Book Summary: The Case Against Education by Bryan Caplan

I didn’t plan this, but my latest summary for The Case Against Education: Why the Education System Is a Waste of Time and Money by Bryan Caplan comes right as the school year begins for those in the Northern Hemisphere.

Estimated reading time: 29 mins

Buy The Case Against Education at Amazon (affiliate link)

Key Takeaways from The Case Against Education

• Education pays. High school grads earn more than dropouts; college grads earn more than high school grads; and those with advanced degrees earn even more.
• There are three different explanations for why education pays:
  • Human capital. Education equips people with the skills or traits that employers value.
  • Ability bias. The people who choose to pursue more education just happen to be the type of people employers value more—education didn’t make them valuable to employers.
  • Signalling. Education certifies that people have skills or traits that employers value. This is similar to ability bias, but education still provides some value under signalling because certification itself is useful.
• Caplan argues that the signalling theory has been underrated and accounts for between 33% to 80% of the total return to education.
• The traits that education signals are not just cognitive ability but also conscientiousness and conformity, which creates a catch-22—you can’t signal conformity in an unconventional way.
• The selfish return from education varies for different types of students at different stages of education. Based on US figures, Caplan calculates that:
  • High school is worth it for almost all students.
  • A bachelor’s degree is only a good deal for Excellent and Good Students.
  • Master’s degrees are at best an average deal even for Excellent Students.
• However, the social return from education is different from its selfish returns. Once you account for signalling, the social returns from education are much lower for all students (except for Poor Students finishing high school).
• The policy recommendations Caplan advocates are:
  • Cutting government funding for education;
  • Putting students on vocational education track earlier; and
  • Loosening child labour laws.
• Caplan truly loves education and believes it can enrich the mind. But the current educational system doesn’t do this particularly well, and it’s no good force-feeding great ideas and culture to students who don’t want to be there.

Detailed Summary of The Case Against Education

Education Pays

Education definitely pays:

• High school grads earn around 30% more than high school dropouts.
• Bachelor’s degree holders earn around 73% more than high school grads.
• Master’s degree holders earn around 122% more than high school grads.

These are raw numbers, without correcting for things like cognitive ability and family background. But even after you correct for such factors, a sizeable earnings premium for education persists.

Why Does Education Pay?

Caplan covers three possible explanations for why education pays:

1. Human capital. Education increases income by equipping people with skills or traits that employers value.
2. Ability bias. Education does not equip people with skills or traits that employers value. Instead, people who are inherently more skilled or productive happen to be those who choose to get more education.
3. Signalling. Education increases income by certifying the skills or traits that employers value.

Both ability bias and signalling agree that employers value skills and deny that schooling enhances workers’ skills. The difference is over the visibility of skill:

• Signalling suggests that skills are often invisible to employers. Education can therefore increase a worker’s income by certifying that they have certain desirable skills or traits.
• Ability bias by contrast is just a type of selection Bias. If the earnings premium was purely due to ability bias, it would disappear once we hold ability constant.

Caplan focuses on the signalling model but acknowledges that the truth is almost surely a mixture of these three explanations. He thinks signalling accounts for between 33% to 80% of the education earnings premium rather than 100%. [Even though he says “signalling” accounts for this much, I think he’s actually lumping signalling and ability bias together here. Put another way, I think Caplan’s claim is that human capital accounts for between 20% to 67% of earnings premium.]

The Signalling Model

Signalling models say that there are different types of people, who differ in some non-obvious way (e.g. IQ, conscientiousness, family background). But there is some visible difference between these types on average, which acts as a signal for anyone wishing to discriminate between the different types.

A signal doesn’t have to be definitive. It just needs to be true on average to be useful.

What does education signal?

Caplan argues that education signals a package of at least 3 desirable traits to employers:

• Intelligence — cognitive ability and IQ.
• Conscientiousness — discipline, work ethic and grit.
• Conformity — submission to social norms. Modern model workers are team players who know and do what is expected, even when it is not made explicit.

If education only signalled intelligence, the explanation falls apart because there are much cheaper intelligence signals like IQ or SAT tests.¹Some people point out that it’s technically illegal to use IQ tests to screen employees in the US due to disparate impact laws. However, Caplan doesn’t place much weight on this because the law is poorly enforced and around 10 to 30% of large employers openly use cognitive ability tests.

But conscientiousness and conformity are hard to signal in other ways. A signal of conscientiousness must be costly and gruelling. With conformity, there’s a catch-22—you can’t signal conformity with unconventional signals. For example, while online education is a great way to build human capital, it doesn’t signal conformity in the same way as a college degree. So employers don’t take people with online credentials seriously.

Credentials help you get your foot in the door. This is particularly hard with your first good job. Then, as your employer gets to know you, credentials matter less. However, employer learning on the whole seems slow. Signals can affect pay even after someone’s been hired because pay scales and across-the-board pay raises mute the effects of job performance on pay. Employers are also reluctant to fire workers. Most employers will retain subpar employees indefinitely unless their business in jeopardy. For most workers, employer learning takes years or even decades rather than months, so signalling is still important.

Many things don’t make sense under the pure human capital model of education and can only be explained by signalling. These include:

• the major disconnect between what is taught in schools and the skills required on the job;
• students’ attitudes to school; and
• the sheepskin effect.

Disconnect between what students learn in schools and the skills required in jobs

There’s a puzzle in that there’s a weak tie between academic curricula and the labour market (i.e. many skills taught in school are useless in jobs) and a strong tie between educational and professional success. In other words, education has a larger effect on income than on job skills.

Caplan accepts that K-12 education teaches some useful job skills like literacy and numeracy. But there are many classes like the arts and foreign languages that teach little to no useful skills in the modern US economy. Even science classes are only useful for the tiny share of students who go on to pursue careers in science or engineering.

In college, the link between what’s taught and what’s required in jobs depends heavily on the major. Majors like business, education, public administration and engineering have medium to high levels of usefulness. Yet even majors with that have little to no relevance to most jobs—such as fine arts, philosophy, women’s studies, history, psychology and theology—come with significant wage premiums. Majors like anthropology, archaeology, English, liberal arts, sociology, history, communications still boost earnings by around 30% compared to high school graduates. Even the worst-paid majors have earnings premiums of around 20%. (These figures are corrected for ability bias.)

Malemployment (when a worker is overqualified for their job) is high, with estimates ranging from 10% to 35%. But malemployed workers still enjoy an earnings premium. Even in occupations like bartender, cashier, cook, janitor, security guard or waiter, which have little or no plausible connection to academia, high school grads still out-earn dropouts and college grads similarly out-earn high school grads.

Caplan blames credential inflation—as credentials become more common, you need more and more education to signal that you are a good worker. In 1950, only 33% of adult males had finished high school, but male workforce participation was actually higher.

Students forget most of what they learn

Much of what we learn is “inert”. Research shows that even students who do well in exams frequently fail to apply their knowledge in the real world.

While education does seem to affect IQ, it only improves performance on narrow IQ tests but not on other cognitive challenges. Many programs designed to boost children’s IQ find that gains disappear after the program ends. During high school, students regress by an average of 1 month during the 3-month summer vacation. Retention is best when people can apply what they’ve learned shortly after learning it. However, there aren’t many opportunities for people to apply what they learn in school because so little of what is taught in schools is practical.

Signalling explains why failing a class has vastly different career consequences from forgetting what you’ve learned. Employers penalise a job applicant who couldn’t graduate from high school because they failed Spanish, but they don’t care if an otherwise identical high school graduate can’t remember a word of Spanish.

Doesn’t education teach intangible skills like critical thinking?

Caplan doesn’t believe that educators are great at teaching intangible skills like “how to think” when they fail so badly at teaching the tangible, self-contained goals (like key facts about science or US history) that they are tasked with. Most people barely learn what they are taught in class, so it seems doubtful they’d be much better at applying it outside the classroom.

Moreover, evidence on transfer of learning is very poor and the effect of education on general thinking skills seems to be weak.

What about discipline and social skills?

Work also teaches discipline and social skills. Both school and work teach people to follow orders and cooperate with others, and there’s no reason to think education would be any better at teaching these skills.

If anything, work teaches these skills in a way that is directly relevant to the world of work, while school teaches them in a more abstract, indirect way. Caplan argues that college today is too easy—while the typical full-time college student in the 1960s spent 40 hours per week studying, this had fallen to only 27 hours at the time of Caplan’s writing. So instead of preparing students for the world of work, school actively underprepares them.

But you never know what skills might be useful

Even if it’s hard to predict what skills will be needed in the future, that doesn’t mean we should prepare students for occupations they almost certainly won’t have.

It’s true that academic professors occasionally come up with insights that revolutionise an industry. But it seems like the best way to discover useful ideas is to search for useful ideas, rather than searching for whatever fascinates you and hope that it turns out to be useful.

Students’ attitudes to schools

If the human capital model were correct, we should see students try to learn as much as possible during their time at school. But we don’t. Students try to get away with as little work as possible to get the highest grades. They look for professors who award easy grades. They don’t bother to attend classes they’re not enrolled in, even when they can do so for free at top colleges (most colleges don’t require student IDs to attend classes). And they celebrate when classes get cancelled, even though they don’t get refunded for the education they’ve missed out.

When college freshman were asked why they’d chosen to go to college, almost 90% cited “To be able to get a better job” as a key reason. Less than half pick “To develop a meaningful philosophy of life”. [I looked this up and found that 82.5% of students picked “To learn more about things that interest me”, while 72.8% picked “To gain a general education and appreciation of ideas”. This suggests that college students are not quite as careerist as Caplan claims.]

Cheating also challenges the human capital theory. If education was only about learning useful skills, there’d be no point in cheating or in schools trying to prevent it. But cheating makes perfect sense under the signalling model because it undermines the signalling value of a school’s qualification.

Sheepskin effect

Although the earnings premium increases as years of schooling increases, the effect is not evenly distributed. There are huge pay spikes when students graduate from high school or receive a college diploma (even after correcting for ability). This is the sheepskin effect.

The sheepskin effect is one reason why Caplan disagrees with labour economists. Labour economists tend to assume all years of education are equal and don’t factor in the risk that someone may attend high school or college and still fail to graduate.

Measuring ability bias

As noted above, the ability bias model suggests people who are inherently more skilled or productive happen to be those who choose to get more education. So to the extent the earnings premium is due to ability bias, it should disappear once we hold ability constant and compare incomes of people with different levels of education.

A difficulty is that “ability” encompasses both cognitive abilities (like IQ) and non-cognitive abilities (like conscientiousness and conformity). The latter is harder to measure and the research is thinner. Caplan estimates that:

• Cognitive ability bias accounts for between 20 to 30% of the earnings premium;
• Non-cognitive ability bias accounts for between 5 to 15%.

Applying the higher end of those estimates (total 45%) reduces the raw earnings premiums listed above as follows:

• High school grads earn around 15% (down from 30%) more than high school dropouts.
• Bachelor’s degree holders earn around 40% (down from 73%) more than high school grads.
• Master’s degree holders earn almost 70% (down from 122%) more than high school grads.

So while controlling for ability shrinks the education premium, it doesn’t completely vanish, indicating some role for signalling or human capital.

What about the Card Consensus?

David Card, a prominent economist, did a literature review finding little or no evidence for ability bias. Many (perhaps most) labour economists embrace this finding, which Caplan refers to as the Card Consensus.

Instead of trying to control for ability, economists who endorse the Card Consensus adopt quasi-experimental approaches such as looking at the effects of changes in compulsory attendance laws. Caplan criticises these studies because:

• States don’t change laws at random (as would be needed in a true experiment) but for reasons that may be related to the labour market. Some make erroneous assumptions — e.g. one study assumes identical twins have identical ability (Caplan argues the more educated twin is usually the smarter one) while another treats the season of birth as random.

Overall, Caplan argues that the studies underlying the Card Consensus are unconvincing and thinks that directly measuring pre-existing ability is a cleaner approach. He also accuses mainstream labour economists of applying double standards—applying unreasonable high standards to evidence that supports the signalling model (e.g. complaining that ability bias studies can’t correct for unmeasured abilities) but not to evidence that supports the human capital model. They also ignore evidence from other disciplines like psychology, education, and sociology.

Caplan suspects labour economists’ judgments are clouded by their personal experience and love of education. But labour economists’ personal experiences are not representative of the typical student’s. The skills taught in school are far more relevant to a job in academia than practically every other occupation.

The Selfish Return to Education

Caplan does some detailed calculations of the selfish (or private) return to education for different types of students at different stages (in the US). I’ve summarised the calculations in a separate post as it’s mostly only relevant to people contemplating how much education to pursue.

In short, Caplan’s conclusions are:

• High school is worth it for almost all students. The only students for whom high school may not be worth it are those who don’t plan to work full-time after graduation or those with slim chances of graduating.
• A bachelor’s degree is only a good deal for Excellent and Good Students. It probably won’t be worth it for most Fair Students, but it might be for those who choose a lucrative major or particularly enjoy college. Poor Students should not bother with college at all.
• Master’s degrees are at best an average deal for Excellent Students. They are bad deals for everyone else.

The Social Return to Education

If the human capital model is correct, education’s selfish returns are also its social returns. If raising everyone’s education actually makes everyone more productive, society as a whole would be better off.

But to the extent signalling is true, education’s selfish returns are not the same as its social returns. If signalling is about certifying existing traits and skills, it makes education a relative or zero-sum game. Like standing up to see better in a theatre, it can’t make everyone better off.²Even if education just certifies existing traits and skills, there is some social value because the economy is more productive when employers can tell which workers are good and which workers aren’t. However, this is relatively minor and would not justify ever-increasing amounts of education.

Effect of education on national income

One way to compare the selfish vs social returns from education is to look at the effect on personal income vs national income:

• In the US, an extra year of education boosts their personal income by around 5 to 10% on average, but an extra year of national education is estimated to boost national income by only 1.3 to 1.7%.
• At the global level, an extra year of education boosts personal income by 8 to 12%, but an extra year of national education is esimtated to boost national income by only 1 to 3%.

Richer countries are generally more educated, but Caplan thinks this is probably reverse causation—i.e. rich countries like to consume more schooling. There’s also tons of measurement error, which makes it hard to isolate any causation effects. At the national level, it’s not clear that education increases living standards at all.

Calculating the social return to education

Like the selfish return calculations, Caplan’s social return calculations differ for different types of students at different stages.

In terms of the social costs:

• For public K–12 education, Caplan estimates a per-student, all-inclusive cost of $11,298 (in 2011 dollars, excluding food and special education for students with disabilities). [In 2024 dollars, this is around $15,700.]
• For college (both bachelors and masters), Caplan estimates a total social cost per student of $8,279 (2011 dollars). [In 2024 dollars, this is around $11,500.]

In terms of social benefits:

• Earnings premium (non-signalling share). Caplan reduces the selfish earnings premium by an estimate of the share attributable to signalling or ability bias (45% cautious view, 80% reasonable view).
• Reduced crime. While education seems to reduce crime, Caplan assumes a greater role for signalling or ability bias (75% cautious, 80% reasonable) because completing high school seems to drastically reduce crime where ordinary school years do not.

Caplan considers but dismisses several other possible social benefits too ambiguous to include, like political participation and fertility rates. He also excludes peer effects (e.g. college graduates are more likely to marry other graduates) because they’re zero-sum. You’re just reshuffling society into different groups, not creating new value.

Caplan admit that his calculations have all the shortcomings of his selfish return calculations, plus several more around the overall share of signalling, and the effect of signalling on workforce participation and crime.

The results

Caplan’s calculations are very sensitive to the signalling share of education. For example, if education is 80% signalling and a year of education raises an individual’s earnings by $5,000, the social return would only be $1,000. But if education is only 33% signalling, the social return would be $3,350.

Based on Caplan’s figures, I’ve summarised the social returns with different signalling assumptions: 33%, 50% and 80%.

As the table above shows, the social returns get lower as signalling increases.

You may notice the high social returns for Poor Students in high school (even higher than their selfish returns). This is because of crime, which has massive social costs. To the extent high school reduces crime in Poor Students, it can be well worth it. (This doesn’t apply as strongly to girls though since they don’t tend to commit as much crime.)

From the above table, I’ve also averaged the social returns at different signalling assumptions across all types of students.

Policy implications

Every government on earth supports education and people consistently want more funding for it. Claims like “education is the most important investment we can make in our children’s future” are emotionally appealing.

Yet these claims aren’t necessarily true. Even if there are some good reasons for education subsidies (such as irrationality, credit constraints and externalities), people rarely ask, “At what point would education spending be excessive?”

Cut government funding

Overall, Caplan thinks that we should estimate each pro-education policy’s social return and compare that to the market interest rate. Even factoring in several positive externalities, Caplan claims his calculations show low and negative social returns, meaning that education subsidies have clearly gone too far (in the US). [Caplan overstates this. It’s really only true under some signalling assumptions.]

Instead of trying to find out how to increase the social return by improving the educational system, we should first “stop wasting resources”. It’s much simpler and quicker to stop wasting resources than it is to find good ways to use them.

Caplan focuses on two ways to cut government funding:

• Cutting fat from the curriculum. Students spend so much time and energy on “useless subjects” like history, social studies, art, music and foreign language, from kindergarten through to college. We should shut down impractical departments at public colleges and make impractical majors at private colleges ineligible for government subsidies.
• Cutting subsidies for tuition. We should raise tuition for public colleges, turn grants into loans and charge borrowers interest rates. We should also impose some tuition for public high school. As the price of school increases, supply and demand will shift until the social return rises to a reasonable inflation-adjusted level of, say, 4%. [As my above table shows, that’s already met for high school at 50% signalling.]

Caplan admits he is a staunch libertarian and would personally favour full separation of school and state. In his ideal world, all levels of education—primary, secondary and tertiary—would be funded by fees and private charity. However, he acknowledges the argument to stop funding primary education is definitely weaker than for high school and beyond. Since almost everyone in the US goes through primary school, it’s really hard to calculate the returns from it.

What about social justice?

Money currently spent on education could be better spent on fixing social problems directly.

The social justice arguments for publicly funding education only work if education teaches useful job skills. But if education is about signalling, government funding just dilutes the value of the signal and leads to credential inflation. Credential inflation hurts everyone, including the most disadvantaged. In the past, a high school diploma signalled more than it does today because dropouts were more common. Yet dropouts were also far less stigmatised than they are today.

Put more kids on the vocational track

Vocational education prepares students for common jobs. Research suggests vocational education raises pay, reduces unemployment, boosts high school completion rates, and even deters crime.

However, the average vocational student is an academic underachiever before they start their vocational school, so they don’t look good next to the average college grad. After controlling for ability, vocational education actually has bigger effects on pay, employment rates and even high school graduation than academic coursework does. Caplan estimates that vocational education’s social return is at least 4 percentage points higher than regular high school’s. For Poor Students, the social return can exceed 7%.

For egalitarian reasons, some people want to put everyone on the academic track initially and only switch kids to a vocational track if they fail academically. Caplan thinks we should steer academically uninclined kids towards vocational education when they’re around 12. He’s worried that when those kids fail on the academic track, they won’t necessarily switch to vocational training. Instead, they may become embittered and drop out.

Loosen child labour laws

An even better “job training program” is work itself. Unfortunately, the term “child labour” is heavily stigmatised. In the US, it’s illegal. Children under 14 generally can’t work at all (with some exceptions for stuff like family businesses and newspaper delivery). Fourteen and 15-year-olds can only work up to 18 hours on school weeks.

Yet, for some kids, school can be just as tedious and dull as work—and they don’t even learn valuable skills or get paid! One study found that looser child labour laws increased high school dropout rates, yet reduced youth crime.

Some worry that child labour laws can cause vulnerable kids to be exploited. Caplan admits that parental oversight isn’t perfect, but we rely on it in virtually every other domain. The US already allows kids to work in family businesses. So why don’t we trust parents to decide whether to let their kids work for someone else?

But isn’t education good for the soul?

Caplan himself loves education. He’s been “in school” for over 40 years and enjoys his job as a professor. He agrees that education can certainly be enriching, and he personally likes poetry even though it’s not very useful.

However, we shouldn’t force-feed great ideas and culture to students who couldn’t care less. This only sparks resentment. Even though enlightenment is virtually free online, most people don’t use the Internet for that (and the ones that do tend to be working adults). Low consumption of ideas and culture is caused more by apathy than poverty.

Moreover, Caplan doesn’t think the current educational system is particularly good at enriching the soul. The curriculum is “ossified”. Instead of making kids consume “high culture” like poetry or Shakespeare, why not offer options like Japanese graphic novels or films from the 1980s? Even assigning a random Wikipedia article could be more interesting.

Alternatively, just let students play. Free play can be extremely enriching and educational. The more time kids spend in classes, the less they have to explore the world. Even if we want to keep kids in schools to allow their parents to work, we could just expand recess times.

Other Interesting Points

• In the 1960s, Egypt guaranteed every college graduate a government job. By 1988, around 67% of male and 80% of female college graduates worked in the public sector.
• Education and work are near the bottom of people’s most enjoyable activities, but work has a slight edge. Elder care is right at the bottom. (Source)
• Self-rated health is apparently a good predictor of objective health outcomes including mortality.
• The effect of education on religion is unclear. As education goes up, people are less likely to believe in God and the literal truth of the Bible (theological religiosity), but they are more likely to attend church (sociological religiosity). International studies typically conclude intelligence reduces both forms of religiosity but the effect of education is not clear.
• The effect of education on people’s politics is also unclear. Increasing education seems to make people more socially liberal but fiscally conservative, and these two effects offset each other. So, overall, the effect seems weak.
• Educated people around the world have fewer kids. The trend persists even after controlling for income, intelligence, democratisation or modernisation. Those who stay in school the longest are most likely to put off having children (if they have children at all). Caplan suggests that if a government wanted to boost fertility rates, it could just cut its education budget.

My Review of The Case Against Education

I was predisposed to like this book. Signalling resonates with my personal experience (though I’m not from the US) and I’ve often argued that higher education is overrated. In my experience, people who defend higher education often implicitly assume the alternative is twiddling your thumbs for 4 years. When you ask them to consider using the same amount of time and money to travel, try out different jobs, or try (and likely fail) to start your own business, they become much less confident that higher education is a great use of time and money. Personally, the most valuable things I’ve learned have been through self-study, travel or from peers rather than through the formal education system.

And there was a lot I did like about The Case Against Education. I liked that Caplan is not afraid to question things that often go unquestioned. I really liked that he tries to tease out the effects of education for those less academically inclined, because I think it’s relatively easy for people like me to underestimate the costs of education for people not like me.

Unfortunately, I also found quite a lot to dislike about The Case Against Education. For one, it’s repetitive. I don’t have an objection to being thorough. I don’t mind that Caplan refers to multiple pieces of evidence to support his claims, walks through detailed calculations, or anticipates and responds to various counterarguments. Given he was arguing against a mainstream academic view (the Card Consensus), I think this was required. However, there were many times where Caplan makes exactly the same point in different parts of the book and I think his arguments could have been far more streamlined.

More importantly, Caplan makes a bunch of sloppy arguments and poorly supported claims in the book. I found this frustrating because those weak arguments diluted his more compelling ones and hurt Caplan’s overall credibility. I’m working on a separate criticisms post (or posts) which will address these in detail.

Although Caplan’s writing is entertaining and easy to read, it came at the expense of accuracy and effectiveness. If he had truly wanted to effect positive policy changes, he’d have been better off confining his calls for reduced government funding to college and pushing to direct more high school students (and existing funding) to vocational ed. But I can’t help thinking that Caplan enjoys making bold, provocative claims more than he wants positive change.

Let me know what you think of my summary of The Case Against Education in the comments below!

Buy The Case Against Education at Amazon <– This is an affiliate link, which means I may earn a small commission if you make a purchase through it. Thanks for supporting the site! 🙂

If you enjoyed this summary of The Case Against Education, you may also like:

• Book Summary: A Mind for Numbers by Barbara Oakley
• Book Summary: Range by David Epstein
• Book Summary: So Good They Can’t Ignore You by Cal Newport

• 1
  Some people point out that it’s technically illegal to use IQ tests to screen employees in the US due to disparate impact laws. However, Caplan doesn’t place much weight on this because the law is poorly enforced and around 10 to 30% of large employers openly use cognitive ability tests.
• 2
  Even if education just certifies existing traits and skills, there is some social value because the economy is more productive when employers can tell which workers are good and which workers aren’t. However, this is relatively minor and would not justify ever-increasing amounts of education.