It's clear the SAT/ACT has predictive power for highly-selective colleges, such as MIT. And therefore, they are valuable for these colleges – especially for the Math scores as MIT suggests. The value of SAT/ACT scores decreases as selectivity decreases or as math abilities matter less for admission (e.g., liberal arts programs).
Here are some related points:
- Harvard considers roughly 4 in 5 applicants to be academically capable of doing the work at Harvard (about 50,000 applicants of which Harvard only accepts 2,000). This data is pulled from their court documents, and my team wrote about it here: https://writingcenter.prompt.com/posts/strong-essays-increas....
- This means that most applicants at highly-selective colleges are very similar academically. Colleges are mostly just using grades, academic rigor, and test scores to determine whether the student will be able to succeed doing the work in college. Absent other information on academic preparation (e.g., not having access to AP/IB classes), the SAT/ACT score can be a critical signal of whether the student can do the work. Students with well-above-the-bar academics are admitted at a 3x clip to those just above the academic bar. But other parts of the application (e.g., essays, athletics) can have a much stronger effect on admissions chances (e.g., a strong personal score, much of which is essay-related, can have a 10x increase on admissions chances).
- Math SAT really is highly predictive of math abilities. When I was with McKinsey, we asked for applicants' SAT scores because it was highly predictive of people succeeding at McKinsey. People hired with scores below 700 struggled to succeed analytically. So, McKinsey used 700 as a bar. MIT is roughly doing the same thing here. Other colleges do this as well.
- Outside of highly-selective institutions, the SAT/ACT can have less predictive power in student success in college than other factors (e.g., GPA). There are a bunch of great analyses at fairtest.org that looks at these exams - e.g., breaking scores down by race.
So overall, we tend to give weight to what we know and what data we're looking at. Most of the SAT/ACT analyses out there are looking across all students. Here, MIT is looking at just their proportion of students. So, both things can be true – the SAT/ACT may not be a useful predictor for the vast majority of students. But scores can (and do) matter for the highest performers, the approximately 1% of high school graduates attending the most selective colleges.
And as MIT states, a perfect SAT/ACT score doesn't matter all that much. All they're using the scores for is to provide an indication of whether the student is above their bar for being able to do the work (e.g., not failing multivariable calculus).
Note: I did go to MIT – some of you may think this is relevant. I also run the largest college essay coaching company globally, Prompt.com. So I've spent a lot of time understanding college admissions.