Sat plus gpa remains the most potent indicator of a student’s college success according to the UCLA study.
1 Like
Dude, I have taught stats at college level, so I am somewhat familiar with the concept of “statistical significance”. I am also very familiar with the fact that a value of 0.05 for p is arbitrary, and not at all accurate in many situations, especially when you have multiple tests. I also know that it is based on a long list of assumptions, which are often violated.
I also know that “important” in a statistical sense is not subjective. An r^2 value of 0.005 will be statistically significant if the sample size is 100,000. There is absolutely nothing subjective about saying that a variable which explains 0.5% of the variation in a dependant variable is not an important independent variable, even though there is a significant relationship.
2 Likes
So you’re saying that good academic performance in HS is the best indicator of academic performance in college?
One point I was making was why the HS name appears to be so predictive of first year GPA. I expect a key factor is course rigor / strength of schedule. As such, I expect that if you consider course rigor / strength of schedule, you’ll find it has significant correlation to first year GPA. Similarly if you look at raw GPA in isolation without considering rigor / strength of schedule, it will give the appearance of transcript being far less predictive than it actually is. This is what studies that compare SAT vs GPA typically do.
Along the same lines when you compare how much additional benefit SAT score adds, the benefit becomes far less when you consider how much it adds to course rigor / strength of schedule + grade distribution of particular HS + which course had lower grades and relevance to major + ECs / awards + LORs + … instead of just what SAT adds beyond HS GPA number in isolation.
Of course the primary goal of college admissions is to not create the class that has the highest possible first year GPA prior to effects of a curve. Colleges may favor students for a less resourced HS who overcame a challenging background, even if the student may be more likely to struggle during freshman year, until they catch up. Both SAT and HS name become less predictive, as you look at academic markers beyond freshman year.
Which UCLA study are you referring to? If you mean the UC senate task force study. it found that combination of SAT + HSGPA explained 23% of variance in freshmen grades at UCLA. That is indeed statistically significant. It didn’t focus on measures of success beyond freshman year. It also didn’t compare potency to anything else of significance beyond SAT and HSGPA.
A low r-square can be considered unimportant. A correlation can have low r-square and high t-stat at the same time. You are correct. Thank you for refreshing the knowledge.
Cornell to require testing for the HS Class of 2026:
10 Likes
Here is the important part:
“to enroll for fall 2025. This means that submitting test scores is not required but recommended”
1 Like
Note that Cornell’s announcement is two stage:
- For fall 2025 entry, frosh applicants to some divisions will be “recommended” to send SAT/ACT scores where sending them was previously “optional”.
- For fall 2026 entry, frosh applicants will be “required” to send SAT/ACT scores.
Probably has to do with timing – for fall 2025 entrants, some applicants may not have much availability of test dates in time for applications. But fall 2026 entrants have plenty of advance notice. I.e. Cornell may have thought about the timing issues more thoroughly than UT Austin did when it recently announced going back to SAT/ACT required and adding an EA round.
Of course, the conventional wisdom around here is that “recommended” should be interpreted as “required unless not possible” for typical applicants from relatively advantaged situations that are the primary demographic of these forums.
3 Likes
Agreed but CB seems to be adding a lot of sites. My recommendation would be to test early fall instead of going the TO route. More importantly, submit those mid/high 1400s if that is what you have.
1 Like
Data from Cornell that led to their decision:
The Use of Standardized Testing in Admissions: Summary of Key Findings
An excerpt: “The vast majority of students who have matriculated at Cornell—with or without known test scores—have performed well here. However, those who were admitted without test scores tended to have somewhat weaker semester GPAs, were more likely to fall out of “good academic standing,” and were less likely to re-enroll semester after semester. These patterns hold true holding constant students’ high school GPAs as well as other personal and high school attributes.”
“Good academic standing” means GPA of at least 2.0. The average GPA at Cornell is around 3.5.
4 Likes
I suspect anyone applying TO from a state with mandatory school-day testing, or a state where testing seats aren’t limited, will be presumed to have a low score. CA kids will be excused from this assumption.
1 Like
Starting what year will they require those scores? If I am a class of 2025, will I still be in “test optional”?
who Cornell - they are test recommended - that means maybe someone gets in without a test…but likely not many.
1 Like
Assuming that you will apply to start in fall 2025, Cornell will be test recommended for divisions that were test optional. If you take a gap year and apply to start in fall 2026, Cornell will be test required.
You should treat recommended as required unless it is not possible to get a test date between now and when you will apply (or something happens like you sign up for the only possible test date and then have COVID-19 that day or other medical issue preventing you from taking the test then).
4 Likes
So I am not against test scores or requiring them and I do think they provide another data point - however, when I see data like these I always wonder if they controlled for other things that can cause lower grades, etc., esp early on - income, less rigorous high schools and so on - I’d love to see if these data hold up if you compared high income test optional kids to high income kids who submitted or poorer kids who submitted vs those who didn’t - while I suspect you might still see a small difference I bet most of the effect goes away — but it would be super interesting if it didn’t.
2 Likes
I have seen multiple posters in this thread suggest that SATs can predict first year success and nothing more. I would be interested in their opinion of the Cornell report.
Cornell did say they controlled for “students’ high school GPAs as well as other personal and high school attributes,” but did not elaborate on what “personal” meant, and whether that included income.
Cornell’s statement on this: “The association with GPA may be attenuating somewhat as students accumulate more semesters of experience at the university. That is, the gap is smaller in the third semester than it is after the first semester. This is encouraging for those students who have persisted, but the robust evidence of increased rates of academic struggle and attrition remains a concern.”
In other words, those who successfully “weathered the storm” gradually caught up, but there were also those who succumbed.
3 Likes
I find it interesting that, unlike previous universities that will accept any sort of external evaluation, Cornell specifically wants the SAT or ACT.
1 Like
…there is still a gap.
There have been posters, with a religious fervor, who have insisted (in the face of studies like these) that standardized test scores show nothing beyond first year success. I don’t think this, or any other evidence, will sway them.
1 Like