^and the 2018 SAT/ACT concordance table will also provide clues about how far off the 2016 tables are. But, it would probably be too embarrassing to provide an updated Old/New set of tables; fortunately for CB, they will be irrelevant soon, albeit not quite as soon as we’d like.
I don’t have any further info relative to the original question I posed on this thread (still waiting for Princeton’s 2017-2018 CDS, for example), but thought that I’d update my comments in #136 where I said that if you compare the 16 Concorded score to the 17 score, it’s a decline of (37) for the 25th percentile, (33) for the 50th percentile, (28) for the 75th. I realized that there was one school which was fairly dramatically bringing the scores down, making the differences greater than they probably should be for what is, at this point, only 18 moderately to very selective schools. (About 3 more than a week and a half ago.)
The median in this case would be a much better approach, and produces a decline of (15) for the 25th percentile, (16) for the 50th percentile, (20) for the 75th. Still significant, but not quite as dramatic. Same caveats regarding the value of this analysis still apply.
Now I’m on the hunt for trying to figure out how total SAT scores at the 75th percentile compare to the sum of SAT section scores at the 75th percentile. (The total should be slightly less than the sum of the individual sections, just trying to figure out how much less… maybe only 10 points.)
Resurfacing for a couple updates:
- A total of 33 of the 36 colleges/universities I was tracking have posted their 2017-2018 CDS in an easily Google-able place. [Gives the side-eye to Grinnell, MIT, and Caltech.] Even with the addition of 15 more schools to the 18 I reported on in post #141, the median decline from the expected SAT by concording the 2016-2017 SAT score to the actual SAT score was essentially the same: a decline of (15) for both the 25th and 50th percentile, and a decline of (20) for the 75th percentile. The raw growth from 2016-2017 to 2017-2018 is significant on the low end – a median increase of 50 points at the 25th percentile and 35 points at the 50th percentile – but unsurprisingly more muted near the top, just 10 points at the 75th percentile.
- Thanks to @evergreen5 for pointing out at some point recently the Princeton 2017-2018 CDS which does exist on Princeton’s website (https://registrar.princeton.edu/university_enrollment_sta/common_cds_2017-2018.pdf), confusingly, isn’t actually posted on Princeton’s CDS page at the moment. (See here for the Princeton score bands: https://admission.princeton.edu/how-apply/admission-statistics) We can now evaluate how much of a bump – or demerit – an applicant might receive based on how their score compares to the median.
Acceptance rate: 6.4%
25-75 percentiles (enrolled students):
- SAT Math: 720-790
- SAT EBRW: 710-780
- ACT Composite 31-35
Acceptance rates at 75% level (1570/35):
- SAT Combined [1500-1600]: 8.2%, but I’m making the assumption that those at 1570 are accepted at nearly 10% (which might make those at 1500 6%).
- ACT Composite [32-36]: 7.8%, but I’m making the assumption that those at 35 are accepted at maybe 9%?
So… maybe a 1.5x acceptance bump? But because of where the numbers fall, take that estimate with a slightly larger grain of salt.
Acceptance rates at 50% level (1500/33):
- SAT Combined – average of 1500-1600 (8.2%) and 1380-1490 (5.0%): 6.6%
- ACT Composite [32-36]: 7.8%, but under the logic of adding 1.2% for a 35, I’m subtracting (1.2)% for a 33%, yielding 6.6%.
So there’s virtually no bump at being at roughly the median.
Acceptance rates at 25% level (1430/31):
- SAT Combined [1380-1490]: 5.0%
- ACT Composite – average of 32-36 (7.8%) and 27-31 (5.3%), tweaked down slightly: maybe something close to 6%? There is a little difference, it would appear, between the 50% and 25% acceptance bands, but not much. Once you fall below that 25% level, it seems to drop off more, but this is a relatively flat acceptance range.
Acceptance rates at 25% and below are hard to calculate because we don’t know the relative number of students in each SAT and ACT band, but given that the bands up to 1370 are at no more than 3.4% and up to 31 are 5.3%, there is more of a dropoff, probably at least a third (in other words, a 4.3% chance) at numbers just below the 25th percentile and more of a dropoff below that.
This data does lend further credence to the idea that there is some (albeit limited) value to being a “star student” relative to the application pool, even at the most selective schools, but below that, it doesn’t matter as much whether the student is at the 55th or the 35th percentile – they’re clearly qualified academically and the school can do a more holistic review. And really, the difference between, say, a 1450 SAT and a 1510 SAT is probably minimal.
@BorgityBorg Actually you cannot add the individual section percentages to get the combined 75th, 50th and 25th percentiles for SAT purposes. There are going to be a lot of students who do much better in one section over another, e.g. a kid may be in the 75th percentile in math but the 25th percentile in EBRW. The combined range is likely to be lower. Also, when comparing likelihood of acceptances starting with such a low base acceptance rate, we should be comparing percentages by rates vs gross differences. In other words, it’s not just the difference between 8.2% vs 6.4% (1.8%) vs 5% (1.4%) but that students in the 1500-1600 range were accepted at 64% higher rate than those in the 1380-1490 range, and a 28% higher rate than the average. That is a pretty significant advantage when you look at it that ways. Brown used to publish a breakdown similar to Princeton, but with narrower ranges (can’t find it now), which indicated fairly significant bumps as you moved up the bands when you look at the change in rates of acceptances.
^I really wish more schools published middle 50 SAT composites in addition to the section ranges, to perhaps have a little more info for the high/low student. I wonder why the CDS does not ask for SAT composites, unlike for ACT.
@BKSquared I agree that it makes logical sense that the combined scores at the 75th percentile are less than the sum of the 75th percentile of the math and the 75th percentile of the EBRW. But I have seen very little actual data in that regard, and in fact, my vague memory of the one data point that I saw was an implication that it’s actually a lot less of difference than one might expect. Like maybe only 10 points (e.g., 750 + 750 at the 75th percentile vs. 1490 at the 75th percentile), but in any case not significant enough to affect students’ decisions. FWIW, Princeton’s 75th composite ACT of 35 is equal to the average of their Math/Science ACT scores.
Anyway, see post #141 for my desire for more data on this point.
As for the rates v. gross differences, I agree completely. Look at #143, where I estimate a 1.5x bump for people at the 75th percentile – that was not 1.5 percent, but literally a 50% bump. That percentage difference seems to generally apply for the couple other schools I looked at earlier in the thread – Amherst and Hamilton. That’s what I think is important, and what I’m curious as to whether that applies more broadly. If a student was looking at a school with a 33% acceptance rate and they were at the 75th percentile, would that mean that they broadly (all else being equal, etc etc) had a 50% chance (33% x 1.5) of acceptance? That’s what I’m theorizing – I’d love to have more data to prove me wrong (or right).
^ I meant to say average Math/English ACT scores, not Math/Science.
Also, I think the issue is especially important for those tippy-top schools. A student sees a 6.4% acceptance rate but thinks, “hey, my score is at the 75th percentile!, so I’ve got a really good shot!” At least with MIT and Princeton, they can see data that says, yeah, they’ve still probably got at best a 10% shot at getting in. It’s way better than being below average, but still not very good. If they can apply the 1.5x factor to other similarly competitive schools that don’t publish the acceptance rate bands by score, then they might approach their chances more realistically.
^ Yes, totally agree with that. Applying to any single digit acceptance rate school is dicey without a major hook. I have always been a big fan of “testing the waters” by applying to a competitive early action school and competitive honors programs for state schools with rolling admissions to get a gauge of the strength of your application. It is easy to compare across statistical factors to see if you are in the ball park, but with low acceptance rates of high stat’s kids, there is little comfort since now you are being judged holistically outside of objective transparent standards. If you are getting good results from competitive EA and rolling admissions schools, it’s probably a good sign that your essays, LoR’s and EC’s are gaining traction in the eyes of AO’s. If you are getting mostly or all rejections, you either are shooting too high or you may need to revisit your essays and LoR’s. Deferrals from not as competitive honor programs may also indicate some weakness.