Case Against Ed.jpg

Few people helped me more on my new book than economist David Balan.  Comments aside, he spent a whole day listening to me explain my spreadsheets.  That’s what I call dedication.  But despite his generosity, his assessment of the project is decidedly mixed.  Here’s David’s review, printed with his permission.

Let me start by saying that the book displays the Caplanian
virtues. It is thoughtful and careful and bold and fair and honest. It involves
big ideas and raises important questions that others either ignore or gloss
over, and it insists upon trying to answer them without letting opponents
retreat into nonsense or technical mumbo-jumbo. It is also clearly and
engagingly written.

In addition, the model that you have created for the book is
something of true and enduring value. Other sub-disciplines have had the idea
of trying to pull all relevant effects together into one single model.
Computable General Equilibrium models in macro is the main example comes to
mind. I hear that those are somewhat out of fashion nowadays, and I’m sure they
have their flaws, but they just make too much sense not to be at least worth
consulting. If you were a macroeconomist, how could you not want to at least
know what such a model says? You have done something similar for education. As we
discussed in your office, there are probably some technical improvements that
could be made to generalize it beyond the “four archetypes” framework, and/or
maybe do something like a Monte Carlo simulation to show that the existing
model is a tolerably close approximation to a richer model. But even as it
stands, it is a big step forward. It would be a great thing indeed if in the
future every empirical researcher would be asked, as a matter of course, to
indicate what parameter of Caplan’s model (or its descendants) should be set to
what value as a result of his or her research.

But for the model to serve that function, it would have to be
given a public-facing interface. That is, it would have to be possible to take
the model, plug in different parameters than the ones you choose (whether
because of new information or because they disagree with your parameter
estimates), and turn the crank and see what answer you get. To my mind, this is
the main value of this project to the discipline of economics, regardless of
the merits of the case against education.

As for the merits of the case, I must say that I am not convinced.
This will come as no surprise to you, but I will try to lay out my reasoning as
clearly as I can, in order to at least precisely isolate the locus of the
disagreement. I hope these comments will be received in the helpful spirit in
which they are intended. I am very happy to discuss any of this with you if you
think it will add value.

Top Level Impression:

At the highest level, the book seems to me to suffer from a flaw
similar to the one that I believe characterized that book by Michael Huemer. It
goes something like this:

1. Identify real, deep, and under-discussed problems with
something important that people accept so uncritically that they barely think
about it (the state; mass education).

2. Propose getting rid of that important something rather than
reforming it.

3. Defend the idea that it can be gotten rid of and not reformed
by reference to a radical alternative that is at worst fatally flawed on its
face, and at best badly under-supported (replace the state with private law
enforcement agencies; replace mass education with mass private employment of

Your proposed alternative to mass education is an utterly radical
change. It would have massive effects of all kinds throughout society. It would
have tremendous distributional effects. It is radically out of sample for any
empirical work that might inform what would happen. My view is that most of
those effects would be much more likely to be negative than positive, but even
if I’m wrong about the mean, you don’t have to be much of a small-c
conservative to worry about the variance of the likely effects of a change that

The book does next to nothing to support the claim that these
effects would in fact be good. It just says that work would teach skills that
are more useful than the ones gained in school, and that the socialization
benefits would also be at least as large. There is nothing else that even tries
to argue that the broader effects of this epochal change would be good effects.

When Huemer and David Friedman argue that we can realistically get
rid of the state because markets will bring about private security agencies
that will keep the peace, they are just being silly. It’s a dereliction of
intellectual duty, to the point that it made it hard for me to appreciate the
very real virtues of some of the other things they’ve written. It’s just a
device that lets them avoid having to say something unfun and unsexy like “The
state has violence at its core, and that’s a big problem, and we need to worry
a lot about how to reform the state to make it less of a problem” and instead
be able to say “Let’s get rid of the state! I have a totally great alternative
that will totally work!” I don’t put “Eliminate mass education” in nearly the
same category as their PSAs. But I do think it’s a (much) smaller version of
the same problem.

I wouldn’t hand the nation’s teenagers over to corporate power
even if I thought you were 100% right about signaling, and even if I thought
there was no chance that education will improve in the future. I would pick the
status quo in a heartbeat. The least of the problems is that the firms would
have no incentive to invest in general (as opposed to firm-specific) human
capital. That’s before we even get to the moral, social, and political effects.

For this reason, I think there would be value in talking a little
bit about some of the less radical (but still pretty radical) educational ideas
that you have discussed on your blog over the years. You could pitch them both
to people like me, who favor mass education, as well as to people who would
prefer to get rid of it but think their efforts are better spent on pushing
reforms. That way, you can bring along at least part of your audience for a
while, even if you are going to lose them on a real push for getting rid of
mass education.

This would include much more vocational education, letting kids
have more access to the library, letting them have more self-directed time,
exposing them to currently-low-status enrichment, like what it might be like to
be a plumber, basically just recognizing that a big part of the school day is
really day care. At a summer camp I used to work at, on Saturday afternoons we
had “optional mandatories” which meant that there were a number of activities,
and you could do any one you wanted but you had to do one. I have always
thought that that would be a great idea for a significant fraction of school
time. I know you have more.

In addition to all of this, I am only very partially persuaded by
the arguments and evidence about the magnitude of the signaling portion of the
return to education. When I was in your office, you said that 1/3 was the
lowest fraction that any halfway sensible person could possibly believe. I am
not sure what number I would offer, but I am inclined to think that 1/3 is much
more like a ceiling than a floor. Below are my reasons why.


To my mind, the most compelling empirical fact that you present in
the book is the gap between individual-level and country-level returns to
education. Something is clearly going on there. And as you point out, it’s not
clear whether the country-level returns are even positive, though they do
appear to be (they are positive in my paper with Steve Knack, for just one
example). But as we have discussed in the past, signaling is not the only
possible explanation for this. Another one is rent-seeking; some of that
privately valuable education is deployed for rent-seeking, which reduces its
productive value for society. To be clear, this story does not require that the
education be for rent-seeking. We’re not talking about pickpocketing classes.
It can perfectly well just be general education, that is either accompanied by
a rent-seeking ethos or more likely just being provided to people who are
likely to become rent-seekers anyway (e.g., children of corrupt oligarchs). As
you’ve pointed out, this story does not work if the education is so bad that no
skills are imparted at all; if there are no skills they cannot be used for
production or for rent-seeking. But most education is not like that.

This idea has a real pedigree, and I don’t understand why you are
so confident that it is not playing a role in generating those cross-country
results. And if rent-seeking is in fact important, it has a somewhat subtle
implication. It can still be part of a case against education, but only if you
think that mass education is largely used for rent-seeking in the contemporary
United States. I don’t know how the U.S. compares to other countries in this
regard, and there is a very good paper to be written that compares the
individual-level vs. country-level returns to education across countries with
bigger or smaller rent-seeking problems. But I would guess that the U.S. would
do pretty well on this score, meaning that an increment of education is a lot
more likely to go for production than for rent-seeking. And even if not, once
we’re talking about radical change, a mass campaign against rent-seeking would
be a better bet than eliminating mass education.


A big part of your argument for signaling is to look at the (low)
level of skills that people actually have, and argue that this represents an
upper bound on the human capital value of education by observing that education
cannot have provided more than 100% of those skills. This is accompanied by
evidence from the psychology literature about skill transfer; people learn
narrow things and can only apply them narrowly and temporarily. You dismiss as
wishful thinking all versions of the idea that they’re “learning how to learn”
and or otherwise indirectly gaining valuable hard-to-measure skills that were
not the direct object of instruction when they are getting so little (or at
least retaining so little) of the easy-to-measure skills that were the direct
object of instruction.

But I believe in soft skills even when there are no hard skills.
At the most basic level, I think the existence of those soft skills can be
inferred from the fact that we have any kind of advanced liberal society at
all. Think for a moment about what society would look like without mass
education. As you point out, many people wouldn’t (couldn’t) know much less
about science or history than they do now. And yet, the society would be
totally different. We don’t have too many contemporary Americans who believe in
witches? I mean really believe in witches. Why not? People have in the past,
and some people in the world do today. Most contemporary Americans believe that
if you lose an election you should turn over power to the winner, even if you
think you might be able to get the military on your side. Why? Why do most
Americans believe that you should vaccinate your kids? Why do most Americans
know a little something about how to prevent pregnancy and the spread of STDs?
Do you really think that things would work this way without mass education? Do
you believe any version of the idea that the existence of novels (which have
only been around for a few hundred years) has fundamentally changed human
nature because for the first time just about everyone in the society has spent
at least a little while in the other fellow’s shoes? And that this can be
valuable even if you’ve completely forgotten the plot of the handful of novels
that you grazed in high school? I really think I do.

I know that in your world everyone would at least be literate, but
would that be enough? Will people get these things from their on-the-job
training in corporate America? This is a version of the “beasts into men”
argument. It is hard for me even to imagine a truly uneducated society that
would not be beastly.

On a more prosaic level, I think you’re looking for evidence of
soft skills in the wrong place. It is an interesting and troubling fact that
people often cannot apply skills across domains, and that they tend not to
retain many of the skills that they do acquire. But some people can, at least
some of the time. A pretty fair number of people. And when they can, it’s
really valuable. In your office, you seemed to suggest that someone could
become a brain surgeon via on-the-job training. I don’t know what to make of
this. Being a brain surgeon clearly requires a wide of general thinking skills.
General thinking skills that they assume you have, or at least have the
rudiments of, before you start training. Some people have those skills. And
where did they learn them if not school? You could argue that school could be
way better at teaching those things. And you could argue for a more vocationally-oriented
program for future brain surgeons. But that training would still have a big
non-vocational component unless you simply defined those skills as occupational
skills. But this would be an error. Moreover, I think Caplan-style introspection
reveals that such skills are reasonably abundant among at least a substantial
minority of students, and that a lot of valuable activity could not get off the
ground without them being reasonably abundant. These are not hard skills that
get transferred from one domain to another; they are soft skills that inform
and support real hard skills for real jobs. It would be better if more people
had them, but they’re difficult to learn and difficult to teach! The fact that
a bunch of people to have them, at least to some extent, is a big deal.

I can think of other examples. I know a lot more microeconomics
than I did in my first semester of graduate school. But I could not walk into
the exam that I took then and pass it right this minute. Heck, I would be a
halfway decently qualified candidate to teach such a class, and I probably
couldn’t walk in this minute and pass the exam that I would write for that
class if I was teaching it! Another example comes from your book, which is the
sports statistics example. Nobody who took the statistics class could reproduce
the formulas, but at least some of them would have a sense of the Law of Large
Numbers or regression to the mean. It’s a fair question whether making them
manipulate the formulas is the best way to achieve that sense. Maybe it’s
really scaffolding for really getting the lessons (you need it to build the
ideas but not afterwards), or maybe it would be better to just spend more time
talking directly about errors that can come from not knowing these concepts. I
don’t know. But either way, that’s exactly the kind of soft learning even when
the hard learning was forgotten that I believe in. Maybe making Daniel San
paint the fence and wax the car is the best way to teach karate and maybe it
isn’t (although I could never see the value in not telling him that it was part
of his training). But either way, it’s easy to imagine some exercise that would
help in a stage of training that was never needed again, and hence forgotten.

Not too long ago, I had occasion to need to use a Lagrange
Multiplier for something. I don’t think I used one since graduate school. And I
had to look up how to use it. I could not have passed an exam about it. But I
knew what one was, and when a particular problem arose, I knew that it was what
I needed.

Bottom line: There are some things that most people pick up, where
civilization would be impossible if they didn’t. Then there are some things
that a lot of people pick up, even if not most, where real productivity would
be impossible if they didn’t. And those things get picked up mostly in school.


Aside from the above claim that most of at least many people
benefits from certain “soft” aspects of education, there are also the benefits
that accrue to only a few people, but are really big. I remember the teacher
who first taught me about supply and demand in I think third grade. She is
mentioned in the acknowledgments to my dissertation. I’m not saying that that’s
the reason I became an economist, but something was the reason, and that something
was some kind of school. And the same is true for just about everyone who ends
up with high human capital. Take any highly productive person you like. Even if
most people don’t learn a lot of science in school, future scientists sure do.
And it’s hard to know who the future scientists are going to be without making
everybody learn some science. And the ones who don’t will at least vaccinate
their kids.

Every future scientist saw his first science in school. Same for
every future arts lover (not art teacher!). Same for every everything. This
doesn’t mean that the exposure was of the right kind or in the right amount for
the right kids. There is plenty to argue about here. But most enthusiasm,
whether productive or personally enriching or both, started in school, and it’s
not easy to see where else it would come from. Certainly that’s true for me. In
school, from teachers. More from good teachers than from bad ones, but not
mostly from superheroes.

Future Improvements in Education:

You dismiss as wishful thinking the idea that education is likely
to be better in the future. Rather, you propose cutting now and adding later if
education manages to improve enough to justify it. But this is not the kind of
thing that can be reversed quickly. If you dismantle the education system it
will take decades to rebuild.

Moreover, I do believe that major future improvements are likely.
The whole idea of rigorously evaluating education and basing policy on the
findings of those evaluations is only a few decades old. Everything else gets
better with rational inquiry, why wouldn’t education? To give one small
example, my cousin recently retired from a career in special education. He told
me that the whole idea of doing anything at all for kids with special needs is
about as old as his career. Before that, they were just labeled as dumb and
that was it. We’ve just barely started even trying.

The Educational Experience:

You have some material on the idea that human capital can be had
in lots of places besides formal education. You can learn online, or you can
even walk into a college class for no credit. From this you conclude that
formal school credentials can add no additional value. This seems like a straw
man to me. Totally self-directed education is something that was always
possible for superheroes. There is that mathematician from a poor Indian
village who turned himself into a world-class genius from a few old math books
that he found lying around. Today, you wouldn’t need to be quite the superman
that he was, but learning with no framework, no context, and no guidance is
simply not within the capabilities of most people. It certainly wouldn’t have
been possible for me. The same point goes for the ability of people to
appreciate high culture via the Internet; they need some instruction and


Relatedly, the book does not address the idea that school not only
provides human capital, but also certifies human capital. This is not the same
as signaling. Imagine a world where school really does provide lots of skills,
but those skills can also be fully gotten on your own online (i.e., you
discount everything immediately above). How would employers know that you have
those skills? You could probably give the answer that you reject as an argument
against signaling: why not just give people a test instead of requiring a
degree? But even if you think that a test can evaluate skills as effectively as
course grades underneath a real teacher, this isn’t a very strong response.
Because in this world, unlike in the signaling world, real, hard-to-acquire
skills are being gained. So it’s not 4 years vs. 3 hours. It’s 4 years minus
the amount of time that you’d have to take to learn the skills on your own vs.
3 hours. That’s maybe still a big hump, but a much more manageable hump for
school to get over in order to justify itself.

Market Test:

Some people object to the signaling story on the grounds that
school has passed the market test. You reasonably reply that it’s no market
test with a trillion dollar subsidy, plus you give a story for why the
equilibrium with high returns to schooling is stable (any deviator will send a
very bad signal and be way worse off). But there is another version of this
objection that I do not believe you have addressed. Even if the equilibrium is
stable with education being mostly signaling, that doesn’t mean that the school
has to give a worthless education. Once all those kids are stuck in school
because of fear of market punishment if they don’t go, there’s nothing to stop
the school from actually teaching them something, or the parents from demanding
it. And even you don’t claim that there’s nothing worthwhile to teach. So if
schools could deviate from the standard curriculum but do not, does that imply
some kind of market judgment that the curriculum has value?

I don’t want to take this point too far. I am not saying that the
curriculum we have is optimal, or even good. I actually agree with many of your
specific criticisms. But it is a little hard to imagine that a curriculum that
is a total disaster has persisted this long.

And to whatever extent you don’t believe that such an inference is
justified, there is room to change the curriculum to improve school. This is
true regardless of how much of education you think is signaling; it’s always
better to improve it unless you really think that you should and can try get
rid of it instead.

Sheepskin Effects:

A big part of your argument rests on sheepskin effects. How could
it be human capital if the rewards are so discontinuous? Signaling does make
sense to me here. But even here there is a human capital story. The student who
goes to college for three years and then doesn’t finish is probably the person
who didn’t learn anything during those three years. It’s still signaling, but it’s
not signaling the presence of a fixed attribute like smarts or
conscientiousness or conformity. It’s signaling the fact that this student
didn’t learn what most other students learned in school, despite managing to
not fail out for three years.

Odds and Ends:

1. You favor the 3Rs, but you disfavor subjects like history. But
you have to 3Rs about something. It might as well be things that are thought to
have other kinds of merit, even if that merit is fairly low. How much business
writing can a kid do?

2. While other important issues are addressed later in the book,
early on it is presented as being all about what skills are valuable for jobs.
I would say up front that other issues will be addressed.

3. I think you underestimate the consumption value of the college
experience, if not necessarily college classes. Relatedly, what you think of as
a bug of college I think of as a feature. Coddled college students who get an
inflated sense of their own importance can be annoying. But I see great value in
there being a time in a young life when you are not accountable to a hierarchy,
when you can explore and grow and turn into your own person without having to
sing for your supper. You’ll have 45 years for that, and that’s plenty of time.
I think of the recent college graduates that we hire as RAs and paralegals at my job. A year earlier, they were college seniors studying the rise and fall
of empires or real analysis or whatever, and now they are doing drudgery for me
and for the lawyers for little money. And you know what, they do it! The fact
that they’ve been coddled doesn’t seem to hurt them at all.

4. As for the gleeful reaction of students when the teacher
cancels class, I do not draw the same inference that you do. It’s just a
self-discipline issue. To give one trivial example, I remember taking a
non-credit karate class in college. I signed up for it of my own free will, and
yet I was happy when class was cancelled, because it meant I could slack off.
In some grand sense, I was worse off, and I knew it; I’d know less karate and
I’d get less exercise. In the moment I was happy.

5. The refutation of the Tyler’s “the market will figure out your
true human capital” objection was strong, but I still think it puts a
significant bite in the signaling story.

6. As we discussed in your office, the amount that people are
willing to pay for an increment of health is probably concave in the level of
health. This might be relevant for your calculations.

7. The material about peer effects on page 330 says that if one
extra kid goes to college and becomes less religious as a result, that the
effect on religiosity of society is ambiguous. Why would that be? You could
tell exotic stories to the contrary, but wouldn’t the simplest and most
intuitive story be that you’ve moved one person from the religion bucket to the
non-religion bucket and that’s it?

8. When we were in your office we talked about the Rosenwald
schools. It might be worth knowing that story.

9. Have you ever heard of someone named Daniel Lilienthal? Someone
told me that he wrote an anti-education book of some kind. Might be worth
looking into.

It is a very impressive work, and it has a lot of virtues. I don’t
believe the basic story, but I am not likely to change your mind about that. So
my backup suggestion is to spend some effort on carving out the parts of the
work that can still be of value to people who are not convinced, including a
public-facing version of the model as well as some discussion of less radical
alternative reforms.