If you just want to cut down the number of applicants, that's even easier - just pick a random sample to look at.
It only makes sense to do it based on features of the individual candidates if you believe the remaining pool is, on average, better. (... or maybe more precisely, "that you will be able to select better candidates from the remaining pool for a given amount of effort").
No, filtering in the presence of an excessive number of applicants makes sense if the resulting pool is not worse than the original pool. They're saying academic performance is not correlated with professional competence. That means precisely that filtering by degree results in a not worse applicant pool.
Why expend additional effort to filter along meaningless axes to wind up with some unknown lesser number of applicants (which you might then want to shrink further) when you can already filter meaninglessly to end up with precisely the number of applications you want (by simply grabbing randomly and dropping the others on the floor)?
Because it lets people feel better about themselves and is - superficially - justifiable.
A lot of what we do in hiring, even in tech, is effectively just ways to whittle down numbers of applicants, but even here on HN there are people who will insist their way of whittling numbers down improves the applicant pool, because hand-waving and superficial logic.
It only makes sense to do it based on features of the individual candidates if you believe the remaining pool is, on average, better. (... or maybe more precisely, "that you will be able to select better candidates from the remaining pool for a given amount of effort").