Category: Reviews

Most people believe education increases students’ skills, and thus, education is a way to improve your life and career through learning how to be an engineer or a… (checks most common majors) …business…person… psychologist. Bryan Caplan’s latest book disputes this claim, arguing instead that education, especially college and graduate degrees, but even high school, is largely signaling, and not skill building. What does this mean?

Caplan believes employers look at your college degree and GPA and see that you are smart, hard-working, and conform to social norms, and it is that information, not skills you have, which gets you hired (mostly). Prior to reading this book, I held views that higher education had some issues, and I was particularly suspicious of the increasing cost of a college education that has significantly outpaced inflation. With new technological breakthroughs that make teaching much easier, is it really that much more expensive to teach undergrads than it was 30 years ago?  With all the new amenities American colleges are adding (study abroad programs galore, student life funding, food, study rooms, etc), it seems like most of the money is not going to teaching, but shouldn’t a free market among colleges force them to compete on price?

The Case Against Education presents a solution to this puzzle, and much more that I had not considered. Signalling is about relative appearance. If the US and USSR both have ICBMs, no one will fire them out of fear of retaliation. But if one superpower develops submarine launched missiles, then there is a relative difference, and now the other must develop submarine launched missiles as well, or they risk being destroyed in a submarine first strike. Afterwards, the new equilibrium re-establishes a balanced peace, but both countries have wasted time and resources building submarines only to get back to the exact same situation. College is an academic arms race; if everyone could agree not to go to college, we’d be in the exact same place but without having to pay all that expensive tuition.

How can this be though? We know college raises people’s earnings, ergo shouldn’t the market fix this by finding better ways to figure out if prospective employees are worth hiring? Not really. College degrees give employers a “free” way to see which prospective applicants would be best to hire, paid for entirely by the applicants (or the government). Most jobs require pretty specialized skills that can only be learned on the job, so what employers are really looking for is intelligence and commitment. Having a good GPA (and getting into a good college) indicates not only intelligence but also the ability to work hard to achieve long term goals. This system is wonderful for employers, and they have no incentive to change their hiring practices to target non college grads who would probably be more productive with direct on the job training. Those hires are riskier on average as some hires will have been unable to enter college, rather than more interested in work. Employers essentially outsource job screening onto colleges at no cost to them. Colleges also have no incentive to fix the system of course, and the government funding that helps make the problem worse doesn’t respond to financial incentives, especially as education is pretty popular. Students can’t escape either as they are stuck in the arms race, and thus the problem persists.

We Don’t Learn Much From School

Nonetheless, the signalling case may not be intuitive. Most people know they learned things in school, after all, we all had tests on the material! I earned a pretty good GPA at my school, and I had to learn a bunch of stuff for it! The Case Against Education‘s second chapter addresses this quite aggressively, and the following few paragraphs are discussions arising from that chapter. First, it points out that a large fraction of what we learn in school isn’t very applicable to our jobs. I really enjoyed my social studies classes, and I certainly use some of the things I learned when writing this blog, but I do this as a hobby. In my actual employment, I have never needed the years of history, government, or economics classes I took. My counter is that Caplan just asserts some classes are worthless. He has some numbers to back up the particular claim that despite most American high schoolers being forced to learn a foreign language, very few actually speak anything other than English as an adult. There also don’t seem to be many useful careers in the social sciences outside academia. Maybe you need to study civics to get a job in government, although not for most basic bureaucracy desk jobs. I think some more concrete numbers would help his case, although I concede he’s probably right on most accounts.

(Caplan also points out that some “useful” classes, like Math, aren’t necessarily that useful if you break them down; almost everyone takes Geometry, yet very few people need to reason about triangles in their everyday life.)

There’s also an excellent counterargument to the idea that knowledge might become useful later (e.g. “Latin might be helpful if I need to know meaning of an unknown word”). Hoarders make the same argument, “I might need this 17th water bottle someday!”. Education costs resources, and we shouldn’t be purposefully spending them on concepts that may never be useful to students.

Caplan also extensively evaluates whether students actually retain what they learn in school years later. The results are incredibly dismal. Even basic literacy and numeracy, which Caplan argues are the most practical skills we learn in school, are pretty awful when tested. Civics, Science, and foreign language skills were similarly terrible. It’s even suggested that 38% of American citizens would fail the citizenship exam.

The claim that school teaches you how to reason or how to think and analyze is also refuted. Moreover, as I would add, if that’s the main benefit of school, why not teach that directly instead of lots of classes based off of memorization? There may be some argument against the exact examples and studies Caplan uses; perhaps the ability to apply statistical knowledge in non-academic areas is slightly better than the book suggests, or perhaps a year of education raises your IQ slightly more than a few IQ points higher, but overall these themes are hard to overcome; students don’t retain much information from decades of education, they don’t become brilliant by being in school, and they don’t learn skills they eventually use on their jobs.

More Signalling Evidence

Other interesting observations by Caplan that impressed me were that top educational institutions give their classes away for free. You can watch many of the best lecturers online without paying anything, but you can literally walk into Duke or Harvard and watch lectures from professors. If schools were charging for the education, there would be barriers to getting into classrooms, but there aren’t, because that’s not why you pay to go to college; you pay for the degree. Also of note, in Chapter 3, Caplan analyzes whether some college degrees are useful in building skills while others are not, which might indicate that the “human capital story” could be true, just not true for every major. He points out that even if you are mismatched in your career from the major you studied in college, you often earn significantly more money than a high school graduate.

This is clear evidence for the signaling model, but time for a little bit of pushback. Caplan estimates signaling’s share at 80% of the value of education. However, this obviously changes with subject level. In the wheat-chaff section of Chapter 3 I was referring to in the last paragraph, the book states that engineers see a 20% decrease in earnings if an engineering degree holder works in a non-engineering field. But the entire premium of engineering is about 60% above HS graduates, adjusting for ability. This means a 20% drop in earnings brings us back down to 28% above HS grads. That means about 53% of engineers’ higher earnings are due to skills, since they lose half of their bonus above HS grads if they are working in a field without those skills. Not to mention there could be some skills engineers pick up that helps them in other areas, despite Caplan’s points otherwise. Of course, humanities majors are much worse for this case, since many of them see zero loss of earnings if they do not work in their field. Anthropology, liberal arts, sociology thus might be demonstrations of pure signalling. Interestingly, their college premiums are pretty close to engineers’ premiums if the engineers are working outside their field.

The point I make here isn’t that Caplan is necessarily wrong about 80%, but rather that I thought this particular discussion clarified where these numbers might be coming from and what the interaction is between my prior concept of “school teaches me useful skills” and this new concept of “school is mostly signalling”. In other words, these compensation numbers indicate that while there are some skills taught in school, large swaths of students are taking classes that do not teach many skills. Signalling may be argued as a reflection of “people study useless subjects”, rather than “school is inherently bad at transferring skills”, which may provoke outright dismissal by some readers.

Another counterpoint to Caplan is that Sheepskin Effects, the effects of graduation on earnings, may be a reflection of ability bias, rather than all signaling. This blog post discusses a possible method, where people who make it to 3 years of college and then drop out may be disproportionately people who could not complete the hardest classes, saved them for the final semesters, and then failed them, causing them not to graduate. Had they spread them out, perhaps they would have failed out earlier, dragging down first and second year benefits, while allowing third year benefits to rise.

The problem is that signalling would still make up a massive fraction of education, even if Sheepskin Effects are partially reflection of ability; Caplan doesn’t discuss professional schools, as they tend to be pretty good about teaching skills, but he also doesn’t mention that even for medical school, the vast majority of required undergraduate classes in the United States are not skill-building. Calculus, physics, and organic chemistry are not necessary for practicing medicine, yet they are still required.

The Case Against Education also discusses how you might calculate selfishly whether college or advanced degrees are worth pursuing. Caplan even includes helpful spreadsheets that you can manipulate yourself to calculate the returns to your own education.

On the other hand, the following chapter on social returns asks if perhaps there are positive externalities to education that might be helpful besides teaching people useful career skills. I found the section that it was hard to find nation-level benefits to economic growth surprising at first, but more realistic given slow US GDP growth despite higher and higher educational attainment.

There’s also a brief, but thought-provoking section in Chapter 6 regarding the impact of education on democracies and policies. Education correlates with higher political engagement, although whether that’s due to ability bias or actual impact of education is not dealt with. Instead, Caplan asserts that whether education’s impact on policy is good or bad “…the social value of participation hinges on the quality of participation”. This is a statement I strongly agree with, but I’m not sure most people would necessarily endorse. He rightly points out that the quality of participation is inseparable from the question of the quality of policy itself, which is way too big a topic for an education book. Nonetheless there is an implication that the general promotion of civic participation is not necessarily good for society, and I suspect such a notion is controversial.

Solutions

Finally, towards the end of the book, Caplan gets into his proposed solutions for the problem of signalling. The Case Against Education makes a strong argument that education doesn’t have great payoffs and wastes resources on relative signalling, and so Caplan suggests we reduce government subsidies for education. Notably, from a libertarian perspective at least, Caplan’s argument rests on the idea that the education free market itself wouldn’t be optimal, as signalling would actually cause an overconsumption of education over what is socially optimal. He actually has a section in Chapter 7 discussing if it would actually make sense to tax education. He makes the cursory libertarian argument that the government should leave people alone unless we know policy interventions will be highly successful. This is probably fair, but if we were to miraculously find ourselves in the position of having no government education subsidies, I suspect that some taxation of signalling heavy education might be socially ideal, if economically and politically untenable.

The book is also aware of how unpopular any calls to reduce education subsidies would be. Nonetheless, Caplan makes a good point that the proper response to poor education effects implies we should stop bad policy until we figure out better ones, not continue them while we debate alternatives. At the very least, college subsidies should be ended. Tuition will rise, but pushing more of the burden on students is what we want; education should only be undertaken from a social cost-benefit analysis if its benefits outweigh its costs. An excellent way to do this is to force individuals to undertake the costs, since they will be incentivized to go to college only if they can study something that will pay for it. This will negatively impact humanities enrollment, but right now much of humanities coursework is subsidized by the taxpayer and seems to be largely signalling. We should save the money.

The Case Against Education also makes the point that the poor are by far the hardest hit by credential inflation. Reducing government subsidies means the poor will have a much harder time getting to college, but it also means you should see a systemic decline in the necessity of college degrees for jobs that don’t need them.

Caplan also devotes a chapter to the benefits of vocational education, and getting young students (especially those who aren’t doing well in school) on the job experience as early as possible. I don’t have much to add, but it seems disturbingly obvious; if school doesn’t teach us much that we remember, and if there exist jobs that aren’t taught in school, but can be taught with work experience, we need to change the cultural aversion to vocational education ASAP. Additionally, I’ve been following many of my friends in medical school and, it’s incredible to me how vastly it differs from the requirements to enter medical school. Important parts of medical “school” is literally on the job apprenticing with actively working doctors, nurses, residents, etc. Meanwhile, med school undergrads spend 8 semesters learning things that are virtually useless in their planned vocation. It’s absolutely bizarre, although well explained by the signalling model.

Finally, I want to briefly discuss Caplan’s explanation for why no one else is talking about education with similar critiques. He places most of the blame on social desirability bias; basically, it’s unpopular and costs us socially if we critique popular views as incorrect. This story make some sense to me: calling for education cuts is often seen as heartless and evil, yet so are lots of calls to cut government spending, and there are plenty of libertarians and fiscal hawks that are ok with taking those views. I think a significant part of the puzzle arises from the fact that the signalling model is not widely known or understood. It’s also counterintuitive, since we have quite plausible explanations for many things signalling suggests, e.g., people with higher education get paid more because education imparts skills, no signalling model required.

Overall, this book was really interesting and has convinced me that signalling is a substantial fraction of the benefits of education. I feel like there was no definitive place where Caplan calculated exactly why he thought signalling should account for 80%, but doing some of my own calculations around education premiums for workers working inside and outside of fields where their degrees were focused, I can see how there is a chance signalling could indeed be as high as 80%. Nonetheless, even if the proportion is much lower, say only 40% or 30% that would be incredibly wasteful for a trillion dollar industry. After reading The Case Against Education, I feel that a significant cut to at least college education subisidies is probably warranted, and further research into the usefulness of education and the signalling model is vital.

Leave a comment on reddit.