I’ve seen this claim made in numerous news articles, but the Plantiff’s expert did 2 analyses – one with the full student body and one with the listed hooks removed. They came to a similar conclusion. The specific Asian regression coefficients for admission decision using full controls (model 6) were -0.33 (0.07) for unhooked and -0.26 (0.07) for full student body. Both suggest a bias against Asian applicants, but the magnitude of the coefficient and corresponding estimated effect is small – many times less than the effect of any hook.
The Plantiff’s analysis suggests that the bias is greater among unhooked applicants than among hooked applicants, which seems like a reasonable possibility. For example, recruited athletes are likely to have a very different admissions process in which athletes sent “likely letters” with endorsement by coaches (and pre-filter on some basic academic/stat criteria) are almost always admitted, regardless of race. This would make it appear than “likely letter” admits have little URM boost compared to unhooked applicants since both URM and non-URMs almost always get admitted. Similarly, it would appear that recruited athletes have little Asian penalty compared to unhooked applicants. A similar effect occurs for other significant hooks – URM boost and Asian penalty both appear to have blunted effects compared to unhooked applicants. The Plantiff emphasizes the unhooked model because it shows a larger degree of bias. Harvard emphasizes the full class model because it shows a smaller degree of bias.
The usual meaning in this context is 5% statistical significance. The Plantiff’s review suggest Harvard’s expert’s model would meet this threshold with minor modifications, such as any of the following. All of these differences in their respective models are points of dispute.
–Do not consider personal rating
–Do not consider parents’ occupation and look at only unhooked subgroup
–Do not consider parents’ occupation, add interaction variable between “disadvantaged” flag and race, and pool applicants from all available years for larger sample size
–Add interaction variable between “disadvantaged” and race, pool applicants, and look at unhooked
Thank you @Data10 for your astute analysis and for pointing out that the plaintiffs had the entire set of data as well. Since most of the bias claims seem to revolve around personal ratings and standardized test scores between different racial groups, what would be considered an acceptable gap if any? I would expect the personal ratings to fluctuate from year to year and be pretty close between all races unless their was some bias shown (that seems to be the case). If a student is in the 95th percentile nationally for the particular standardized test and the rest of their profile is very good, is that good enough? What particular percentile is too low? Most of the issues I have with racial preferences revolve around not hearing about Asian-Americans reaching elite schools with lower end (25th percentile or lower) standardized test scores (although one student who testified was one such student).
My biggest concern with this and similar lawsuits is the shallowness of understanding of college admissions on the part of those bringing such lawsuits, as well as on the part of some cheering from the sidelines.
Not many of the truly outstanding among my Asian clients apply to Harvard; it is much more common for under-qualified students to apply, actually. However, I did have one well-qualified applicant about 6 years ago who applied. He had very much going for him, but what people fail to understand is that it’s not just about “well-qualified” or even “outstanding.” It’s also about Yield. Historically, and that is equally true today, Harvard has protected its Yield more vigorously than Tufts University or any self-protective LAC with “Ivy resentment.” H wants to be sure the student will enroll, which is probably why they invited Diep, and which is hardly the first time I have seen this. (And if so, that fact will not necessarily appear in Diep’s or any other such student’s file, regardless of how pleading or bitter complainants “demand” that the committee make physical notes about such facts.)
Despite the supposed claims in one of the citations in an earlier post, H does in fact seek out low-SES valedictorians from public high schools – vals for whom a H offer is the best opportunity of any kind they have received so far in their lives and could be their ticket to a far more privileged life. They are not as interested in offering high-SES vals admission because they most often know or assume, rightly so, other offers from similar elites, and H hates competing on that level.
H is also often hesitant to admit students --Asian or non-Asian-- who are likely admits to STEM-focused schools such as MIT and CalTech when it is quite obvious that such students are likely admits there. This is what happened to the student I mentioned above. His profile was decidedly MIT/CalTech, so even though his first choice school was H, and his personal statement explained that, H was taking no chances on being turned down.
In the past, yield was a factor in rankings. And even though by 2004 it had been officially eliminated from the USNWR list of factors, internally Yield can be an important business factor for U.S. colleges, just as ED is important in the business model.
I could discuss other factors, but I have to go to work.
I don’t know anything about the plaintiff organization, but I would not underestimate the understanding, experience, intellect and talent of the attorneys representing both sides in this obviously big case.
@epiphany I believe that the elite schools definitely want to protect their yields, but the lowest yields by race are from African Americans (at least at Harvard). If yields were the main reason for rejecting very qualified candidates who may not attend, it would stand to reason that they would admit less African Americans (or find ones more likely to attend) because of that same reasoning, but the percentages of African American admits have grown over the last 20 years. That is a signal to me that the yields are not as important to Harvard as their diversity numbers. The plaintiffs are not concerned with the diversity goals of Harvard and believe they will get more seats at Harvard for Asian Americans (which looks to be the probable result based on statistical analysis if race is removed as a preference in admissions).
Doesn’t affirmative action and URMs hurt yield? URMs have the lowest yield %. The problem is that any decent URM applicant has a good shot at every top school because of affirmative action. That’s why every kid that makes the news for getting into all Ivy League colleges is URM.
Yes, in the final year range listed in the lawsuit, the yields by race among high rated admits were:
Asian – 84%
White – 65%
Hispanic --55%
Black – 51%
However, the same document also mentions that the yield rate among early applicants (average among all races) was nearly 95%. Since then, Harvard has notably increased the percentage of students admitted early. In more recent classes, the majority of matriculating students were admitted during the early action round. I expect yield considerations relate to this increase in portion of students admitted early. Yield bragging rights may also be a consideration, as Harvard does not want to appear to be a 2nd choice to Stanford’s yield. Restrictive early action is also a way for students to indicate Harvard is a first choice.
Being an early action applicant was associated with a significant boost of chance of admission in the lawsuit, after all controls, meaning students with similar hook status, scores, and reader ratings were more likely to be admitted if the applied REA than RD. In the Plantiff’s model degree of this boost was ~80% the magnitude of the boost for being Hispanic and several times stronger than the estimated penalty for Asian.
There are many reasons students may be rejected. How do you know it was due to yield protection? If Harvard was big on yield protection beyond favoring REA, then I’d expect to see an odd relationship between various stats and chance of admission. For example, we might see that applicants with high scores had a better chance of admission than applicants with near perfect scores. However, the previously linked acceptance rate by SAT score showed the opposite effect with the chance of admission getting increasingly steep as score increased, up to and including getting a perfect 1600/36. The effect was even more extreme looking at AI (combination of GPA, SAT and SAT II) instead of SAT score. The closer stats were to perfect, the higher the admit rate. There also does not appear to be any indication of not favoring the highest rated applicants on the holistic components of the application, after controlling for hooks. Instead being one of the few applicants who received the maximum rating of 1 instead of a 2 was associated with a huge increase in chance of admissions, such that the majority with a 1 in any of the 4 major rating categories were admitted.
Harvard may or may not be concerned about losing STEM admittees to MIT/Caltech, but Fitzsimmons’ testimony does signal a bias against STEM applicants at Harvard. That bias could explain a large part of the perceived bias against Asian American applicants (particularly Asian American male applicants) since majority of them are STEM applicants.
I believe the quote posted in news stories was they were wary of admitting too many engineering and CS majors because a “whole bunch” would end up “will end up happily ever after at M.I.T. or Caltech.” The number of CS concentrators at Harvard in recent years is below. Nearly quintupling in the past 8 years is an incredible rate of increase. At the current rate of increase, CS will overtake economics to become Harvard’s most popular concentration next year. It’s likely that Harvard has some concerns about the rate of increase and wants to avoid becoming Harvard Institute of Technology, where the overwhelming majority major in tech. I’d want to see the full quote in context before assuming wanting to avoid admitting too many prospective CS majors is primarily an issue about yield. The “because” wasn’t part of Fitzsimmon’s quote. He may have said that they considered prospective concentration with the high rate of CS increase, and a whole bunch of rejected prospective CS majors would end up “happily ever after” at great Institute of Technology schools, like MIT or Caltech, without indicating yield protection.
According to IPEDS, Asian students are overrepresnted among CS majors, and all other races are underrepresented, so limiting prospective CS majors would impact Asian applicants far more than any other race. I listed the percentages below. These use IPEDS racial definitions, comparing the most recent graduating class to the entering class 4 years ago.
Asian – 40% of CS majors, 19% of entering class
White – 35% of CS majors, 43% of entering class
Hispanic – 8% of CS majors, 10% of entering class
Black – 3% of CS majors, 7% of entering class
However, based on the rate of increase in CS majors, I doubt that this limitation was notable during the lawsuit analyzed years, which is consistent with the conclusion from the lawsuit analysis. It found that the regression coefficient for a CS concentration was higher than all other concentrations, but did not reach statistical significance due to the high SE. There may have been a small boost for prospective CS majors during the lawsuit years, but nothing to indicate a penalty.
What is interesting about this is that popular STEM majors at Harvard (e.g. computer science) do not specify grade or GPA requirements to declare the major, suggesting that their departments have plenty of capacity, unlike the departments of economics and visual & environmental studies, which do have grade or GPA requirements higher than C or 2.0 to declare the major, presumably to keep enrollment within departmental capacity.
Unlike the economics majors, most declared STEM majors at Harvard could probably meet ANY grade/GPA requirements Harvard would set, so this type of requirements wouldn’t accomplish the goal of limiting the number of students in those majors.
I’m not talking about Asians competing with URM’s. The Yield concern exists within the White and Asian segments, which compete internally and with each other.
Based on the grade distribution for Harvard CS 50 at https://cs50.harvard.edu/2018/fall/faqs/#what-is-cs50s-grade-distribution , setting a minimum grade of A- in CS 50 to major in CS would weed out about 30% of the students, while setting a minimum grade of A would weed out about 62% of the students (based on percentage of letter grades that are that grade or higher – presuming that SAT/UNS grades are non-majors).
So yes, Harvard could limit enrollment in the CS major with grade or GPA minimums if it felt that that was necessary to avoid overflowing the department’s capacity.
I should have said that AVERAGE grade or GPA won’t do the trick. Would it do the trick if it’s based on the grade of a SINGLE course? Of course, but it wouldn’t be fair to the students.
Even when using a GPA over several courses, they could weed out more aspiring CS majors by setting a higher GPA requirement, if that is necessary to keep enrollment within department capacity.
I don’t think the economics GPA requirement was added to limit enrollment under a departmental capacity. Instead I expect it is more to flag the small minority who are not sufficiently knowledgeable and get them to take Harvard’s tutorial review class. After they take the review tutorial, they can declare the concentration, even if they initially had a low grade.
The number of economics concentrators decreased significantly during the few years prior to the GPA requirement, then increased significantly since adding the GPA requirement. I expect the number of economics concentrators is more correlated with stock market indexes and popularity/pay of Wall Street jobs. If the stock market and Wall Street becomes more popular in comparison to CS and Silicon Valley, then I’d expect the number of economics concentrators to go substantially higher.
That’s not necessarily true. Some colleges, such as MIT or Stanford, offer early applicants little advantage, unlike Harvard. He could choose to SCEA to Harvard and then RD to MIT and Stanford. If he wanted to major in CS, he likely preferred the other two schools.
Thank you for your responses, @suzyQ7 and @1NJParent . He applied Early Action to H and RD to MIT, CalTech, and Stanford. He was admitted to MIT and CT but accepted his offer to Stanford in CS, yes. But at the time of application to H he was being guided emotionally more than intellectually, which is not surprising and which often happens. The admissions committees often know better than the student knows where his profile indicates he belongs.
Any more questions about this particular student, you can PM me.
