Lauren Rivera, an assistant professor at Northwestern's Kellogg School, has written a paper on the way that elite employers screen resumes (no ungated version found). Her claim is that top law firms and investment banks select for prestige in academic background, not for content. Various bloggers have commented.
It is odd that the soft firms, which market themselves to clients as being super-smart repositories of brainpower (of course this is largely a fiction; see point 3 above), would rely so heavily on university admissions committees. They effectively outsource a big chunk of due diligence on their most important investment (human capital) to a group of people whose judgement they somehow trust, but perhaps without detailed understanding. When I was on the faculty at Yale I knew people in admissions and it's not clear to me that they were the best able to spot potential in 18 year olds. In studies of expert performance admissions people are less good at predicting UG GPA than a simple algorithm. (The "algorithm" is simply a weighted sum of SAT and HS GPA!)
But this doesn't matter if the success of HYPS grads becomes a self-fulfilling prophecy. Once soft elite firms and large parts of the rest of society (in particular, clients) have accepted the idea that elite universities should be trusted to do the filtering, these schools will automatically produce large numbers of successful alumni -- the imprimatur itself has value. The outsourcing of human capital filtering is more dangerous for hard elite firms, with their more objective criteria: if they find that Yale grads aren't actually any good at pricing derivatives, writing code or designing chips, then they'll have to adopt a different filter.
Why don't firms that hire 22-year-olds ask college seniors to take the GMAT or GRE or LSAT and have those scores sent to them?
I think of this sort of employment screening as a form of cultural prejudice, somewhat akin to racism. The only way that racism can pay off economically for a firm is if the other firms that it deals with are also racists, so that it gets more revenue or lower costs in return for conforming to shared prejudices. The harder it is to measure quality objectively, the more chance there is for a racist equilibrium to assert itself. For example, if it hard to measure quality of research objectively in the humanities or the social sciences, then you can develop a community of prestigious degrees and journals that is based entirely on a set of shared prejudices. Similarly, you could get a racist equilibrium in occupations where shared prejudice causes value to be more correlated with irrelevant characteristics than with objectively measurable skills.
I call this post "Grist for Bryan" because he is a strong proponent of the signaling model of educational credentials. I am inclined to doubt that the world is quite so perverse.