Arnold makes an intriguing remark about education:

If college were truly a utilitarian good, all it would take to turn
these edifice-complex campuses into ghost towns is a good hack for the
accreditation process.

But he’s skeptical because:

[A]t the high levels, college is a status good. Let me repeat that going
to a top college today is like belonging to the right church in 1850 or
the right country club in 1950. When you are supplying a status good,
ostentatiously wasting money on buildings can increase demand.

Arnold is dangerously close to a variant on the signaling model of education – a model he’s previously rejected because:

[I]t suggests that there is a huge unexploited profit opportunity for
employers and employees who can come up with alternative signals. And
yet nobody tries to set up a system for identifying and hiring smart
high school graduates.

Notice that you can use Arnold’s objection to the standard signaling model of education to attack Arnold’s “status good” story: Why oh why hasn’t anyone come up with alternative, low-cost ways to signal status?

A big part of my answer to Arnold, as I’ve said before, is that education doesn’t just signal intelligence and conscientiousness; it’s also signals another character trait employers pragmatically cherish: conformity.  This leaves us in a catch-22, because experimenting with new ways to signal conformity is a strong signal of… non-conformity!