematical_Olympiad]United”>United States of America Mathematical Olympiad - Wikipedia)</p>

<p>The table in its present form shows clear declines, by the way; it supports the Asian decline pattern. Before getting into that, I’ll await your explanation as to whether the table does, in fact, represent the USAMO counts to any accuracy. If I have misread your table, let me know.</p>

<p>
[quote=siserune</p>

<p>]

Your count was 394, which is off by 48, or more than 10 percent:

[/quote]
</p>

<p>For year 2009 data, 1 page was not imported and processed (page 3 of 9) in the earlier post. Below is the complete data set with the amended 2009 data. I also caught a couple of typos (all corrections are in red).</p>

<p>…Class 1985 … Class 1986 … Class 1987 … Class 1988 … Class 1989
9th: … … … … … … … … … … … … … … … … 1/3(33%) … 0/1(0%)
10th: … … … … … … … … … … 1/3(33%) … 1/5(20%) … 0/17(0%)
11th: … … … … … . 2/20(10%) … 1/24(4%) … 10/38(26%) … 4/40(10%)
12th: … 7/37(19%) … 7/43(16%) … 5/28(18%) … 15/63(24%) … 11/83(13%)</p>

<p>……… Class 1990 … Class 1991 … Class 1992 … Class 1993
9th: …… 0/1(0%) … . 1/3(33%) … 1/1(100%) … 1/3(33%)
10th: … 2/13(15%) … 5/14(36%) … 4/22(18%) … 4/8(50%)
11th: … 9/39(23%) … 13/40(33%) … 16/50(32%) … 7/36(19%)
12th: … 7/54(13%) … 16/57(28%) … 17/55(31%) … 13/58(22%)</p>

<p>….…… Class 1994 … Class 1995 … Class 1996 … Class 1997 … Class 1998
9th: …… 3/8(38%) … 2/6(33%) … 2/11(18%) … 3/10(30%) … 2/4(50%)
10th: … 7/24(29%) … 5/26(19%) … 7/26(27%) … 6/19(32%) … 5/23(23%)
11th: … 5/43(12%) … 9/50(18%) … 15/66(27%) … 10/53(19%) … 10/56(18%)
12th: … 7/57(12%) … 13/63(21%) … 13/76(17%) … 14/66(21%) … 12/64(19%)</p>

<p>… Class 1999 … Class 2000 … Class 2001 … Class 2002 … Class 2003
9th: … 1/9(11%) … 5/14(36%) … 4/19(21%) … 4/19(21%) … 19/59<a href=“39%”>/color</a>
10th: … 2/25(8%) … 8/31(26%) … 8/31(26%) … 14/60(23%) … 31/90(34%)
11th: … 13/47(28%) … 13/48(27%) … 16/49(33%) … 8/50(16%) … 19/65(29%)
12th: … 11/64(17%) … 16/64(25%) … 28/69(41%) … 20/84(24%) … 30/83(36%)</p>

<p>… Class’04 … Class’05 … Class’06 … Class’07 … Class’08
9th: … 16/48(33%) … 16/47(34%) . 18/32(56%) … 9/25(36%) … 26/35(74%)
10th: . 42/111(38%) . 25/70(36%) . 34/68(50%) … 27/65(42%) … 77/126(61%)
11th: . 18/49(37%) … 22/57(39%) . 27/61(44%) … 41/82(50%) … 74/124(60%)
12th: . 27/82(33%) … 23/76(30%) . 48/112(43%) . 46/130(35%) . 68/129(53%)</p>

<p>… Class’09 ……… Class’10 ………. Class’11 ……… Class’12 ….…. Class’13
9th: … [color=red]29/62(47%) …. 31/62(50%) …. 42/71(59%) … 46/61(75%) .16/22 (72.7%)
10th:… 54/110(49%) … 61/110(55%) … 67/119(56%) . 30/42(71%)
11th: . 64/119(54%) … 71/130(55%) … 50/87(57%)
12th: . [color=red]69/139<a href=“50%”>/color</a> . 45/100(45%)</p>

<p>

</p>

<p>Well, if the thought help you sleep better, …</p>

<p>More data “issues”, still no data. </p>

<p><a href=“NCL:”>quote</a>]
For year 2009 data, 1 page was not imported and processed (page 3 of 9) in the earlier post. Below is the complete data set with the amended 2009 data. I also caught a couple of typos (all corrections are in red).

[/quote]
</p>

<p>It doesn’t inspire confidence that it’s still wrong – and that’s just 2009.</p>

<p>I performed, again, a hand count for 2009, annotating a printout. The results were the same as my previous count except for an additional US youngster (total 41).</p>

<p>The correct total for 2009 appears to be, once again, different from yours.</p>

<p>441 US grades 9-12 student = Total 514 (wiki was correct) - 32 non-US (28 Canada, 2 Korea, 1 Puerto Rico, 1 Guam) - 41 US youngsters (grades 4 to 8).</p>

<p>Your new count is 449, which is off by 8:
449 = 61 (9th graders, class of 2012) + 119 (10th graders, class of 2011) + 130 (11th graders, class of 2010) + 139 (12th graders, class of 2009).</p>

<p>Since your data are “imported and processed”, apparently by a computer program, why should it be more reliable for the other years? </p>

<p>You had similar data “issues” with the earlier table (adding Canadians to the USAMO winners). It’s great that somebody is counting — the p-values for the tests on your current version of the tables are already amusingly low! But it would be good to know that the data are real. Is there some reason you’re refusing to expose your database to statistical calculations with the R code for those calculations posted publicly?</p>

<p><a href=“fabrizio:”>quote</a><br>
So apparently siserune didn’t bother to adjust NCL’s table to reflect the true selection paths, but that is fine; he has done so now, and I will follow his new correction.

[/quote]
</p>

<p>There is no “new correction”. Everything was already correct in 782, a posting that you say is wrong but but whose supposedly wrong contents you can’t or won’t quote here. </p>

<p>

</p>

<p>Your mistakes this time are as follows.</p>

<ol>
<li><p>Not quoting the posting you dispute (782). Apparently you can’t argue with what was written, so substitutions are preferred.</p></li>
<li><p>Assuming that a 50% empirically observed (3 of 6) qualification rate for one year’s IMO refutes the idea that the true Asian qualification rate is something lower (such as 40% or 45% or whatever). There have been a few years with a woman on the team, 1/6 = 17% of the team members, and nobody would cite that as an “increase” or “overperformance” relative to some lower fraction in the selection pool, such as 5% female representation at that year’s MOSP or one in the Top 24.</p></li>
<li><p>If the true expected share of USEAsians on the IMO (the statistical average when you re-run that year’s olympiads over and over on the same students) were any number between 41.6% (5/12) and 50%, then, in a simplified model where the number of selectees is only 2 or 3 in each year, in most years the observed count would be 50%. In the same way, in most years the observed number of females is 0.00%, 0/6, but there is nevertheless a higher underlying qualification rate and it is not correct to say that 0.00 is the female rate even though NCL-style tables would indicate this for particular years. </p></li>
<li><p>To estimate the “true” female or Asian rates we have to use some averaging process that aggregates the multi-year data. In 782 I used the simplest one, just adding up the different years’ data. Probably some more sophisticated weighting process is better that takes into account the difficulty of the selection. But it’s pointless to go to such lengths for 4 years of data when, **as I wrote in 782<a href=“but%20you%20don’t%20quote”>/b</a>, there is plenty of other data saying the same thing more sharply. The main interest of NCL’s old and new tables is that they were presented as refutations but in fact support the underperformance thesis. </p></li>
<li><p>Recall that I posted my ideas before NCL’s table or Espenshade’s new book appeared. You mistakenly assume that my ideas somehow are tied to the table or the book. They aren’t; the tables and book are supporting evidence, and there is plenty of other evidence from other sources.</p></li>
</ol>

<p><a href=“NCL:”>quote</a></p>

<p>… Class’04 … Class’05 … Class’06 … Class’07 … Class’08
9th: … 16/48(33%) … 16/47(34%) . 18/32(56%) … 9/25(36%) … 26/35(74%)
10th: . 42/111(38%) . 25/70(36%) . 34/68(50%) … 27/65(42%) … 77/126(61%)
11th: . 18/49(37%) … 22/57(39%) . 27/61(44%) … 41/82(50%) … 74/124(60%)
12th: . 27/82(33%) … 23/76(30%) . 48/112(43%) . 46/130(35%) . 68/129(53%)</p>

<p>… Class’09 ……… Class’10 ………. Class’11 ……… Class’12 ….…. Class’13
9th: … 29/62(47%) …. 31/62(50%) …. 42/71(59%) … 46/61(75%) .16/22 (72.7%)
10th:… 54/110(49%) … 61/110(55%) … 67/119(56%) . 30/42(71%)
11th: . 64/119(54%) … 71/130(55%) … 50/87(57%)
12th: . 69/139 (50%) . 45/100(45%)</p>

<p>

</p>

<p>Well, if the thought help you sleep better, .

[/quote]
</p>

<p>Well, consider the classes above. I assume they cover (more than) the period of “modern demographics”. </p>

<p>Don’t you find it interesting that for the classes with full data, 2004-10: </p>

<ul>
<li>the 12th grade rate in the table is always lower than 11th grade (probability 1/128);<br></li>
</ul>

<p>AND - the 12th grade rate is lower than 10th grade in 6/7 cases (prob 1/8);</p>

<p>AND - the 12th grade rate is lower than the grade 9 result in 6/7 cases (prob 1/8)?</p>

<p>These are independent, so the probability of such a pattern of decrease is p = 1/128 * 1/8 * 1/8 = 1/8192. </p>

<p>These effects are there for obvious reasons:
– grade 11-12 USAMO qualification is a lot harder than the grade 9-10 standard.</p>

<p>– the effect of the earlier start of the recent-demographics E.Asian (mostly Chinese) students is reduced as students age. </p>

<p>The 9th grade numbers are noisy for similar reasons to the 8th-and-below figures. If you drop the 9th grade numbers from consideration the probability is still low, 1/1024 < 0.001. The 9th grade pattern seems to be continuing for the class of 2011, with the 9th grade rate being higher than the 11th grade rate (which if previous trends hold, will then be higher than the 12th grade rate).</p>

<p>I’d love to hear the explanation for all this in terms of Chinese (or East Asian) American demographic trends.</p>

<p>Apparently siserune does not like quoting his own tables, perhaps because the very tables he constructed himself do not support his hypothesis. It would explain why he felt the need to find “fresh evidence” as well as why he insinuated that I dishonestly modified his tables to suit my argument; he knows that his tables do not support his hypotheses. There were two tables in post 782 of [the</a> older but related thread](<a href=“http://talk.collegeconfidential.com/college-admissions/790609-do-elite-colleges-discriminate-against-asian-students-53.html]the”>http://talk.collegeconfidential.com/college-admissions/790609-do-elite-colleges-discriminate-against-asian-students-53.html), as follows:</p>

<p>

</p>

<p>

</p>

<p>siserune claims that these tables show two selection paths: USAMO –> MOSP –> IMO and USAMO –> Top 24 –> IMO. siserune argues that as the selectivity increases, East Asian representation decreases because East Asians have a fixation on “low hanging fruit” and employ “less sophisticated educational strategies” than siserune’s mythical superrace. Yet, a casual glance at the tables show that East Asian representation does not monotonically decline with increasing selectivity. Again, the only year that supports his “consistent” decline hypothesis is 2008, which was the year that three “low hanging fruit” and “less sophisticated educational strategies” students earned 42/42 on the IMO.</p>

<p>Now that I have quoted siserune’s table, we see that none of the percentages I quoted were wrong. siserune himself did not deny this; he simply refused to check his own tables as a delaying tactic.</p>

<p>

</p>

<p>So you admit that your data analysis is garbage because you can’t draw conclusions from the percentages? To quote Jamie Foxx from Law Abiding Citizen, I think we’re done here.</p>

<p>Edit</p>

<p>ERC claimed that it is not known why Asian students underperform. I reiterate what I wrote earlier: siserune ought to contact Dr. Espenshade and present his “low hanging fruit” and “less sophisticated educational strategies” explanations. siserune’s ideas are indeed his own, but they don’t have to be. He clearly thinks he’s right, and I am simply encouraging him to spread his mythical superrace gospel to Dr. Espenshade in the name of social science.</p>

<p>MODERATOR’S NOTE: </p>

<p>After months of activity and thousands of views, this thread is closed. The new thread </p>

<p><a href=“http://talk.collegeconfidential.com/college-admissions/1366406-race-college-admission-faq-discussion-10-a.html[/url]”>http://talk.collegeconfidential.com/college-admissions/1366406-race-college-admission-faq-discussion-10-a.html&lt;/a&gt; </p>

<p>is open for your polite, thoughtful discussion.</p>