Caltech and Swarthmore (conflict of interest notice: I have degrees from both) are luxury brands with good success (as brands measured by endowment). They are

Caltech and Swarthmore (conflict of interest notice: I have degrees from both) are luxury brands with good success (as brands measured by endowment). They are quite different from the ivies as described by Gladwell. I was always surprised that Caltech had the highest SATs of entering students: surely Harvard, Princeton, etc. had the ability to outdraw Caltech. Now I understand it: highest SATs is part of Caltech's luxury brand but not that of the ivies, just as high (er) performance is BMW but not Mercedes, or Lamborghini but not Rolls Royce. Another feature of Admissions that Gladwell doesn't mention in this article is one I discovered when reviewing 400 applicants a year for perhaps 15 spots in the Electrical Engineering PhD program at University of Rochester. Your job is to create a great department, NOT to find the 15 most deserving of the 400. My own evolution on the graduate admissions committee was from poring over the details of hundreds of applications, to using numerical screens to bring myself down to perhaps 60 that I would look at in some details. One of the great things about having so many applicants is I could afford to be wrong about some of my filter metrics. Perhaps limiting myself to 4 or 5 universities in China (for Chinese applicants) was not necessary, but having done so I still had 10X as many good applicants as we could admit. Is interpreting the admissions job as creating a great university rather than "fairly" picking the top applicants a feature or a bug? On reflection, I think it is clearly a feature. If you were hiring for a startup you would instantly recognize this approach to your selection process as the only one that vaguely made sense. Why would it be different when trying to create quality at older bigger places? But the kicker that makes it a sure thing as the right policy is this: it is IMPOSSIBLE to identify the best few % of a large pool of excellent applicants. When of two policy choices, one is impossible, it is likely it is not the optimum.