Arnold's post on segregation makes several points on the signaling model of education. I'm here to rebut them. Arnold's in blockquotes:
1. Where Bryan sees college as a useful signaling device for those
who are cognitively gifted, I see it as a useful segregation device for
As I've saidseveraltimes, I see college as a useful signaling device not just for intelligence, but for two "Vicky" traits: conscientiousness and conformity. Which makes me wonder: If college is where the Vickies go, won't college be a strong signal that you're a Vicky? If so, Arnold's model morphs into mine.
2. The segregation model predicts that as the society gets
wealthier, the dollar cost of college will get higher. The signaling
model would not necessarily predict that. In fact, it would predict
that the market would try to find less expensive signals.
Au contraire. Not only does the the signaling model predict that a higher payoff for college will increase demand for signals; it predicts that if the price of signaling falls, people need to increase their quantity of signaling to remain separate from the pack. As I've explained before:
Many economists assume that market forces will somehow figure out a way
to make signaling costs disappear. But as far as I can tell, they never
explain why signaling costs would be easier to eliminate than any other
costs. And on reflection, the truth is precisely the reverse:
Signaling costs are especially hard to eliminate. Why? Because
when you make signaling cheaper, agents' natural response is to signal
more intensely or on another dimension.
Let me illustrate my claim with a prediction: The typical engagement ring will always cost several weeks' income.
If industry figures out how to cheaply synthesize gold and diamonds,
we'll start making engagement rings out of something else - platinum and
rubies, or ivory and T-rex teeth. Why? Because one major function of
engagement rings is to signal commitment with an expensive gift! To
separate the sheep from the goats, the signal has to be expensive enough
to convince the goats to give up.
3. The segregation model predicts the emergence of institutions like
Boston University and George Washington University, which require much
more money than brains to attend, and yet which have fairly high
I'm happy to admit that, in addition to their other functions, colleges are social clubs. I suspect that this social club function is especially important for religious colleges (think Brigham Young) and less-selective private colleges. But even if students in "clubby" colleges are implausibly apathetic about impressing future employers, belonging to any selective club almost automatically sends a signal. As long as (a) the average graduate of BU or GWU possesses special traits that employers value; and (b) employers can't costlessly measure these traits, a BU or GWU degree will pay off in the labor market.
4. I think that if either the utilitarian model or the signaling
model of higher education were correct, I would be sure to collect on
any bet I make with Bryan about the demise of colleges. If college as
we know it manages to persist for another two decades, it will be thanks
to the segregation model.
Arnold's right about what he calls the "utilitarian model," better known as the human capital model. But contrary to Arnold, signaling models readily predict the persistence of costly, inefficient customs. Indeed, it's the persistence of costly, inefficient customs that inspire much of the signaling literature.
Given Arnold's faith in educational innovation, I have to ask: If entrepreneurs can figure out cheaper ways to teach students, why can't they figure out cheaper ways to segregate students? Suppose Harvard is just a Vicky Club. On Arnold's account, there's no reason why an upstart Vicky Club couldn't come along and offer Harvard students Harvard-level segregation for a fraction of the cost. In the signaling model, of course, this wouldn't work: Quitting Harvard to join an "upstart Vicky Club" sends a godawful signal to employers and the world.