The title of this thread is about the Ivy League. The vast majority of the NACAC survey colleges are less selective ones that accept the majority of applicants – very different admissions criteria from the Ivy League. You can get some clues about how the weightings may differ at selective private colleges by looking at the regression coefficients in appendix B. Appendix B suggests lower admit rate private colleges tend to have the following differences compared to the typical surveyed NACAC college.
Criteria That is Generally More Important at Selective Private Colleges
Essay is much more important
LORs are much more important
ECs, Work, and Awards are much more important
Race, Legacy, First Gen, and Gender are much more important
Ability to Pay is much more important
Grades in College Prep Classes are somewhat more important
Strength of Curriculum is somewhat more important
Interview is somewhat more important*
Demonstrated Interest is somewhat more important*
*More/Less important for private, Little difference for low admit rate
Overall GPA is slightly more important
Criteria That is Generally Less Important at Selective Private Colleges
Test Scores are somewhat less important*
*More/Less important for private, Little difference for low admit rate
Essentially every criteria except test scores tends to be weighted more heavily at selective private colleges than at the average NACAC college. I don’t find this result shocking, as highly selective colleges are more likely to emphasize holistic criteria and less likely to focus on stats. The reverse is true for less selective publics, which are more likely to focus on stats and less likely to emphasize holistic criteria.
We have much more detailed information about Harvard’s admission system due to the lawsuit analyses. Harvard has 4 core ratings on which they rate each applicant on a scale of 1 (best) to ~5 (worst) – academic, extra-curricular, personal, and athletic. These 4 ratings have a similar degree of influence between a rating of 3 (typical) to 2 (well above average), which is the most common distinction among applicants. Academic rating does not appear to be more influential than everything else, as is commonly assumed. Harvard also rates applicants in several subcategories such as teacher LORs, counselor LORs, and interview. These sub-ratings influence some of the core ratings above, as well as the overall holistic decision.
Both sides of the lawsuit created models that could explain the majority of variance in admission decisions using these ratings and many other criteria. The model created by Harvard’s expert could explain 64% of variance in admission decisions, with a r (correlation coefficient) of 0.8. When only the following criteria was removed from the admission model, the model created by Harvard’s expert lost the following degrees of explanatory power.
Full Model – Can explain 64% of variance
LORs and Interview Removed – Decrease by factor of 50% to 32% explained
Personal Rating Removed – Decrease by factor of 19% to 52% explained
Academic Rating Removed – Decrease by factor of 17% to 53% explained
EC Rating Removed – Decrease by factor of 13% to 55% explained
Athletic Rating Removed* – Decrease by factor of 12% to 56% explained
*Estimated based on individual contribution
Removing LORs + Interview were by far the most influential. Among the core categories, personal rating was most influential. The two most influential areas both have little direct dependence on scores, although there is no doubt that there is a correlation with test scores – students with higher test scores are more likely to have top LORs than the average applicant.
Among the categories above, the analyses suggest test scores have little direct influence beyond the academic rating… A report by Harvard’s Office of Internal research found that a combination of test scores and GPA stats could explain ~70% of variance in academic rating – the vast majority of variance in academic rating. It’s unclear exactly how much relative influence transcript and test scores have towards the academic rating, but highly selective private colleges generally claim that transcript is much more influential than scores. If this is true, then removing test scores alone is expected to decrease predictive ability by a small fraction of the 17% reduction listed above… most likely well under 10% decrease in variance of admission decisions explained.
In short, removing test scores from admission consideration is not expected to influence the overwhelming majority of decisions in past classes since the different admission criteria tend to be correlated with one another. Students who ace the rest of the application and have incredible grades, course rigor, LORs, essays, ECs and out of classroom activities, interview, personal traits, … compared to the overall Harvard pool rarely bomb the SAT. The more criteria that is significantly considered, the less likely SAT is to change the decision based on that other criteria.