lf: “Thibault, I don’t agree the “de facto threshold is actually a 1550 SAT score for unhooked middle- and upper-middle class applicants.” What I see, is: hit 750 and fine. Then attention turns to the rest- which is where those top performers often miss. This is the vital part.”
Err… that’s exactly what the term “threshold” describes: an initial gating mechanism, or screen if you like, after which the non-academic criteria are applied to the resulting shortlist of applicants. You’re restating my point.
But the problem with the non-academic criteria is that there is no attempt to explain how those criteria are weighted or applied.
What is almost certainly going on is that the adcoms are bending their application of those criteria to achieve predetermined outcomes for certain targets related to “building the class”-- what data scientists would call “optimizing” for those desired class-wide outcomes.
The only question is at what stage of the process this bending occurs: Is it happening only at the last stage, via intervention by the head of the adcom, or does it occur subtly at each step of the admissions process, via the lower-level readers’ revisions to their scoring of the subjective non-academic criteria.
This latter approach btw is precisely what used to happen at UCLA back in the day, according to Prof. Tim Groseclose-- see his book, “Cheating.” Prof. Groseclose also dumped the (masked) data for several years of UCLA undergrad admissions onto the internet for you to download.