Education, Top Stories

Bryan Caplan’s ‘The Case Against Education’ — A Review

A review of The Case Against Education: Why the Education System Is a Waste of Time and Money by Bryan Caplan. Princeton University Press (January 2018) 416 pages. 

Almost no issue unifies commentators across the political spectrum as support for education, though their motivations strongly differ. For the left, affordable education is a great leveler for disadvantaged groups and also a force for cosmopolitanism. For the right, it represents an equality of opportunity that can substitute for a generous welfare state. But almost everyone seems to believe that more people with high-quality education means better and more productive workers. 

Among the exceptions are a minority of economists studying education. Beginning with Nobel Prize winners Kenneth Arrow and Michael Spence in the 1970s, these economists proposed that people with more years of education earn more not merely because of the skills and knowledge they accumulated during their time in school (“human capital”) but largely as a function of the information their degree signals to employers. The Case Against Education  by George Mason University economist Bryan Caplan is the most thorough and compelling synthesis of the research in favor of the signaling model of education. The essence of the book is an empirical claim — that only a fraction of the extra wages that graduates earn can be explained by skills and knowledge — and a set of policy proposals to end wasteful spending on unnecessary education.


The signaling model is compelling to many skeptical autodidacts in part because people spend years toiling in classrooms learning things that their careers do not require. Caplan calls this puzzle the “ubiquity of useless education”, and it is not just a matter of comparative literature majors never using critical theory for their day job. Even heavily quantitative and practical majors like computer science are filled with theory courses that are never used by the vast majority of professional software engineers. Caplan even takes aim high school education, arguing that large chunks of class time are devoted to coursework which only improves job performance in very rare occupations.

Yet there is also a vast empirical literature that analyses the effect of education on wages, and at first glance, it is not kind to Caplan’s signaling model. For example, we might ask if the estimated extra earnings of more educated people (called the “wage premium” in the literature) is confounded by the fact that they also tend to have traits like higher intelligence. David Card, an economist at UC Berkeley, reviewed the literature and found that the bias in the estimate was pretty small. His most convincing evidence looks at the earning difference of identical twins with different levels of education, where the ability bias only explains about 10% of the wage difference. So the rest of the premium must be explained by skills and knowledge, right?

A shallow read of the literature might suggest so, and that education raises human capital by making workers more productive. But the virtue of The Case Against Education is that Caplan synthesizes many lines of research that approach the question with different methods. By showing that signaling is the only model consistent with all the convergent facts, his conclusion is more robust than any that comes from a thin slice of the evidence. Aside from the ubiquity of useless education, he looks at three lines of research most consistent with the signaling model.

First, he looks at how much learning students forget over time. People who take a foreign language or geometry class forget more than half of what they learned within 5 years (though the more math they took, the more they retained). Using the National Assessment of Adult Literacy survey, he shows that only about half of high school graduates achieve “intermediate” or “proficient” mastery of basic quantitative questions (an example of a task of this level is calculating the total cost of ordering specific office supplies from a catalog). Despite most high school graduates taking several years worth of science courses, less than a third know that atoms are bigger than electrons. All this leads Caplan to ask: “Why do students rejoice when the teacher cancels class?” A human capital model where the “capital” is an accumulation of classroom learning is difficult to square with these celebrations.

What if students forget the specifics but learned how to learn and think critically? Here too the evidence is shockingly bad for the human capitalist camp. In quantitative courses, students diligently learn how to solve the problems that appear in their homework sets and tests, but their ability to use this knowledge correctly is highly sensitive to contextual clues. According to one survey, just 28% of students who completed a mechanics course could identify the number of forces acting on a coin midair after it is tossed. There is strong evidence from meta-analysis that years of education raises IQ by 1 to 5 points a year. But an IQ score is just a proxy for the latent g factor, or “genuine” intelligence as Caplan calls it. It may be that students’ IQ scores go up because education trains them to answer the kinds of questions on IQ tests, while the g factor remains the same. Some research suggests this to be the case, and it is certainly consistent with the finding that improvement on cognitive training tasks does not transfer over to other tasks. 

There are other ways that more schooling could make better workers. A good education could boost “non-cognitive skills” by making students more extroverted and agreeable — that is, they’d be more socialized for white collar work. Another possible channel is that education gives people access to better jobs with more skilled peers, whose combined effect might otherwise be mistakenly attributed to the schooling itself. But these are not what education advocates typically point to in making their case for education.

Most of the rewards of education come in the final year

Caplan’s second (and strongest) line of evidence of signaling are “sheepskin effects”: the tendency for the bulk of the rewards of education to come in the final year. (The name comes from the fact that old diplomas used to be made with sheepskin.) In the human capital model, returns should be fairly linear and steady within classes of degrees, since a high school or college senior takes about the same course load as a freshman. A steady increase in human capital should lead to steady increases in wages. But surveying ten studies, Caplan finds that for 9th through 11th grade, each extra year of education raises wages only by about 4%, while finishing the 12th grade raises wages by 16%. Similarly, each of the first three years of college only raise wages by about 6%, while the last year raises them by 30%.

The results are more consistent with a model where students who fail to finish their degree signal a lack of work ethic, conscientiousness, or some other set of traits employers want to avoid. Alternatively, it represents a bias in the kinds of people who drop out, but Caplan cites studies that cast doubt on this possibility.

Last, Caplan looks at the effect of increased years of education on national income. If signaling explains the bulk of the education premium, then more degrees mostly just means inflating away the relative significance of credentials. What we would expect if this was true is that the national return on education would be lower than the return to the individual. Caplan reviews the evidence and finds consistently that an extra year of national education brings a “puny” 1.3% in extra income to a nation (though there are exceptions in this line of research, some careful work finds national returns can exceed private returns under certain conditions). The implication here is that the gains from education are mostly zero-sum, where credentials help educated workers at the expense of others rather than adding much to the economy as a whole.

This arms race for educational credentials in most apparent in Taiwan and South Korea where college attainment has exploded in recent decades. Thirty years ago, about 20% of South Koreans had some tertiary education; today about 70% of 25-34 year olds do. Rather than unleashing extraordinary growth as a human capital model would predict, the results have been devastating to lower skilled Koreans who thought college was their ticket to a better life:

Seongho Lee, a professor of education at Chung-Ang University, criticizes what he calls “college education inflation.” Not all students are suited for college, he says, and across institutions, their experience can be inconsistent. “It’s not higher education anymore,” he says. “It’s just an extension of high school.” And subpar institutions leave graduates ill prepared for the job market.

A 2013 study by McKinsey Global Institute, the economic-research arm of the international consulting firm, found that lifetime earnings for graduates of Korean private colleges were less than for workers with just a high-school diploma. In recent years, the unemployment rate for new graduates has topped 30 percent.

Equally convincing is research that looks at the effect on students when the number of years to achieve an equivalent credential is suddenly reduced or increased for everyone. For example, in the German state of Saxony-Anhalt, the duration of secondary school was reduced by a year. Women’s wages went up and men’s went down only by 2-4%. The opposite of this experiment occurred in China, which added a year of schooling to achieve the same credentials in the 1980s. Economists found only a 2% increase in wages from this change. There are also well-designed studies with the opposite conclusion in more limited contexts: Colombian economics students whose post-university wages dropped significantly after their university eliminated some rigorous quantitative courses as a requirement for graduation.

Policy Implications

The signaling model gives us a fresh perspective on the rapid rise in the cost of university tuition, which have tripled in price over the last few decades (though it is less severe than it looks, since many students don’t pay the full cost). This rise is sometimes blamed on Baumol’s Cost Disease, a phenomenon where salaries of workers who do not become more productive over time still go up because employers are in competition with industries whose workers are becoming more productive. The problem with this characterization is that professors’ salaries have mostly stagnated at the same time, while the number of administrators — who are not so central to building students’ human capital — has exploded. According to the best research on the topic, the rise in tuition is most related to the rise in the college wage premium and availability of student loans. If true, a free college for all plan might just allow university administrators to capture a substantial portion of the subsidies.    

But Caplan is not just content to end the gravy train for useless administrative bloat. He gives a series of policy suggestions. First: remove from graduation requirements “useless” school subjects. Second, remove subsidies for education so that their unsubsidized price greatly reduces the supply of students. He even toys with the idea of taxing education, though he dismisses out of hand on libertarian concerns. His most plausible proposal is to vastly increase enrollment in vocational training programs.

Even for those of us convinced by his empirical arguments about signaling, these policy suggestions are unsatisfying and scary because we have no model of what a developed country without useless coursework and education subsidies looks like. In Switzerland, with its extensive vocational training programs, about 30% of young people still graduate from university. There is some evidence that education accelerated the industrialization process, and without a counter-example, it is not hard to imagine that a less educated population might miss out on future waves of innovation.

Thanks to Caplan’s excellent point that education signals conformity and conscientiousness, not just intelligence, we know what a replacement for the current model won’t look like: a decentralized system of learning with certification tests. This is the hope of many autodidacts who dislike the gatekeeping power universities have over us. But it won’t work in the current equilibrium: these certificates might signal social dysfunction, just like GEDs (General Equivalency Diplomas) are inferior replacements to high school diplomas. One exception is a field like software engineering, where students could put up examples of their code on Github and demonstrate their skills quickly in a technical interview (and where traits like conformity are less important.) In this case, a “teacherless university” is possible, but it is likely impossible to scale to all other fields.

The most politically viable route to defunding higher education arises from the culture wars. Thanks to the rising level of political extremism and hostility to political diversity in the academy, Republican support for these institutions has plummeted. The leftist commentator Freddie deBoer has written: “I am increasingly convinced that a mass defunding of public higher education is coming to an unprecedented degree and at an unprecedented scale…the monsters are coming, and I am afraid.” He might be right, but if The Case Against Education is to be believed, we have a lot less to lose than deBoer thinks.


Follow Noam on Twitter @NoamJStein