What this covers
Bryan Caplan joins Russ Roberts to defend the thesis that formal education functions primarily as a sorting mechanism rather than a vehicle for learning. The conversation centers on the claim that once ability differences are controlled for, roughly 80 percent of education's earnings premium flows from signaling—the credential's power to certify pre-existing traits like intelligence, conscientiousness, and conformity—rather than from human capital acquisition. Caplan marshals evidence across economics, psychology, and educational practice to argue that schools are vastly more effective at labeling workers than at teaching them to think, reason, or apply knowledge across contexts.
The conversation moves through several empirical and conceptual grounds. On the empirical side, Caplan highlights the sheepskin effect—the outsized payoff for completing a degree relative to attending classes—and the observation that curricula remain largely irrelevant to modern labor markets, both of which sit uneasily with human-capital explanations. On the psychological side, he examines a century of research into transfer of learning, which has found vanishingly little evidence that schooling teaches generalizable thinking skills. He proposes thought experiments, like the diploma vs education thought experiment, to isolate which aspect of schooling people actually value. Roberts enters as a skeptical interlocutor, raising questions about whether IQ has been overrated, whether deep understanding requires time and narrative rather than immediate recall, and whether common-sense concepts like opportunity cost truly remain inert after formal study. The exchange also touches on ability bias, the limits of falsifiability in measuring learning, and whether conformity and reliability—rather than raw intellect—are what degrees genuinely signal to employers.
Caplan argues that most of the earnings premium from education comes not from skills acquired (human capital) but from signaling — schooling certifies pre-existing traits like intelligence, conscientiousness, and conformity — so much of formal education is a wasteful, expensive sorting mechanism rather than genuine learning.
- After subtracting ability bias (~45% of the apparent effect), roughly 80% of the remaining causal effect of education is signaling rather than human capital
- The sheepskin effect — outsized rewards for completing degrees — and irrelevant curricula are hard to explain on human capital grounds
- Educational psychology has searched for 'learning how to learn'/transfer effects for a century and found very little
Schools signal conformity and willingness to comply with institutional demands rather than teaching genuinely transferable thinking or practical skills.
- One of the main things education signals is conformity to social expectations, which is distinct from being hard-working; employers value workers who put the job first and are agreeable, and avoid hiring defiant personalities, so finishing a degree certifies willingness to comply with institutional demands.
“the sheepskin effect is a clue that one of the main things that it signals is conformity to some social expectations. Which isn't quite the same thing as being hard working.”
- Educational psychology has searched for a century for evidence that schooling teaches general 'learning how to learn' or transferable thinking skills and found very little; people are bad at taking what they learn in one context and applying it to a different area or a practical task (e.g., knowing the Pythagorean Theorem but not using it to build a birdhouse), so the claim that education 'teaches you how to think' is mostly wishful thinking.
“these guys have been studying this, learning how learn, learning how to think, idea for a century... there's very little to it. It's mostly, it's mostly wishful thinking.”
- The IQ gains from education appear largely 'hollow' — school functions like IQ test prep that raises measured scores via teaching to the test, but does not translate into genuinely greater practical intelligence, so the 1-3 point per year gain mostly doesn't mean anything in real-world matters.
“the gains look very hollow. Which means that it looks mostly like it's just teaching to the test... it functions like IQ test prep. But to the idea that it's genuinely making them smarter in some practical way, is... probably just hollow.”
- If no designable test — however hard one tries — can demonstrate that a person learned something, then they didn't learn it; claiming a person has been transformed in a way that is inherently untestable and unobservable is wishful thinking, not evidence of learning.
“if there is no designable test that can show that people learn something, then they haven't learned it. You might say the test is bad, in which case I would say, 'Fine. Design a better test, and then show it to me.'”
- Advanced science students who have spent years studying the experimental method fail to apply correlation-versus-causation reasoning when asked about a practical issue outside their exact subject (such as the effect of sleep on student performance), offering only platitudes — showing they have not internalized the reasoning the way someone like Roberts spontaneously does.
“I cite this one study where they asked, you know, advanced science students in college about the effect of sleep on student performance. And... all these kids could do was say, 'Well, it can't hurt to get enough sleep.'”
- If economics education genuinely changed thinking, students would routinely walk out of movies they are not enjoying since the ticket money is already sunk — one of the easiest applications of opportunity cost — yet they don't, indicating the lesson wasn't internalized; Roberts counters that walking out misapplies the concept because many unenjoyable activities are worth finishing.
“what fraction of your students walk out of movies after you talk about opportunity cost?... That's like, basic opportunity cost. And yet, how many students are persuaded to do that because they took an economics class?”
Education's payoff mostly comes from the degree itself, not from learning, as shown by the sheepskin effect and curriculum irrelevance to labor markets.
- Most of the payoff from education comes from crossing the finish line — completing the degree (the 'sheepskin effect') — rather than from years of partial attendance; this effect is enormous for college and large for high school, which is bizarre under human capital (why would the final year pay so much more?) but natural under signaling, and ability-bias counter-models fail because measures of ability don't change the graduation-to-regular-year payoff ratio.
“most of the payoff from education comes from crossing the finish lines, from diploma years. Especially for college, the sheepskin effect is enormous.”
- A major reason to favor signaling over human capital is that much of the curriculum is irrelevant to the modern labor market: under half of high school students' time is spent on reading, writing, and math, most college majors (communications, psychology, history) are non-vocational, yet graduates in these majors still earn far more than high school graduates even after adjusting for ability — which makes sense only if the degree certifies rather than trains.
“the first one is just to look at the curriculum and to see how irrelevant to the modern labor market so much of the curriculum is.”
- Faced with the choice between a Princeton education without a diploma or a Princeton diploma without an education, a pure human-capital theorist should obviously prefer the education, yet for most people the choice is a genuine head-scratcher — and the fact that it is confusing rather than a no-brainer is itself strong evidence that education is largely about convincing others you are worthy (signaling) rather than acquiring skills.
“you can have either a Princeton education without a diploma, or a Princeton diploma without an education. Which gets you further on the job market?”
After removing ability bias, roughly 80 percent of education's causal effect on earnings is signaling and only 20 percent is human capital skill acquisition.
- Drawing on industrial psychology, IQ is the single best predictor of job performance across virtually every occupation, and education raises IQ by roughly 1-3 points per year (a newer meta-analysis suggests 1-5), giving human capital a plausible channel; conscientiousness is the second-best predictor, capturing reliability, punctuality, and similar traits.
“the most, the single most important predictor of job performance is indeed IQ... the second best predictor of job performance is conscientiousness, which captures all these other things you are talking about.”
- After subtracting ability bias — which Caplan estimates removes about 45% of the apparent education effect (roughly 30% from IQ and another 15% from work-ethic measures) — the remaining 55% is the genuine causal effect of education, and of that causal effect roughly 80% is signaling and only 20% is human capital.
“what I really say is that 80% of the causal effect is signaling... adding in my preferred estimate for ability bias is that subtracts off 45% of the apparent effect. And then, we have to divide that 55% between 20% human capital and 80% signaling.”
- There are three main explanations for why college graduates earn more: human capital (school trains useful skills that raise productivity), signaling (school certifies that you are a premium worker without necessarily teaching useful skills), and ability bias (the more able would have earned more anyway, so the premium is partly spurious correlation).
“the first story is called human capital... The main story that I am pushing in the book is called signaling... And then, the last story is called ability bias.”
Education is largely a waste funded by taxpayers while benefiting professors, making transparency about this a moral obligation despite personal cost.
- Because the education system is largely a waste of time and money — fantastic for professors but a rip-off for taxpayers who fund it — Caplan sees himself as a whistleblower obligated to inform taxpayers they are being ripped off, even though widespread acceptance of his book would harm his own enjoyable job.
“I do at least feel like I need to be a whistleblower and let taxpayers know they are getting ripped off.”
- Academic incentives are so weak that a normal person given a tenure-track job would do almost nothing — not publish, give nearly everyone A's — and the fact that professors do more than the minimum reflects conscience and genuine love of their subject rather than institutional pressure, which also explains grade inflation tending toward universal high grades.
“think about what a normal person would do, given these incentives. I mean, I think most people, if they were given a tenure-track job, they would do virtually nothing. They wouldn't publish anything.”
- The book describes education as it actually is, not what it could be; most classes are bad classes, and the current education system barely teaches people how to think — even though a well-taught course could do better, that is much harder and not what the system delivers.
“I'm talking about education as it really is. I'm not talking about what education could be... 'Look, does the current education system actually teach people how to think?' And that's what I'll say, 'Barely.'”
- Following the Epicurean recipe of setting expectations very low produces happiness and meaning; Caplan derives satisfaction from a handful of appreciative readers or a single engaged student rather than mass impact, treating that as enough to justify his work.
“I'm a big fan of the Greek philosopher Epicurus, whose recipe for happiness is: Set your expectations super-low. And I try to do that. And it works.”
Economists reject signaling on disciplinary grounds and ideological preference for education rather than on evidence from psychology and sociology.
- Chicago economists like Gary Becker and James Heckman, who taught Roberts, are strong human-capital proponents unsympathetic to signaling, while the signaling theorists who won Nobel Prizes (Spence, Stiglitz, Arrow) were high-status but signaling was never a popular empirical theory — more a gadfly on the side of the profession.
“I don't know that signaling was ever a popular theory, even in the 1970s. So, I think it's important to distinguish between the high status of the theorists--like Michael Spence and Stiglitz who are doing signaling models--and yeah, they got Nobel Prizes.”
- Mainstream labor economists remain skeptical of signaling for several non-evidentiary reasons: the best evidence lies outside economics (in educational psychology and sociology) and is ignored on disciplinary grounds; distinguishing signaling from human capital requires extra work beyond a simple income-education regression; a functionalist bias assumes whatever exists socially makes sense ('how come every country does it this way'); researchers were lifelong good students who romanticize education and engage in 'intellectual incest' by using what they studied; and left-leaning enthusiasm plus bipartisan appeal make subsidizing education attractive.
“a lot of the best evidence comes from outside of economics. And economists, like almost everybody else, having not invented [?] syndrome... they don't actually care to study educational psychology.”
- Research is most trustworthy when investigators find conclusions unwelcome to themselves — educational psychologists are idealists who love education and want to prove it works, yet after a century of trying every way to detect transfer learning they glumly conclude it isn't there, which is exactly the kind of disconfirming result against self-interest that should be believed.
“the people who go into it are educational idealists. They love education; they want to find out that education is wonderful. And yet, you go and read the Educational Psychology, and then they are glumly saying... 'I can't find anything either.'”
Signaling value of credentials persists even when undeserved because employers retain workers for legal and emotional reasons rather than firing them.
- Although one might expect a fake or undeserved credential's signaling value to erode quickly as employers learn the truth, it actually diminishes only slowly — over decades — because firms resist firing for legal reasons and basic human empathy ('once you name the puppy'), unhappy employers help bad workers leave by hiding the truth ('de-hiring'), and replacement costs make it cheaper to tolerate a worker performing below expectation.
“there's actually a lot of resistance to firing people in modern economies. Some of that is legal... But a lot of it is just basic human empathy, where once you name the puppy, once you get to know another human being, you don't want to go and fire them.”
- Students cheer when an entire class is canceled — not merely when they individually skip — because what matters is that no classmate gains an advantage and employers will never know, which fits the signaling model better than a human-capital story where missing instruction would be a loss, or a simple laziness/myopia story.
“What students like is when the whole class is canceled, so no one has to go... for school it matters, because you say, 'Well, now... employers will never know.'”
- Even granting education credit for 100% of what people know, the amount adults actually retain is shockingly small — for example, about half of Americans cannot say how many senators each state has — demonstrating that much teaching is in vain because the knowledge isn't held as adults.
“when half of Americans can't do that, I want to say they don't know it. In any sense. And all the efforts to teach them about the number of senators per state were in vain, if they do not know it as adults.”
- Modest child labor is unfairly stigmatized: working a few hours a week in a job a child doesn't mind teaches useful skills and earns money, and the standard objections (it distracts from school, it robs children of childhood, it is boring and confining) apply equally to school itself — revealing a double standard where children's suffering in school is acceptable but unpleasant paid work is condemned; from a signaling view, work experience is socially preferable because it primarily acquires skills rather than just showing off.
“why is it okay for a kid to suffer and be bored in school, and to have their movement confined and not to be free to act like a kid; but they can't do it for money?”
- Most people who read insightful work like Tetlock's Superforecasting find it interesting but return to their normal lives unchanged, whereas genuine internalization requires treating new true knowledge as a moral imperative to repent and restructure one's daily decisions — a rare 'Puritanical' response that explains why reading rarely transforms thinking.
“Most people read Tetlock's Superforecasting and say, 'Oh, yeah. So interesting...' And then they go back and live their normal lives.”
- Economists focus on means rather than medians because all their econometric methods are built around means, even though for income data the mean is distorted by a numerous right-hand tail of high earners among college graduates.
“it's not the one that economists focus on, because all of our econometrics are set up around means.”