I wouldn’t expect a GC to dissuade a student from applying based on that difference, which was the suggestion I responded to, but YMMV I suppose.
MODERATOR’S NOTE:
Let’s move on from parsing the difference between 2.6% or 2.8% or 4.6%
@Data10 @MWolf
Thank you for the explanations. I somehow (naively) thought “yield protection” is below the perennial number 1/2 college of the nation. ?
I respect MIT/Mudd/CalTech more now! ?
In the case of a non-ALDC applying to Harvard, in order to be in the academic decile with at least a 5% chance of admission, a white(non-Hispanic), African American, Hispanic, or Asian student would need to be in the 8th decile(top 30%), 3rd decile(top 80%), 4th decile(top 70%), or 8th decile(top 30%), respectively. So the admit rate for non-ALDC white(non-Hispanic) or Asians is very low if they’re not at least in the top 30% of applicants, academically.
The 8th decile for Harvard’ Academic Index is a 229.9+. Note this is based on previous admissions cycles and higher scores may be necessary to be competitive in the current admissions cycles, as apps increase.
On the GPA conversion to an 80 point scale:
80 - 4.0(unweighted)
79 - 3.91-3.99
78 - 3.81-3.9
77 - 3.72-3.8
So the Ivy League academic index is heavily test score biased compared to rank or GPA.
40% two best different SAT subject scores
20% best SAT or ACT math
20% best SAT or ACT RW
20% rank or GPA
But compression of the score for rank or GPA may mean that its importance is even less. The same may apply for the SAT subject test for math level 2.
Mudd has ED, which is a much stronger form of yield protection. Only MIT/Caltech have totally unrestricted EA among privates.
The difference between EA and RA aggregate acceptance rates aren’t as dramatic at Stanford, but the data aren’t as granular as in the Harvard case. Is protecting yield less of a concern at Stanford?
This AI score is used to determine minimum stat requirements for athletes, as specified in Ivy League athletic conference rules. I would not assume colleges emphasize this AI score in admission decisions for non-athletes.
The Ivy League conference academic index is calculated as 1/3 Converted GPA, 1/3 SAT/ACT, and 1/3 SAT II. They favored rank over GPA at the time of the lawsuit, but this has changed with fewer HSs providing rank. The converted GPA scale is 98%+ = 80, 4.3+ Weighted = 80, 4.0+ Unweighted = 80, … ; 90% / 3.33 UW = 69; 80% / 2.33 UW = 53; … They also have conversions for 5.0 scale, 6.0 scale, other countries, etc. For the class of 2019, the components were:
Average CGPA = 77.0
Average SAT/ACT = 74.8 (2244 SAT / 33.1 ACT)
Average SAT II = 76.2 (762)
Whether they emphasize it or not, it has useful predictive value for non-athletes. In the Arcidiacono paper, non-ALDC whites in the top decile had an admit rate of 15.27%, while the corresponding fifth decile had an admit rate of 2.57%. Non-ALDC Asians in the top decile had an admit rate of 12.69%, while the corresponding fifth decile had an admit rate of 1.86%. The AI, and for that matter the academic rating, is less useful for predicting whether ALDCs or URMs are admitted.
The reader guidelines for academic rating are largely specified in grade + score stats, so it makes since that a measure of grade + scores would have a good correlation with academic rating, such as the academic index (AI) calculation. AI alone explains 70% of variance in academic rating – the clear majority. As such, AI is expected to be correlated with admissions decisions, like academic rating is correlated with admissions decisions. There should be a correlation for all groups – ALDC, URM, non-ALDC, non-URM, etc. An example showing admit rate by academic rating for different ALDC categories is below.
Non-ALDC Admit Rate by Academic Rating
Academic Rating = 1: 68% admit rate
Academic Rating = 2: 10% admit rate
Academic Rating = 3: 2% admit rate
Academic Rating = 4: 0.0% admit rate
Academic Rating = 5: 0.0% admit rate
LDC Admit Rate by Academic Rating
Academic Rating = 1: 97% admit rate
Academic Rating = 2: 49% admit rate
Academic Rating = 3: 18% admit rate
Academic Rating = 4: 3.5% admit rate
Academic Rating = 5: 0.0% admit rate
Athlete Admit Rate by Academic Rating
Academic Rating = 1: 100% admit rate*
Academic Rating = 2: 96% admit rate
Academic Rating = 3: 87% admit rate
Academic Rating = 4: 80% admit rate
Academic Rating = 5: 50% admit rate*
*very small sample
However, AI appears to add little the prediction for the other non-academic administration ratings beyond controls. The alumni interview ratings were more interesting, with a significant positive correlation between AI and alumni overall rating and a significant negative correlation between alumni personal rating, after controls. Both of these alumni interview correlations were far weaker than the academic rating. Specific regression coefficients for AI contribution to the subratings are below, after all controls. They show how much influence a large 1 standard deviation change in academic index has on the subrating.
Regression Coefficient for 1 SD change on Academic Index
Academic Rating: +3.75 (explains 70% of variance in academic rating)
EC Rating: +0.1
Personal Rating: -0.15
LOR #1 Rating: +0.15
LOR #2 Rating: +0.15
GC Rec Rating: +0.05
Alumni Personal Rating: -0.4
Alumni Overall Rating: +0.75
It’s important to subdivide this by race. If I did my math right, the non-ALDC admission probabilities are as follows:
Academic Rating - 2 or better
white(non-Hispanic) - 9.6%
African American - 49.0%
Hispanic - 24%
Asian - 8%
Academic Rating - 3
white(non-Hispanic) - 1.2%
African American - 7.7%
Hispanic - 4.5%
Asians - less than 1%
Something interesting is that even among the non-ALDC admitted whites, 21% had an athletic rating of 2 or higher. This might be a top athlete in a sport that Harvard doesn’t have a team in, or it could be a potential walk-on to a varsity team with at least some support from a coach. ** If you treated these potential walk-ons as athletes, then only 44% of white admits are not ALDCs or unhooked. **
My back of the envelope calculation suggests about 68% of Harvard’s class falls into one or more of the following categories: URM, international, ALDCs, or athletic tip(having an athletic rating of at least two, but not being recruited).
^If you add disadvantaged to the mix, about three in four Harvard admits will be international or hooked to some degree.
H says:
https://www.politico.com/f/?id=00000166-9690-d166-a77e-9f9c92f10001
Honestly that could be my kid or hundreds of others at her smallish public high school. Mine was a two sport athlete, varsity captain of one.
I have to think 2s and 3s are pretty common. @roethlisburger do you have the % of students (applied or accepted?) that fall into which rating?
I think being international is an anti-hook, a negative. What makes you think otherwise?
The new 2023 reader guidelines go in to more detail about the athletic categories as quoted below. 2’s do include the comment “possible walk-on to a varsity team.” However, I wouldn’t consider this a hooked group, just like I wouldn’t consider a students with really impressive non-academic ECs/awards as a hooked group.
Harvard Reader Guidelines: Athletic
The percentages in the different categories for the lawsuit period are below:
Percentage Athletic Ratings
1 – <1% of applicants, 11% of admits
2 – 9.5% of applicants, 16% of admits
3 – 50% of applicants, 41% of admits
4+ – 39% of applicants, 32% of admits
Percent of applicants who list varsity athletics as “primary EC” – 23%.
Percent of applicants who list JV athletics as “primary EC” – 11%.
The athletic rating is interesting in that it is the only rating where there does not appear to be a major penalty for getting a 4. Nearly 100% of non-ALDC applicants who receive a 4 in any other rating subcategory besides athletics are rejected;, yet a large portion of the entering class has a 4 athletic rating. The regression model found little difference in admission chance between getting a 3 or getting 4, with controls for rest of application and characteristics.
To the extent applicants are competing for slots in separate buckets in the admissions process, US citizens and internationals aren’t competing against each other for the same spots.
Right. And the international “quota”, if you will, seems to be about 10% of the class. So a much smaller bucket.