commonplace book redux

List of passages I highlighted in my copy of “The Case Against Education”.

On conformity and conscientiousness

Heterodox signals of your strengths, in contrast, automatically suggest offsetting weaknesses. Suppose you scored well on the SAT but never went to college. Employers will readily believe you’re smart. But if you’re so smart, why didn’t you go to college? As long as your conscientiousness and conformity were in the normal range, finishing college would have been a snap. Once employers see your SATs, they naturally infer you’re below average in conscientiousness and conformity. The higher your scores, the more suspicious your missing diploma becomes. […]

The further outside the box your substitute signal of conformity, the more it backfires. Try telling employers, “I’m not Jewish, but I keep kosher to prove I can conform to intricate rules.” They’ll take you for a freak. […]

On employer learning:

Give people a chance, observe how they do, fire them if they don’t measure up: a “Hire, Look, Flush” personnel policy sounds both profitable and fair. Yet group identity and pity get in the way. After a firm hires you, you’re part of the team. […]

Employers do have one guilt-free way to reverse a bad hiring decision. Human resources calls it “dehiring.” Instead of firing the unwanted worker, help them jump ship. Privately urge them to find new opportunities. When firms call for a reference, shade the truth—or lie. […]

For most workers, employer learning takes years or even decades, not months. Two seminal studies of employer learning found that during your first decade in the workforce, the ability premium sharply rises, while the education premium falls 25–30%. A subsequent prize-winning article found the education and ability premiums plateau after roughly ten years of experience; the education premium stops falling, and the ability premium stops rising. […]

The more fundamental reason why signals durably affect pay, though, is employers underreact to what they learn. Why? Because they want to match pay and perceived productivity without seeming unfair. When employers spot poor performance, they could swiftly respond with wage cuts, demotions, or terminations. The catch: such “unfair” measures are bad for morale—and make employers feel guilty. […]

A subpar worker can profit from their fancy degree long after their employer sees their true colors. The degree lands them a good job. As truth unfolds, the typical employer responds with stingy raises, not outright pay cuts or demotion. This slowly erodes the value of the signal, but squeamish firms show mercy long before they sync pay with performance. If and when the employer vows to eject the underperformer, both prudence and pity tell them to informally “dehire” rather than blatantly fire. As long as the subpar worker lands another position suitable for their paper persona, the cycle of disappointment, mercy, and deception is reborn. […]

A litmus test:

Imagine this stark dilemma: 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? For a human capital purist, the answer is obvious: four years of training are vastly preferable to a page of paper. But try saying that with a straight face. Sensible versions of the signaling model don’t imply the diploma is clearly preferable; after all, Princeton teaches some useful skills. But you need signaling to explain why choosing between an education and a diploma is a head-scratcher rather than a no-brainer. […]

Counter-intuitive effects of education subsidies:

Yes, awarding a full scholarship to one poor youth makes that individual better off by helping send a fine signal to the labor market. Awarding full scholarships to all poor youths, however, changes what educational signals mean—and leads more affluent competitors to pursue further education to keep their edge. The result, as we’ve seen, is credential inflation. As education rises, workers—including the poor—need more education to get the same job. Where’s the social justice in that? Imagine the government subsidized wedding rings for the poor. Anyone ready for marriage can go to any jewelry store in the country, knowing—whatever their income—they can buy a diamond ring. The snag: diamond rings are largely a signal of marital commitment. If diamonds were cheap as plastic, other gems would adorn our rings. They’re valuable because they’re costly. Once the government makes them affordable to all, then, diamond rings signal little or nothing. Doesn’t this “level the playing field”? Only for a heartbeat. Once the nonpoor see diamond rings don’t signal what they used to, they procure a snazzier ring to separate themselves from the pack. Thanks to government subsidies, every suitor can afford a wedding ring, but so what? Society is functionally as unequal as ever. […]

Subsidies don’t just hurt the poor by fueling credential inflation. They reshape hiring and promotion to the poor’s detriment. Picture a society where half the population can’t afford college. In this setting, reserving good jobs for college grads is bad business. “There are plenty of qualified candidates who didn’t go to college” is not wishful thinking, but literal truth. Education still signals something, but lack of education is not the kiss of death. When asked, “Why didn’t you go to college?” “I couldn’t afford it” is a great excuse. Heavy subsidies take it off the table.

Does school build human capital?

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 […]

Despite the shortage of long-term retention studies, we can fall back on a compelling shortcut. Instead of measuring the enduring effect of education on adult knowledge, we can place an upper bound on that effect. It’s a two-step process. Step one: measure adult knowledge about various school subjects. Step two: note that schools can’t be responsible for more than 100% of what adults know about these subjects. What people now know is therefore an upper bound on the school learning they retain. My shortcut is easy to implement. Surveys of adults’ knowledge of reading, math, history, civics, science, and foreign languages are already on the shelf. The results are stark: Basic literacy and numeracy are virtually the only book learning most American adults possess. […]

Barely half of American adults know the Earth goes around the sun. Only 32% know atoms are bigger than electrons. Just 14% know that antibiotics don’t kill viruses. Knowledge of evolution barely exceeds zero. Knowledge of the Big Bang is actually less than zero; respondents would have done better flipping a coin. […]

If you throw a coin straight up, how many forces act on it midair? The textbook answer is “one”: after it leaves your hand, the only force on the coin is gravity. The popular answer, however, is “two”: the force of the throw keeps sending it up, and the force of gravity keeps dragging it down. Popular with whom? Virtually everyone—physics students included. At the beginning of the semester, only 12% of college students in introductory mechanics get the coin problem right. At the end of the semester, 72% still get it wrong. […]

Learning how to learn?

The clash between teachers’ grand claims about “learning how to learn” and a century of careful research is jarring. Yet commonsense skepticism is a shortcut to the expert consensus. Teachers’ plea that “we’re mediocre at teaching what we measure, but great at teaching what we don’t measure” is comically convenient. […]

Effects on IQ:

“Flowers for Algernon” is science fiction, but life mirrors art. Making IQ higher is easy. Keeping IQ higher is hard. Researchers call this “fadeout.” Fadeout for early childhood education is especially well documented. After six years in the famous Milwaukee Project, experimental subjects’ IQs were 32 points higher than controls’. By age fourteen, this advantage had declined to 10 points. In the Perry Preschool program, experimental subjects gained 13 points of IQ, but all this vanished by age 8. Head Start raises preschoolers’ IQs by a few points, but gains disappear by the end of kindergarten. […]

In any case, suppose each year of school permanently made you a whopping 3 IQ points smarter. According to standard estimates, this would raise your earnings by about 3%, leaving a supermajority of the education premium unexplained. […]

This section, comparing three models of education (human capital, signalling, ability bias) was excellent:

Given a clear explanation, most people readily grasp the conflict between human capital and its rivals. Yet even experts occasionally confuse signaling with ability bias. Both stories agree that employers value workers’ skill; both deny that schooling enhances workers’ skill. The two stories diverge on the question of visibility. In a pure signaling story, employers never see your skill. So if your skills mismatch your credentials, the labor market rewards your credentials, not your skills. In a pure ability bias story, in contrast, employers see your skill plain as day. So if your skills mismatch your credentials, the labor market rewards your skills, not your credentials. […]

Table 3.2: Human Capital, Signaling, and Ability Bias [WYSIWYG = “What You See Is What You Get.”]

Story	Visibility of Skill	Education’s Effect on Skill	Education’s Effect on Income
Pure Human Capital	Perfect	WYSIWYG	WYSIWYG
Pure Signaling	Zero	Zero	WYSIWYG
Pure Ability Bias	Perfect	Zero	Zero
⅓ Human Capital, ⅓ Signaling, ⅓ Ability Bias	2/3	1/3 x WYSIWYG	2/3 x WYSIWYG

In 1999, a comprehensive review of earlier studies found that correcting for IQ reduces the education premium by an average of 18%. 7 When researchers correct for scores on the Armed Forces Qualification Test (AFQT), an especially high-quality IQ test, the education premium typically declines by 20–30%. 8 Correcting for mathematical ability may tilt the scales even more; the most prominent researchers to do so report a 40–50% decline in the education premium for men and a 30–40% decline for women.

Internationally, correcting for cognitive skill cuts the payoff for years of education by 20%, leaving clear rewards of mere years of schooling in all 23 countries studied. The highest serious estimate finds the education premium falls 50% after correcting for students’ twelfth-grade math, reading, and vocabulary scores, self-perception, perceived teacher ranking, family background, and location. 

A thinner body of research weighs the importance of so-called noncognitive abilities such as conscientiousness and conformity. The results parallel those for IQ: noncognitive ability pays, and correcting for noncognitive ability reduces the education premium. Correcting for AFQT, self-esteem, and fatalism (belief about the importance of luck versus effort) reduces the education premium by a total of 30%. The sole study correcting for detailed personality tests finds the education premium falls 13%. The highest serious estimate says that once you correct for intelligence and background, correcting for attitudes (such as fear of failure, personal efficacy, and trust) and personal behavior (such as church attendance, television viewing, and cleanliness) further cuts the education premium by 37%.

There are admittedly two big reasons to mistrust these basic results: reverse causation and missing abilities. The former could systematically overstate the severity of ability bias. The latter could systematically understate the severity of ability bias. Need we fret over either flaw?

Reverse causation. When you estimate the education premium correcting for ability X, you implicitly assume education does not enhance X. If this assumption is false, correcting for X leads to misleadingly low estimates of the effect of education on earnings. The best remedy for this “reverse causation” problem is to measure ability, then estimate the effect of subsequent education on earnings. Research on cognitive ability bias routinely applies this remedy—and uncovers little evidence of reverse causation. The comprehensive review article mentioned earlier separated studies into two categories: those that measured IQ before school completion and those that measured IQ after school completion. If reverse causation were at work, studies that relied on IQ after completion would report more ability bias than studies that relied on IQ before completion. In fact, both categories yield similar estimates of cognitive ability bias. 

Researchers who rely on the AFQT and related tests reach a similar result: When you correct for cognitive ability in 1980, the payoff for posttest education falls at least as much as the payoff for pretest education. Correcting for mathematical ability in the senior year of high school shaves 25–32% off the male college premium and 4–20% off the female college premium. What about reverse causation from education to noncognitive ability? Truth be told, relevant research is sparse. A few papers grapple with the issue, with mixed results. Most research, however, either measures noncognitive ability and education at the same point in time, or fails to distinguish between the effect of pre- and posttest education. The shortage of evidence hardly shows reverse causation is a serious problem, but caution is in order.

Missing abilities. Correcting for ability doesn’t fully eliminate ability bias unless you measure all relevant abilities. Are there any important abilities we’ve overlooked? Family background—via nature or nurture—is a plausible contender. Perhaps wealthy families use their money to help their kids get good educations and good jobs. Maybe college is a four-year vacation for rich kids—and a status symbol for their parents. Perhaps children from large families get less educational and professional assistance from their parents. Maybe well-educated workers come from high-achieving families—and would have been high achievers even without their schooling. The mechanism is hard to nail down, but most researchers find correcting for family background reduces the education premium by 0–15%. On reflection, though, correcting for family background probably “double-counts.” Both cognitive and noncognitive ability are moderately to highly hereditary, so you should correct for individual ability before you conclude family background overstates school’s payoff. This caveat matters. Rare studies that correct for intelligence and family background find that correcting for intelligence alone suffices. Armed with good measures of cognitive and noncognitive ability, we can probably safely ignore family background. The most troubling evidentiary gap: researchers usually settle for mediocre measures of noncognitive ability. Most studies that correct for noncognitive ability rely on one or two hastily measured traits and find only mild ability bias. Yet when asked, employers hail the importance of workers’ attitude and motivation—and the study with the best measures of noncognitive ability finds large ability bias. Until better measures come along, we should picture existing results as a lower bound on noncognitive ability bias rather than a solid estimate. So how severe is ability bias, all things considered? For cognitive ability bias, 20% is a cautious estimate, and 30% is reasonable. For noncognitive ability bias, 5% is cautious, and 15% is reasonable. Figure 3.1 shows education premiums correcting for both abilities, assuming equal bias for all education levels. […]

Why don’t more companies use IQ tests?

IQ “laundering.” Human capital purists often protest, “Why on earth do workers signal ability with a four-year degree instead of a three-hour IQ test?” My response: employers reasonably fear high-IQ, low-education applicants’ low conscientiousness and conformity. Other critics of the education industry, however, have a more streamlined response: American employers rely on educational credentials rather than IQ tests because IQ tests are effectively illegal. Thanks to the landmark 1971 Griggs vs. Duke Power case, later codified in the 1991 Civil Rights Act, anyone who hires by IQ risks pricey lawsuits. Why? Because IQ tests have a “disparate impact” on black and Hispanic applicants. To escape liability, employers must prove IQ testing is a “business necessity.” […]

Sheepskin effects: graduation is much more valuable than learning

Labor economists normally neglect sheepskin effects. By default, they assume all years of education are created equal, then estimate “the” effect of a year of education on earnings. 2 Yet economists who trouble to look almost always find pay spikes for diplomas. 3 High school graduation has a big spike: twelfth grade pays more than grades 9, 10, and 11 combined. In percentage terms, the average study finds graduation year is worth 3.4 regular years. College graduation has a huge spike: senior year of college pays over twice as much as freshman, sophomore, and junior years combined. 4 In percentage terms, the average study finds graduation year is worth 6.7 regular years. […]

correcting for ability usually modestly cuts the effect of both years of education and diplomas—holding the relative payoff for diplomas steady. […]

The Onion and the politics of education:

Who could oppose investment in our children, our people, our nation, and our future? The Onion, the best parody site ever, once ran an article titled, “U.S. Government to Discontinue Long-Term, Low-Yield Investment in Nation’s Youth.” 8 In it, Secretary of Education Rod Paige takes a calmly analytical approach that would cost any politician their job: “Testing is exactly the sort of research the government should do before making spending decisions,” Paige said. “How else will we know which individuals are sound investments and which are likely to waste our time and money?” […]

Bryan Caplan, The Case Against Education, 2017