<p>Since when does “slowly but surely” turn into “decimated”? But, hey, I’m an aspiring empiricist. I politely request an explanation as to how Asians are “decimated in pure merit selection such as math competitions.” You yourself presented a “corrected” table later in post 782, which apparently does not suffer from the so-called “denominator of 6” problem, and which doesn’t even corroborate your new “decimated” claim. Let’s take a look at what happened to the percentage of Asian students between 2006 and 2009 as they went from USAMO to MOSP to the Top 24 to the Top 12 and finally to the Top 6 IMO team.</p>
<p>2006: Decreased -> Increased -> Increased -> Decreased (Near perfect match between USAMO percentage and IMO percentage)
2007: Decreased - > Decreased -> Increased -> Increased (Near perfect match between USAMO percentage and IMO percentage)
2008: Decreased > Decreased -> Increased -> Decreased (Only year that actually supports siserune’s “decimated” assertion)
2009: n/a -> n/a -> Decreased -> Increased (Asians made up 57% of the USAMO qualifiers but “only” half the IMO team)</p>
<p>Of course, maybe there’s something I’m not seeing. Maybe siserune will explain what he meant by “decimated,” because going from 57% to 50% doesn’t count as decimation to me. And maybe he’ll finally explain what he meant by other groups’ having “more sophisticated educational strategies.”</p>
<p>NCL’s right on this one. There is no point in discussing ERC’s documented Asian underperformance as measured by class rank at graduation with siserune as he clearly believes that East Asians are intellectually inferior to Jews, non-Jewish whites, and South Asians, evidenced by his continuing refusal to apologize for or clarify his “more sophisticated educational strategies” and “low hanging fruit” remarks, and his newfound insistence that East Asians’ being 41.6% of all U.S. IMO team members between 2006 and 2009 is somehow evidence of being “decimated” in the competitions relative to Jews, non-Jewish whites, and South Asians.</p>
<p>Edit</p>
<p>For the record, in 2008, the only year that satisfies siserune’s “decimated” assertion, the top-scoring U.S. competitor was none other than the ethnic Chinese competitor Alex Zhai, who earned a perfect score alongside two PRC nationals.</p>
<p>If the believe of Asian intellectual inferiority can help some people sleep better at night, I’ll just say, hey, whatever works.
[/quote]
</p>
<p>My comments (and the data) are consistent with Asian intellectual superiority, inferiority, equality or any other relative standing compared to other groups. It’s an unrelated issue.</p>
<p>If you think discussing data on Asian underperformance (relative to credentials) is somehow racist, but it is not racist at all to be discussing for the 50th time the superior Asian performance (in raw numbers) compared to other groups, or the lower performance of affirmative action minorities — I wish you luck finding buyers for that theory. For my part, I think data is data and open for discussion. </p>
<p>
</p>
<p>Please display any item of evidence you claim that I “selectively picked, distorted, or made up”.</p>
<p>The USAMO counts were posted by you, so I assume those are considered OK, even if you did pump up the Asian numbers by adding foreigners where convenient. </p>
<p>The statistical significance calculations concerning the counts were manipulated downward in your postings, in precisely the manner I explained (now and earlier), as quoted above. You don’t contest that.</p>
<p>My own statistical significance calculations were apparently correct in method and result, and you haven’t contested that, either.</p>
<p>Finally, your claim of increasing Asian population being enough to cause the observed declines was just refuted a couple of postings above, and (does this sound familiar by now?) you haven’t answered that, either.</p>
<p>When you have a concrete objection, let us know. Don’t expect everyone to just take your word for it that the evidence is phony. </p>
<p>
</p>
<p>It’s very funny that these are exactly the pseudo-explanations advanced for the effect of the White variable in the same regressions and the 2004 ones, i.e., there are hidden variables driving the race effects. Sure, maybe in the 2004 results whites have better “personal qualities”, “leadership”, etc. And maybe in this case you will tell us how lower-class Asians have solidity, temperament, and character. </p>
<p>You see, you missed the point of Espenshade’s statistical models, 2004 to 2009. The entire Asian discrimination interpretation of his resuts is based on the fact that he finds that being Asian is a negative statistical predictor of admission given all the other variables in the regressions, in this case including SAT, SAT-II, ACT, grades, number of AP exams, leadership/awards, athlete and legacy status, quality and type of high school. However, his full model of admission (no. 6, the one with race/class interactions) predicts that with the same grades, scores, AP exams, etc, changing an applicant’s race variable from White to Asian would increase admission chances substantially for lower and working class applicants. The size of the predicted effect was large.</p>
<p>With respect to the high performance data that I’m allegedly omitting, there is none. Espenshade did not find any measure of achievement (more stringent than admission itself, indicating a “higher level of selection” in the sense of the current discussion) where Asians outperform whites academically after admission to selective universities. This is unexpected, because they had higher credentials coming in: the Asians display “underperformance”. I read through much of the book but don’t own it, so if you found some performance measure I missed, do let us know.</p>
<p>
</p>
<p>As I predicted, Espenshade’s methodology comes under attack when it doesn’t produce the desired result, under-admission of Asians. </p>
<p>10 percent of the data set is a substantial number and while some noise is expected in allocating the pro-Asian admission effect between lower- and working-class applicants, it’s questionable whether a completely phantom effect could show up as strong as it does.</p>
<p>This is all very funny, but what is funnier is that this objection doesn’t apply to the Espenshade regressions in the Asian underperformance discussion — the models of percentile class rank. He doesn’t consider race/class interactions there, so there is far less dependence on the size of the categories.</p>
<ol>
<li>Group academic performance in college is difficult to capture with some simple metric.
[/quote]
</li>
</ol>
<p>I have a sense of deja vu. Isn’t this exactly the stereotypical white excuse when higher Asian SATs, grades, math competition results, or other numerical measures are compared with the Asian admission rates, or Espenshade’s admission regression predictions? SATs aren’t a good measure, grades don’t reflect personal qualities, sports are the key to success, etc. The nasty oversimplified numerical metric must be replaced by one that makes the group look better — in the other case whites, in this case, Asians. This is known as “rationalization”. </p>
<p>But nevertheless, let’s consider your point. Can you suggest a metric that is clearly more stringent (a “higher level of academic selection”) than admission to elite colleges and happens after matriculation, where (US, East) Asians overperform their credentials, or their numbers in the underlying pool of candidates? If USEAsians are 40 or 50 percent of the best (top 200) incoming students at Harvard or Princeton or MIT, what is the fraction who are making valedictorian or Phi Beta Kappa or getting elite graduate fellowships?</p>
<p>
</p>
<p>We get it. Espenshade is stats god when he finds discrimination and incompetent when he finds Asian underperformance.</p>
<p>
</p>
<p>It seems that Espenshade fit the model using only the students who graduated, so the comparison is apples-to-apples. What this model says is “among students who graduated, being Asian was a substantial negative predictor on class rank”. Unless the graduation standard was lowered for Asians, this comparison would not penalize them, because there isn’t a special “extra 3% of low performing Asians who barely managed to graduate” group being added to drag down the Asian numbers, although the Asian graduation rate might be higher by 3%. </p>
<p>So flaw #1 appears not to exist.</p>
<p><a href=“2”>quote</a> The model did not consider the effect of group aggregation into different majors, asians are heavily concentrated into engineering and natural sciences, and others are more into humanities and social sciences.
[/quote]
</p>
<p>The model specifically was designed to control for the effect of different majors. There are variables in the regression indicating a major in Engineering, or Natural Science, Humanities and a few others. </p>
<p>There are a few flaws with your alleged “flaw #2”. </p>
<ol>
<li><p>The size of the Asian effect is comparable to or larger than the entire effect of majoring in Engineering or Natural Science. I think Asian and Engineering had a similar effect in model 1 and Asian and Natural Science were comparable in model 2, and those were the worst majors, but I don’t have the table in front of me. </p></li>
<li><p>Whites in engineering and natural science are also affected.</p></li>
<li><p>The credential-adjusted propensity for whites to major in engineering and natural science may be higher than for whites in general. i.e., the relative Asian preference for these majors may be partly credential-driven. So on a credential adjusted basis, as in this regression, whites may be taking a comparable beating to Asians in science and engineering.</p></li>
<li><p>After adjusting for items 2 and 3 above you are discussing second or third-order effects on the results. But the Asian effect is of the same magnitude as the entire effect of majoring in science.</p></li>
<li><p>The models fit the data quite well. Either R or R-square for model #2 was 80 percent, I don’t recall which. This reduces the scope for alternative explanations and subtle clustering effects.</p></li>
<li><p>Espenshade’s model is consistent with what is seen in the data. Median class rank was 5 points lower for Asians. His model predicts about a 5 point difference for Asians on the average after taking immigrant status into account, as I wrote in the earlier posting. </p></li>
</ol>
<p>Your “clustering” explanation does apply to Espenshade’s admission models (and data). Asians are, as Espenshade finds and as is visible demographically, highly clustered by high school. Insofar as valedictorian or top few percent class rank is important for admission to the upper schools, Asians are effectively fighting over half or a third the number of such awards compared to whites, if they are 2 or 3 times concentrated by population in the high schools they attend. This will show up as an Asian SAT penalty in the 2004 Espenshade & Chung regressions.</p>
<p>In the same way, if Asians are clustered by major more so than whites, and there are higher entry requirements or limited spaces in some of the preferred majors (engineering, premed), this will also appear as statistically equivalent to an Asian SAT penalty.</p>
<p>siserune seems to believe that if a group makes up X% at the beginning stage of a selection process, then it should always make up X% at each increasing level of selection. Anything more or less is evidence of over- or underperformance. (After statistical significance has been calculated, of course. [Except when it doesn’t help him.])</p>
<p>siserune casually dismisses it, but fact of the matter is, when the denominator changes from several hundred to less than ten, it really shouldn’t surprise anyone that some groups might be “overrepresented” when compared to the initial selection stage and some groups might be “underrepresented.” Suppose that there are 100 people in the initial selection stage, 96 of whom are East Asian and 4 of whom belong to a mythical superrace who employ one of siserune’s “more sophisticated educational strategies.” Suppose that in the final cut of ten selectees, eight are East Asian are two are mythical superrace. Since East Asian people made up 98% of the initial cut but “only” 80% of the final cut, siserune would argue that they have been “decimated” in competition because 80% is less than 98% and 20% is greater than 4%.</p>
<p>Well, 80% of what? 98% of what? 50% of what? 4% of what?</p>
<p>Ah, it all becomes clear now. 80% of 10. 98% of 100. 20% of 10. 4% of 100.</p>
<p>Personally, I’m still wondering why “Asian” is the relevant categorization, as Asia is a large and diverse continent. (And, indeed, not all persons from Asia are categorized as “Asia” by the federal definitions.)</p>
<p>I guess it’s good news that Fab has (finally!) progressed to posting numbers and calculations, even if the numbers are hypothetical and the analysis spotty.</p>
<p>
</p>
<p>Except when it happens year after year and in every analogous selection that one can come up with. </p>
<p>It’s true that you can’t take single years with a “denominator of 6”, or in your case a single hand-picked year (2008) with a “denominator of one”, as evidence of much. Such easy dismissals are not quite so credible when a larger number of similar experiments, some with denominators a lot higher than 6, all point in the same direction, and nobody seems able to identity experiments that refute the pattern.</p>
<p>
</p>
<p>In your hypothetical example, one group (non-Asians) passes the selection at a rate six times higher than that of the other (Asians; the odds ratio is (2/8)/(4/96) = 6.0). To downplay this by comparing 80 to 98 would be misleading.</p>
<p>If that pattern continued in this selection in other years, and other analogous selections showed similar patterns each year, and no analogous selections were known that didn’t display such effects, yes, it would raise questions. Depending on the amount and nature of data available it might also qualify as highly statistically significant, but that is a question for calculation and analysis, not arrogant declarations that it’s all just from random chance.</p>
<p>A real life example is graduation rates. If (hypothetically) Asians were graduating from Princeton at a rate of 98 percent, whites at 96 percent, and blacks at 94 percent, who would care about a few percentage points here or there, right? But those numbers, if consistent over time, would show whites and blacks being 2 and 3 times more likely to not graduate than Asians, which is quite a difference. If the same is also true at all the other Ivy League schools, that might or might not be statistically significant at any one school considered alone, since the total number of dropouts is small, but the pattern is a lot more significant when considered collectively. Some explanation is in order. For example, the NON-graduation rates could be driven by the admission of athletes, or affirmative action, with the white numbers reflecting the first effect and the black numbers both effects. The odds ratios are the first thing to look at, and they are often a hint at the mechanisms driving the differences. </p>
<p>Here’s a real example from the math competitions. One year of data (the first one available), denominators not tiny and not huge, no amazing significance on its own but it offers some hint of the underlying processes.</p>
<p>In 2010 the US national math olympiad was split into a Junior olympiad (USAJMO) for grade 10 and below, and the grade 12-and-under olympiad. I used NCL’s method from his earlier count of USAMO qualifiers: classify as “East Asians” those with Chinese/Korean/Japanese/Vietnamese names from the 50 US states. Results:</p>
<p>67.2 percent (148/220) of US qualifiers for the national Junior Math Olympiad (USAJMO grades 10 and under), are East Asian.</p>
<p>57.4 percent (151/263) of US qualifiers for the much more advanced USAMO (grade 12 and under, selector for the IMO team training camp), are East Asian.
Separating the data by grade level sharpens the pattern.</p>
<p>46 percent (46/99) of US grade 12 qualifiers for the USAMO are East Asian. (There were 44 clearly E.A. names and another 1-2 that I considered ambiguous but marked as Asian for present purposes.).</p>
<p>As a comparison, the winners of the grade 8 and under national competition (“Mathcounts”) are more like 70+ percent East Asian and this level of dominance has been the rate for many years.</p>
<p>We see, again (and again, and again) that as the selection level and the grade level is increased, the E.Asian representation drops. The odds ratios are substantial when going from 67 to 57 to 46 percent, worse than the 0.712 “discrimination” found in the Espenshade & Chung 2004 study. Refuting NCL’s assertions (again), two years of growth of the US Asian population share is very, very far from being enough to account for those odds ratios.</p>
<p>What this suggests is that one reason the E.Asian numbers get deflated at the higher selections is that those selections happen later. The advantages of starting early are dissipated over time, as everyone in the pipeline learns more. To the extent that this effect continues in college, it would make being Asian a negative predictor of performance (i.e. underperformance is expected), and a meritocratic selection would account for that.</p>
<p>I can only assume that siserune’s decision to ignore my assessment–siserune seems to believe that if a group makes up X% at the beginning stage of a selection process, then it should always make up X% at each increasing level of selection–is evidence of his agreeing with said assessment. If siserune does in fact believe that, the question still remains, why should “group representation” remain constant at increasing levels of selection in any given selection process? </p>
<p>To reiterate, siserune claims that as the selection level is increased, the East Asian representation drops. Yet, let us examine siserune’s own table in post 782 of the thread linked above, which was a “corrected” version of NCL’s table. siserune calculated the percentage of East Asian participants at the USAMO, MOSP, Top 12, and Top 6 stages of selection. We see the directional percentage changes as follows:</p>
<p>2006
USAMO to MOSP: Decrease
MOSP to Top 12: Increase
Top 12 to Top 6: Decrease</p>
<p>2007
USAMO to MOSP: Decrease
MOSP to Top 12: Decrease
Top 12 to Top 6: Increase</p>
<p>2008
USAMO to MOSP: Decrease
MOSP to Top 12: Decrease
Top 12 to Top 6: Decrease</p>
<p>2009
USAMO to MOSP: n/a
MOSP to Top 12: n/a
Top 12 to Top 6: Increase</p>
<p>siserune would have us believe that there is a perfect negative correlation between East Asian representation and selection level (ie. “as the selection level and the grade level is increased, the E.Asian representation drops”). Yet, his own “corrected” table does not prove his hypothesis. And the only year that had monotonically decreasing East Asian percentages, 2008, was in fact the year that U.S. competitor Alex Zhai earned a perfect score of 42/42 alongside two PRC nationals.</p>
<p>siserune appears to have a habit of only using evidence that supports his argument and discarding any evidence that hinders his argument, even if the evidence is his own!</p>
<p>Only an individual who believes that evidence behaves like a long call option payoff chart could argue that East Asians are “decimated” in elite math competitions.</p>
<p>Espenshade et al. indeed document Asian underperformance in their 2009 book, but siserune’s hypothesis is plainly illogical as it is premised on a highly dubious assumption: “group representation” should remain constant at increasing levels of selection in any given selection process.</p>
<p>The MAA provides a list of the winners of the Putnam Exam dating back to 1994. Using the years 2006 to 2009, the same years NCL and siserune used for the IMO results, we find the following:</p>
<p>2006 - 4 of the 5 Putnam fellows were East Asian
2007 - 2 of the 5 Putnam fellows were East Asian
2008 - 3 of the 5 Putnam fellows were East Asian
2009 - 3 of the 5 Putnam fellows were East Asian</p>
<p>siserune would have us believe that if the percentage of Putnam test takers was greater than 80%, 40%, 60%, and 60% in those respective years, then East Asians were “decimated” in the Putnam exam because as “the selection level…increased, the E.Asian representation [dropped].”</p>
<p>Edit</p>
<p>Once again, siserune seems to know something that other people don’t. Espenshade et al. did not offer any explanation for why Asians underperformed in their 2009 book. Perhaps siserune should contact Dr. Espenshade and proffer his “less sophisticated educational strategy” and “low hanging fruit” hypotheses?</p>
<p>in comparison to other races i really don’t think blacks or hispanics are underrepresented. as for native americans, i think i can agree with them being URM.</p>
<p>^ Doesn’t matter what you think. In general, US colleges consider Native Americans, Hispanic Americans and African Americans URMs and it is usually those three groups that receive the admissions handicap.</p>
<p>CollegeConfidential will convince you that it is a magic bullet, but it’s really not. It is an advantage, yes, but stories about kids who barely graduated and have below 1400 SATs getting into Harvard and Yale because their last name is “Menendez” are either extreme outliers or out and out lies. Don’t think you can coast on your URM status any more than you can coast on the fact that your legacy status or your athletic abilities, because you’re not the only person who has any of those things and colleges won’t take someone who they don’t think they want.</p>
<p>As others said, I’m not sure your ethnicity will give you any boost - if it does, I highly doubt it will be as much as for African Americans, Hispanic Americans, and Native Americans.</p>