>>>What it's saying is that they're unfortunately the best proxy historically available
And yet most of the global industries of today were formed by people with little or no formal education (Carnegie and steel, for one). And then ruined by those who have had it.
The point is that regardless of the merits of a Yale degree, it doesn't really come close to measuring, "will this person be good at this job?" Signaling is a shortcut, something you do because you don't have time or energy to measure something more directly aligned with what you need. It's not wrong, just a blunt instrument. Yale will graduate tons of dud candidates, and far lesser schools will graduate hoards of great ones, even if they do so at a lower rate. So the question is, can we create a more precise tool for measuring what we actually want to measure? We should certainly be trying. A Yale education might be well correlated to, say, the ability to learn quickly, but a good test for the ability to learn quickly would be even better correlated. If the talents that got you into Yale, not to mention the stellar education you got there, cause you do better in those measurements, then good for you! (None of that is to say this particular company is successful at this measurement, BTW.)
Another thing that went unstated in the article is value. Yale (and MIT) grads are expensive! Penn State grads, relatively speaking, are not, which creates a huge incentive to find the good ones and hire them instead. So at the very least, you should agree there's a big opportunity in a more precise tool for separating the chaff from the wheat, even if you think all the Yale wheat is pre-husked.