"Race" in College Applications FAQ & Discussion 12

Applying to College
3701

The personal qualities rating is supposed to assess “applicant’s humor, sensitivity, grit, leadership, integrity, helpfulness, courage, kindness and many other qualities.” The academic decile index is a mathematical function of SAT and GPA/rank stats. These are obviously not the same metric. Asian applicants as a whole have higher AI SAT + GPA AI stats than other races, but don’t have a corresponding higher average personal qualities rating. However, within any race including Asian, the applicants with the higher stats were more likely to have a high personal qualities rating than applicants with lower stats, but it was far from a perfect correlation. Specific numbers are below for unhooked.

Percent of Applicants with 1-2 Personal Qualities Rating: Admission Readers
White – 21%
Asian – 18%
Hispanic – 19%
Black – 19%

Percent of Applicants with 1-2 Personal Qualities Rating: Alumni Interviewers
White – 50%
Asian – 50%
Hispanic – 41%
Black – 43%

I agree that the percentages look a little suspicious, but it’s still a relatively small effect among Asian applicants. The Plantiff 's analysis predicted that removing personal rating only would increase Asian admission share by ~1%… I think the greater impact is the athletic rating. Among unhooked non-recruited athlete applicants, the following percentage received a 1-2, which is associated with an increased chance of admission.

**Percent of Unhooked Applicants with a 1-2 Athletic Rating **
White – 13%
Asian – 5%
Hispanic – 8%
Black – 7%

As has been said repeatedly, putting a greater weighting on stats and lower SES is not the same as " fixing an anti-Asian policy." At the time, Berkeley had a high weighting on stats compared to most other selective colleges. Average stats are below for new freshman. I’m not sure what you are expecting see, but it certainly doesn’t suggest “fixing an anti-Asian policy”. Note that I am counting Filipino as Asian.

(year before new policy): White SAT = 1262 and GPA = 3.85, Asian SAT = 1275 and GPA = 3.95
(year after new policy): White SAT = 1261 and GPA = 3.85, Asian SAT = 1274 and GPA = 3.95

The bigger changes occurred among Hispanics, which fits the Hispanic historic trend changing from a sharp increase with the rapidly increasing Hispanic (and Asian) CA HS senior population to a sharp decrease upon the change in admission policy…

(year before new policy): Chicano SAT = 996 and GPA = 3.48, Latino SAT = 1084 and GPA = 3.55
(year after new policy): Chicano SAT = 998 and GPA = 3.60, Latino SAT = 1101 and GPA = 3.67

3702

That’s not what the Arcidiacono analysis you’ve been posting direct quotes from found, . I’ve posted specific numbers from that analysis earlier in this thread and won’t repeat them. One hint is you should consider the relative portion of class that is unhooked for Asian and White, rather than assuming the number all hooked spots will increase linearly, if URM preference is removed.

3703

A while back, I read somewhere that black student enrollment at Harvard has been remarkably consistent right around 170 students each year in recent years, whereas it has varied for other groups. If that is correct, that is very unlikely to happen by random chance, suggesting a hard quota rather than a soft one.

3704

Asians are ranked substantially lower on personality ratings by Harvard’s AO than whites in the same academic decile.

whites
academic decile 10 - 29% (occurrence of 2 or better)
9 - 27%
8 - 25%

asians
academic decile 10 - 21% (occurance of 2 or better)
9 - 20%
8 - 17%

Despite the fact that (for each racial group, higher academic index deciles are associated with better personal ratings; and Asian-American applicants have the highest academic indexes,
Asian-American applicants have the lowest shares receiving a 2 or better on Harvard’s personal rating of the four main racial groups

3705

@hebegebe This is in the CDS. I hope I copied them over correctly.

2017-18, 160 first years, 523 total
2016-17, 132 first years, 473 total
2015-16, 109 first year, 406 total
2014-15, 118 first year,s 463 total
2013-14, 116 first years, 443 total
2012-13, 107 first years, 436 total
2011-12, 116 first years, 453 total
2010-11, 99 first years, 457 total

3706

Harvard’s yield is quite a bit lower on Black admits than non-URMs, which makes predicting enrollment difficult. I expect hebegebe is referring to is the lawsuit finding about the single race black admit rate for all practical purposes being identical to the non-black domestic admit rate for the classes of 2017-19. This works out to ~170 black single race admits per year. Arcidiacono calculated a 0.2% chance of the nearly matching admit rate happening randomly during the specified years. Card disputes this calculation, as I recall, due to the large number of possible year-variable combinations. . If the single race black admit rate is held constant with the overall domestic admit rate, then this results in a larger number of black admits with Harvard’s policy of increasing recruiting efforts towards not likely admission Black HS students, which began in class of ~2011 and appears to have continued since then.

3707

Well, so much for that theory. I will try to find where I read that.

3708

I didn’t see your reply before I posted my response yesterday, @Data10, so thanks for that information. Apparently I confused admits with enrollments.

For anyone used to statistics, this is a pretty obvious “smoking gun” that demonstrates that Harvard has an illegal hard quota on the number of black admits. There are two reasons why, that when combined together, make it difficult to reach any other conclusion.

The first reason is that, even without any preferences, a nearly constant number of admits for a certain group just doesn’t happen by chance very often. I haven’t done the math, but am not surprised by the 0.2% that Arcidiacono came up with.

The second reason is that Harvard’s OIR model showed that it wouldn’t get to around 170 black admits without a pretty significant preference. But preferences alone would also get to a somewhat random number each year.

3709

The exact numbers are below. The predicted is the rate of admits that would occur if the rate of single race Black admit rate was set to the non-Black domestic admit rate (IPEDS Black applicants * IPEDS non-black domestic admit rate) . Black is defined using IPEDS/federal racial definitions, which differs from the racial definitions used on Harvard’s website, so numbers do not match Harvard’s website.

Class of 2017 – 2788 IPEDS Black applicants: Predicted 172.7 admits, Actual was 172 admits
Class of 2018 – 2688 IPEDS Black applicants: Predicted 175.3 admits, Actual was 177 admits
Class of 2019 – 2905 IPEDS Black applicants: Predicted 176.7 admits, Actual was 176 admits

Card’s claim is that there are many possible combinations of race and years, increasing the chance to 17% that any one of these many combinations would have this degree of similarity randomly. As usual, I think both sides present the stats is a way that exaggerates their position. Yes, there are a lot of other possible year and racial combinations that should be considered when calculating chance that of one of the many possible combinations would being similar. However, most of the combinations Card includes are not meaningful and Harvard would have little reason to manipulate. If you limit it to URM combinations that have a significant portion of students (I am considering at least 1% of students to be significant) and limit years to the one in which the corresponding racial definitions were used, then I get a ~3% chance of occurring randomly using Card’s methodology . I think a better methodology would be to consider the historical average and SD and use that to calculate chance of similarity.

3710

and as far as Berkeley what’s needed is white and asian application numbers and acceptances before and after along with a link to the data. otherwise showing your data is irrelevant

3711

When you weren’t satisfied with the admission share results, you specifically asked for GPA and SAT scores multiple times. Now that the GPA and SAT scores didn’t support your view, you have been asking for application numbers and acceptances multiple times.

Unfortunately I don’t have an apples to apples comparison in the specified years. I only have a 1987 report and the standard federal reporting which began in the mid 90s. The 1987 report only provides the total – new freshman + advanced standing. It doesn’t separate the two, so I only list new freshmen admit rate for 1995.

Before (1987): Total – White 3,555/11,286 = 31% admit rate: Asian 1,375/5,032 = 27% admit rate
After (1995): Total – White: 3.748/10,310 = 36% admit rate, Asian 3,505/10,388 = 34% admit rate

After (1995): New Freshman – White: 2,921/7,728 = 38% admit rate, Asian 2,879/8,395 = 34% admit rate
After (1995): Advanced Standing – White: 827/2,582 = 32% admit rate, Asian 624/1,993 = 31% admit rate

Note that the number of Asian applicants roughly doubled from ~5k to ~10k in this period, while the number of White applicants declined. This rapidly increasing number of Asian applications is the primary reason for the much larger portion of Asian students at Berkeley during this period. A secondary, but less influential factor is the smaller gap between the White and Asian acceptance rate. This smaller gap is expected with a modified admission policy that primarily auto admits non-URM, not low SES applicants by highest stats, without considering the rest of the application (see earlier post). That’s not a misprint about the acceptance rate going up. The overall acceptance rate across all races also increased during this period, as listed in the linked document.

The 1995 report is at https://academic-senate.berkeley.edu/sites/default/files/aepe_2002_report.pdf . I’d prefer not link to the 1987 report and have a huge series of misinterpreted quote posts from a 400 page document.

3712

@collegemomjam Those numbers for African Americans who flow through HBCUs (undergrad or grad school) still surprises me, especially with most HBCUs populating a specific portion of the US (Southeast and East Coast). One thing that the data shows is the importance to African American students in STEM fields and African American students who shoot for certain terminal degrees.

My family could move on from HBCUs and just focus on USNWR rankings to pick a school and I have definitely tried to show my kids what the elite schools have to offer, but they see the love that my wife and I have for our HBCUs, and see our friends from undergrad living fantastic lives (Son’s godfather “retired” at 39 and travels the world doing what he wants, is one such example). My kids have been asked a few times why they would even consider HBCUs with their “stats” and I think it comes back to the outcomes that they have seen. They have talked to some of our family friends who have gone to some of the most amazing grad schools (Harvard, MIT, Duke, Vanderbilt, Emory, Tulane, etc.) and know that they all started from the 3 small HBCUs that currently make up the Atlanta University Center (AUC).

@ucbalumnus I am not arguing that calling HBCUs “self-segregation” is wrong, I am saying is that I don’t ever hear of White students at school such as Auburn (85% White) choosing that school due to “self segregation”. I just hate that term because I will see commentary that says that white students like to to be with like students, but it is never described as “self segregation” for an entire college campus. The only time that I hear about self segregation when it comes to race at PWI’s deals with minorities (Asian students, African American students, and Hispanic students) grouping off with their racial group on a PWI campus,

@SatchelSF The one part of the Harvard trial that I thought was proven by the plaintiff based on what I read was around possible illegal quotas.

@Data10 You are the CC Data Champion of the World.

3713

Note that in my previous post, I was writing that the fact that they are still 80+% black is evidence that non-black students are self-segregating away from black students, rather than black students self-segregating. (Obviously, the history of historically black schools was not that of self-segregation by black students, since it was imposed segregation back then, although one can argue that it was self-segregation by white students enforced on everyone.)

Even though it effectively is to some extent (though perhaps a better example would be UAH and AAMU, two public universities in the same metro area). But then is it any surprise that a term with negative connotations like “self-segregation” tends not to be used by commentators to describe people of their own group?

3714

@ucbalumnus Thank you. You even chose 2 schools that I know very well (my wife’s hometown institutions). Your last point is what I was trying to get across, but you said it much better than I.

3715

@ucbalumnus I agree that all races participate in self-segregation to some degree but you’re not making a good case. Almost all the HBCUs are low quality schools. Student wouldn’t schools in the HBCU tier unless they couldn’t get into other schools. The white student ‘in the market’ for a HBCU is choosing between a ranked 101 white majority school over a ranked 100 HBCU school. In-group preference doesn’t have to be high for it to be a tipping factor in this case.

The fact that white students don’t target or apply to HBCUs isn’t a smoking gun. It could just be that black people apply to HBCUs at a much higher rate than normal (which is true) rather than white people applying to HBCUs at a lower rate than normal.

As ChangeTheGame mentioned, apparently smart African Americans (who get almost automatic admissions into top schools with the combination of affirmative action + decent academics) are declining attractive schools for HBCUs. That is a strong sign of self-segregation. They’re taking a ranked 100 school over a top 15 school to be with people of their race.

Your case would be better made with white flight in growing Asian communities. Top school districts in safe, green communities in the Bay Area and LA are attractive places that are losing white people. White people are presumably forgoing these amenities to live in communities with other white people. Same for UCLA and Berkeley

3716

I’m not caught up on the recent discussion on the case.

To be clear, Arcidiacono’s model is objectively better.

  1. Hooked and unhooked applicants are obviously evaluated differently, so all hooked (except for race hook) applicants should be ignored
  2. Harvard has an academic rating floor of 3 (but most acceptances have ratings of 2+, with many suspicious hooked applicants who get in with a 3). All applicants below this floor add noise to the data because these applicants are treated as if they aren't even in the applicant pool by Harvard admissions.

All analysis should use data with unhooked applicants with academic ratings of 3+ (ideally 2+).

3717

@UndeservingURM When it comes to College rankings the one thing that makes it hard to compare elite PWIs and HBCUs are the endowment levels. If you take a look at the USNWR top 50 and compare it to the schools with the largest endowment levels in America, the 2 lists would have a lot of overlap. Having a lot of money makes having a high USNWR ranking likely. It is amazing that HBCUs are ranked at all with the size of their endowments. My daughter’s best friend (at a top 3 USNWR school) and my daughter were both taking a Chem class 1st semester and the difference in her honors Chem class and that student’s was negligible from a lecture perspective, but where they see the difference is with the resources provided from a lab perspective. That may not be the case with all classes and majors, but I am confident that my daughter will get what she needs to be successful for graduate school one day (but she will need to do research every year at schools with the latest cutting edge equipment to close the resources gap).

3718

That’s the portion they want you to focus on.

…because you are then arguing about the seats left over after athlete, legacy, development etc have played a significant role in the admission of roughly half the class.

“Arguing over crumbs” would definitely be putting it too strongly, yet that’s the phrase that comes to mind.

3719

“It builds in a cultural bias tilted to those who buy into the testing (and retesting) mentality. So the heavily prepped repeaters are not necessarily more talented in math”

The students getting into Cal Tech (if I may generalize from the bay area) are not testing repeatedly, in fact they usually just take it once, maybe twice. And they do not prep for the math portion of the SAT or any of the subject tests. They do a practice test and that’s just to get a feel for the format. They do prep more for the verbal and essay parts of the test.

As others have said the SAT and ACT cannot differentiate between the good math and the really talented in math, that’s why they encourage kids taking the AIME and participating in AMO.

And yes, the Cal Tech students are more talented in math, no question, maybe the most talented in the country. Only MIT, Harvard, Standford, maybe Chicago can compete with Cal Tech students. I thought I was good in math growing up, these kids are on another level.

3720

I know plenty of kids who scored 1500+ on the SAT prior to 13 years old. Among math competition kids, it is not uncommon to see 800 SAT math level 2 and 5s in AP calculus in elementary school.

The idea of test prep is a total red herring for the top kids. Take a look at the AIME questions, and try to solve some. Then look at who gets the best scores. In recent years, it is usually the 6th grade and under group. Just how much time do you think they have had to prep?

Caltech gets probably the highest average math group in the country, but the very top kids are at MIT, and then Harvard in my opinion.