Attention Caplanites: School is Less Wasteful Than You Think!
10 reasons I'm skeptical of The Case Against Education
Bryan Caplan’s The Case Against Education has a lot of purchase in the blogosphere. If you haven’t heard of it, you should put down this post and go read his article in The Atlantic to see it at its condensed best. To this day, very few essays have influenced the way I perceive the world so much.1
But while Bryan does an admirable job of advocating a contrarian thesis that is far more defensible than many of its critics suppose, I am now much less convinced of the Caplanian position than I once was. To more fully flesh out my thinking on these matters—and in belated response to Ives Parr—I present ten reasons to doubt that the education system is a waste of time and money.
1. Bryan has the Burden of Proof
We can speculate that cutting spending by 80% would be a great boon, but when the burden of proof is against you, speculation can’t surmount it.
Bryan Caplan, The Case Against Education.2
To anyone not deeply siloed on these issues, it should be obvious that Caplanians need to provide much stronger support for their position than those closer to the status quo. Bryan himself acknowledges minimal concrete evidence for radical reforms of the type he rhetorically supports.3
Yet many in the blogosphere appear to believe just the opposite. For instance Ives Parr writes,
Advocates of socially undesirable views are unjustly held to a higher level of scrutiny… Those who advocate continuing to spend exorbitant amounts of time and money on a program should establish beyond a reasonable doubt that it is working. Since markets tend to weed out costly and unworthwhile endeavors, I believe the current education system survives through coercion and subsidy. (emphasis added)
While it is reasonable and appropriate to regularly scrutinize both current and prospective expenditures—especially when it comes to historically new interventions of the type Parr summarizes in section five—expecting a “beyond a reasonable doubt” standard from those who don’t want to cut the bulk of U.S. educational spending is a bridge too far.
2. The Sheepskin Effect Isn’t Just Signaling
Many people point to the fact that students with a degree earn a large premium relative to those with nearly the same amount of schooling but no degree as knockdown evidence for signaling.
But in fact there are competing explanations for the so-called “sheepskin effect”. Here’s one from Nick Huntington Klein:
What schools actually *do* when they allow you to continue in your education is, effectively, measure what you’ve learned and see if it passes some minimum standard. If you don’t, you drop out. We end up with those failing the (lax) minimum populating the dropout years. They’ve learned little, so they earn little. In the final year, you see everyone who passes the minimum, whether they learned just enough or WAY MORE than enough. The final year contains a wide range of big learners, so on average there’s a big jump in earnings that year.
Intuitive, based on literal actions people perform, and a totally-HC explanation of sheepskin effects.
Considering that Bryan uses the size of the sheepskin effect to set his lower bound for wasteful signaling,4 the mere existence of sensible alternatives is a big deal.5
3. Schools Teach Conformity
Education doesn’t just signal intelligence and conscientiousness; it also signals another character trait employers pragmatically cherish: conformity.
Though most of us prefer to emphasize the value of our intellect and individuality to the world, it’s hard to deny that employers demand conformity.
Conformists do what they’re told and don’t look for ways to skirt the rules. They’re more likely to assimilate into corporate culture and are easier to monitor, motivate, and predict in a wide variety of circumstances. When it comes right down to it, you can probably pay them less for doing the same amount of work.6
But while Caplanites readily assert this point to explain why the signal from education can’t easily be replaced by market alternatives, they conspicuously neglect the challenge it poses to their broader assault on schooling’s societal worth: so long as otherwise useless and arbitrary curricula succeed at teaching students to conform, they still contribute to human capital formation.
Beware lest social desirability bias blind you to this ugly yet important truth…
4. Cheating is Scarcely Punished
The signaling model implies a lot of vigilance against cheating. As Bryan explains,
Detecting and punishing cheaters preserve the signaling value of your school’s diploma. When more of your students cheat their way to graduation, firms that hire your students are less likely to get the smart, hardworking team players they’re paying for. Every time your school expels a cheater, you protect the good name of your graduates—past, present, and future.
And yet… there seems to be very little expulsion. At the comparatively strict7 University of Virginia, a mere 1104 students were expelled between 1919 and 2017 (less than 5% of last Fall’s total enrollment alone.) Of these, a disproportionate number were reported by a single teacher in a single year (the big spike) suggesting that even the slightest effort at the margin could uncover many more cheaters.8
In more representative samples, cheaters themselves report that they seldom face consequences, and I’m inclined to believe them.9
5. Fadeout is Weaker than Advertised
Bryan is a strong proponent of the fadeout hypothesis, which holds that most of what students learn is forgotten over time.10 But how well does he actually defend this position?
Since very few researchers actually measure post-graduation fadeout, Bryan largely relies on a single study:
A rare—and discouraging—exception: One major study tested roughly a thousand people’s knowledge of algebra and geometry. Some participants were still in high school; the rest were adults between 19 and 84 years old. The researchers had data on subjects’ full mathematical education. Main finding: Most people who take high school algebra and geometry forget about half of what they learn within five years and forget almost everything within twenty five years. Only people who continue on to calculus retain most of their algebra and geometry. (emphasis added)
But is this really so damning of retention? Once you look past the “discouraging” framing, there’s a lot to be optimistic about. From the source abstract:
Even in the absence of further rehearsal activities, individuals who take college-level mathematics courses at or above the level of calculus have minimal losses of high school algebra for half a century.
This is the single best piece of evidence for fadeout, and to me it primarily demonstrates the effectiveness of practicing beyond the point of initial mastery—what psychologists call overlearning. If calculus helps students retain their algebra, might that suggest that algebra helps students retain their sums and fractions? It’s an important question I wish Caplan did more to contemplate.11
Instead he proceeds to rattle off all of the ways in which the American public is generally ignorant as reported in nationally representative tests and surveys. To uncharitably summarize, this portion of the book goes a little like this— “Did you know only X% of Americans can correctly say how many senators they have? Hahahaha, the education system MUST be terrible!”
While there is some merit to reporting an upper bound on the things students remember from their time in school, this sort of exercise largely fails to do even that for many of the things employers actually care about their workers retaining [hint: not high school civics].12 Does learned ability to communicate, meet deadlines, or sustain focused cognition fadeout? These data can’t tell us.
Coverage concerns aside, Bryan adopts a misleading presentation of the results he does have. Many of his tables report Americans’ knowledge in the aggregate—including dropouts and others with very little schooling—and corrected for guessing, which is a good way to shock the reader with very low numbers that ultimately have very little relevance to the question at hand: how much does education improve student knowledge relative to what they would have known without it?
If ignorance is pervasive at a baseline, then it’s perfectly reasonable to suppose that the public would be even less enlightened without our education system.13
6. Signaling vs. Human Capital Can’t Truly Be Settled
Because nearly all results can be explained using either signaling or human capital, the vast majority of empirical results are void of theoretical contribution to this particular "why" question.
Human Capital vs. Signaling is Empirically Unresolvable (Huntington-Klein 2020)
Economists place a lot of emphasis on testing the distinct predictions of competing models to determine which theories hold up and which can be falsified. But as Huntington-Klein establishes in his published paper on the topic, signaling and human capital are extremely tricky to disentangle empirically.14
While I’m sympathetic to Bryan’s dissatisfaction with natural experiments (33:01), accumulating lots of internally valid estimates from a wide range of different contexts—and largely abandoning any general model—might be the best we can hope for.
7. Signaling != Waste
Assume for a moment that Caplan is being overly conservative—education is not 50% signaling or even 80% signaling, but 100% signaling.
If all workers compete with each other for a single job, my gain is your loss and every second spent in school is pure waste. It then follows that Caplan’s education-cutting proposal is a massive step in the right direction.
But the second you introduce multiple types of jobs, signaling ceases to be zero-sum,15 and Bryan’s identity function mapping from percentage signaling to percentage that should be cut no longer works.1617
8. Learning on the Job isn’t that Great
Couldn’t we all just receive on-the-job training more directly suited to the tasks that need to be done? I have my doubts.
For one thing, teachers select into the profession based on some pretty good qualities—they are often patient, nurturing, and empathetic. Most genuinely desire to help students and their schedule commits them to at least go through the motions of actually doing it.
In contrast, people I’ve met in the working world tend to have comparatively little time and interest for teaching or mentorship. Apart from the many individual coworkers and bosses that expect you to show up and figure things out with minimal direction, there is a more general atmosphere of “we’re paying you” that is less conducive to asking questions relative to the classroom or office hours.
There is also the uncomfortable reality that even faculty at low-ranking schools are likely to be smarter and more accountable than the person tasked with training in many workplaces. Firms lack sufficient incentive to allocate very much talent toward formal training because their employees can then take that training and use it demand higher wages or run to a competitor. What training that does exist is almost all dreadful propaganda in the service of avoiding lawsuits.
Until AI technology can deliver on its potential for radically improved mentorship at scale, the basic structure of our education system (with significant room for improvement at the margin!) might still be our best bet.
9. Context is that which is scarce
Schools make virtually no one fluent in a foreign language. Only 0.7% claim to have learned a foreign language "very well" in school; another 1.7% claim to have learned a foreign language "well" in school. Since these are self-reports, true linguistic competence must be even worse. The hard truth: if you didn't acquire fluency in the home, you almost certainly don't have it.
Bryan Caplan18
At first glance, there are few courses that seem like a bigger waste of time than foreign language credits. If students don’t actually learn to speak the language, what else could they possibly be learning?
Here is one answer…
Foreign language mainly teaches you that it’s possible to learn a foreign language, that there are other cultures out there and you can learn about them, that this is something gringos actually do sometimes, etc…
Lots of missionaries don’t attain fluency either, but exposure to foreign language and culture appears to teach them something all the same.
Seen in this light, even the most ineffective classes we make students sit through likely hold some redeeming value.
10. Role Models, Peer Group, and Identity
Give them a fine reputation to live up to, and they will make prodigious efforts.
Dale Carnegie
Why do so many people make New Years’ Resolutions? In part, because it is much easier to motivate attempts at self-improvement when others around us are likewise committed. Just months or weeks later, many of these resolutions fall apart for the exact same reason.19
In contrast, schooling provides an ongoing coordination mechanism to motivate learning above and beyond what many of us are likely to do on our own and directs such efforts toward a coherent, organized sequence of topics and materials to study. While there are some genuinely brilliant self-guided learners, I suspect that there are many more who due to conformity, laziness, or indecision would fail to generate much useful knowledge and internal ordering without that structure.
College inspires students to become a certain kind of person by placing them in a social context where authority figures are highly learned and peers have certain goals and aspirations.20 After all formal instruction has ended, the university further impresses upon graduates a particular sense of scholarly identity, an affiliative pride that can be continually reinforced by branded merchandise, alumni donations, and legacy admissions. The ritual of commencement largely serves to induct departing students into a kind of priesthood, and while frequently mocked, is nonetheless enduring and widely celebrated.
Who knows? Perhaps even the sheepskin itself plays a role.
Thanks for reading! If you enjoyed this, I would appreciate if you would subscribe or share this article with one of your friends—over-educated, under-educated, or otherwise.
Apart from its relevance to education, Bryan’s essay introduced me to the George Mason tradition of economics blogging. It was love at first sight.
Unless accompanied by a link, all Caplan quotes are from The Case Against Education.
See the chapter “What I Really Think” for the quote I give above and the preceding sentences I paraphrase. For further evidence of rhetorical support, see here.
Caplan sets his lower bound at 50%. See the section “The Sheepskin Effect”, or find the Solomonic ruling excerpt in this Zvi post.
For another criticism of interpreting the sheepskin effect as all signaling, see Noah Smith, who argues that completing just a little bit more schooling isn’t costly enough to explain such a large wage premium.
Acemoglu models the labor market this way in a recent paper,
Obedience is a useful characteristic for employers, especially when wages are low, because independent workers require more incentives.
UVA is one of the few universities with an honor system (most likely there are between 60 and 100, with the number decreasing over time).
From the Bicentennial report cited above,
The significant spike in number of dismissals in 2001 comes from an influx of reports from a single class… [one] professor for the introductory physics class “How Things Work” created an algorithm designed to flag sections of papers submitted for his class that were substantially similar. The professor discovered a large number of identical passages and reported over 100 cases in a single year. As a result of those cases, 48 students were found guilty at a hearing or left admitting guilt.
This is in line with Bryan’s own recent assessment—“colleges generally give cheaters a slap on the wrist”.
Bryan often presents this point as the core of his broader argument. See Bryan’s opening paragraph at the Times,
My argument in a nutshell: First, everyone leaves school eventually. Second, most of what you learn in school doesn’t matter after graduation. Third, human beings soon forget knowledge they rarely use.
The word overlearning doesn’t appear at all in Bryan’s book, which is a puzzling omission given that he has blogged a lot about this concept himself in the past and acknowledges its potency:
As a previous post documented, students who master a subject by overlearning have excellent retention.
If memory serves, Bryan’s “The Myth of the Rational Voter” would classify much of the knowledge tested as trivia that doesn’t pass cost-benefit for an individual to bother learning because it doesn’t increase their human capital (and hence their wages).
A point that Bryan himself acknowledges in the context of NAAL testing,
“Illiterate and innumerate compared to what?” is a fair response. Conceivably, in the absence of English and math courses, all Americans would be “Below Basic” in all three categories. From this perspective, the NAAL puts a fairly high upper bound on schools’ total effect on Americans literacy and numeracy. Eighty-six percent of Americans exceed “Below Basic” for prose; 88% exceed “Below Basic” for documents; 78% exceed “Below Basic” for quantitative. For each of the three categories, 13% are actually “Proficient.” While these results are meager given the typical American student’s years in English and math, they’re way better than nothing from employers’ point of view."
Nick’s insights come in three speeds: there is the Medium article version which was written earliest and is the most accessible (but barely touches the econometrics), there is an official article published in Empirical Economics in 2020 which is very thorough and a little mathy, and there is a working paper that is essentially in between the two. Here is another excerpt from the last of these:
In this framework, I show in Section II the conditions necessary to identify the human capital or signaling shares of the return. I then argue in Section III that these conditions cannot be realistically met for three reasons:
There are too few observable mediating variables that can be assigned to only one of human capital or signaling
Both theories place heavy emphasis on unobservable mediating variables which prevents falsification
Situations in which all these concerns can be overcome are too heterogeneous to be able to build a general model of education returns.
(formatting adapted for readability)
There’s also a Twitter thread if you’re into that. You can follow me here.
This is because signaling improves the match quality between workers and jobs, increasing the productivity of the overall economy. See NHK’s medium piece for a simple numerical example, CTRL-F “doubled production”.
“An identity map is a function that always returns the value that was used as its argument, unchanged.” [Wikipedia]
Another interesting point worth mentioning is that just as signalling is not necessarily wasteful, human capital formation is not necessarily productive either.
From NHK,
Although the argument is more strained, it is possible for education to reduce productivity by building human capital if the skills attained allow graduates to enter industries that rest on rent-seeking or negative externalities.
See “Measured Learning”.
Once a few people give up on their resolutions, it’s easier for others to lose motivation and quit, in an ongoing spiral. Though the tradition seems to do a lot of a good anyway!
See for instance the growing evidence on the importance of role models.
Photo Credit: Yuhan Du
There has been a related message that parenting quality is not very important, based on lack of effect on IQ. That argument suffers from a lot of the same flaws.
Hahaha, the oblique reference to "the black robes of a false priesthood!"