While writing my final reply to Bill Dickens, I often wished I were writing six short posts instead of one long one.  My favorite could-have-been free-standing points (Bill’s in quotes, I’m not):

1. Policy implications.

Further, you must believe that there is a more
efficient way to solve the problem of figuring out who is fit for what type of
job than running them through the educational gauntlet. I take it from your
reply you think less schooling and more work is the way to accomplish this.

Right.  Some
signaling is socially productive because it improves the match between
workers and jobs.  But it’s privately optimal for students to far
exceed the social optimum. 

My preferred
policy is simply to end government subsidies for education.  It might
be even more efficient to go further and impose a Pigovian tax on
education.  But since education is a mix of human capital creation and
signaling, and government has already made a mess of things, I think
the all-things-considered best approach is separation of school and state.

2. Against incredulity.

In your reply you stated that education’s value added was
maybe 20% of the private returns to education. Does this mean that you think
fully 80% of the value of time and other resources spent on education are
wasted? Wow! Given how much we spend on education and how much time we spend in
school that is an enormous inefficiency! 
How is that possible?

you told me that only 20% of oil production was socially efficient,
because the other 80% had large negative externalities – pollution,
global warming, whatever.  Would “How is that possible?” be a
reasonable reply?  I don’t think so.  In both cases, the point is that
individuals and firms are advancing their own interests by ignoring the large negative side effects of their behavior.

3. My comfort zone.

…Bryan, I hope you appreciate the irony here.
In our first public debate you are taking the side of widespread market failure
while I’m defending the ability of private ingenuity to work its way around
whatever problems the world throws in its way – the reverse of our usual roles.
I must say I’m enjoying being in this position because I normally grant that it
is the person arguing against efficiency who has the burden of proof.

It’s not
an ideologically comfortable position for me.  But first-hand
observation of the education industry, plus the psychological evidence against
the empirical importance of “learning how to learn,” “training the
mind,” etc., leave me little choice.  On a deeper level, admittedly, I
am taking a characteristically Caplanian position: Government is making
a genuine market failure worse because voters systematically overestimate the social benefits of education.

4. Why signaling is stable.

To begin with, you don’t have to jump all the way to the
most efficient possible system in one leap. A small innovation that gives you a
slight advantage should be enough to allow a private institution to prosper and
shouldn’t cause a big adverse selection problem as so many factors go into
college choice.

This sounds very reasonable.  But reread your last sentence carefully.  If the advantage is only “slight,” then a slight adverse
selection problem is probably all it takes to cancel out the benefit. 
So there’s not much reason to be optimistic about stepwise progress to
a low-signaling world.

5. Educational innovation: When it works, when it doesn’t.

Your argument that innovation is
impossible because of adverse selection is untenable in the face of massive
amounts of innovation in the educational system everywhere.

I never said that all educational innovation generates adverse selection.  You get adverse selection when you try to make education easier
– when you reduce the number and difficulty of meaningless hoops
students have to jump through.   Co-ed dorms and high-tech classrooms
don’t attract losers.  Cutting math and foreign language requirements
do.  And as I said earlier, it only takes a slight adverse selection problem to cancel the benefits of a slightly more efficient curriculum.

6. My bottom line.

I’m standing my ground.  I’m happy to admit that I’ve got a 10
percentage-point standard error around my “80% signaling” estimate. 
But a priori arguments about how easily market forces would undermine
signaling don’t sway me.  I’ve spent too many years learning – and
teaching – material that’s all-but-useless in the real world.