Bryan Caplan on College, Signaling and Human Capital

Episode Summary

Executive Summary: Russ Roberts and Brian Kaplan debate the economics of higher education, arguing that the college wage premium is driven less by human capital than by signaling and student selection. Kaplan stresses that graduation, not mere attendance, is what matters, that ability and persistence vary widely, and that the social return to ever-more schooling may be negative because education functions as an arms race.

Main Topics: Rising college wage premium (Priority: 5/5): The conversation begins with the sharp increase in the earnings gap between college and high school graduates, and whether this reflects true productivity gains or other forces. Selection, heterogeneity, and graduation risk (Priority: 5/5): Kaplan argues that college returns vary enormously by ability and likelihood of completion, so averages are misleading and incomplete students often get only a small payoff. Signaling vs. human capital (Priority: 5/5): The central theoretical dispute: whether college raises earnings mainly by teaching useful skills or by certifying preexisting qualities like intelligence, conformity, and work ethic. Sheepskin effects and the value of completion (Priority: 5/5): A major empirical focus is the large premium associated with finishing a degree, versus the much smaller payoff to merely attending college. Critique of educational measurement (Priority: 4/5): The discussion highlights how standard metrics like years of schooling and average returns can obscure underlying differences in ability, major, and completion probability. Policy and the case for subsidies (Priority: 5/5): Kaplan argues that if education is largely signaling, subsidizing more schooling can be socially wasteful because it intensifies a credential race rather than creating net new value. Examples and analogies from labor markets (Priority: 3/5): Marriage, banking defaults, cancelled classes, and free access to Princeton are used to illustrate why credentials and completion signals matter.

Key Arguments: The college premium rose dramatically over time, but this does not prove college itself became more productive; it may mainly reflect changes in who attends and finishes. Average returns are misleading because college students are not a uniform group; ability, work ethic, and persistence differ sharply, and these differences affect both graduation and earnings. The premium to merely attending college is much smaller than the premium to graduating, suggesting the credential itself matters more than accumulated coursework. The sheepskin effect supports signaling: employers reward the diploma disproportionately, not just the learning acquired en route. Many students cheer class cancellations because the labor market rewards the credential, not the lost instructional content, if everyone is equally affected. Education acts like an arms race: if more people get degrees, the credential’s relative value declines, encouraging others to get even more schooling. Subsidizing education can be socially inefficient if schooling mainly reallocates jobs through signaling rather than creating new productive skills. International evidence, major-specific wage gaps, and low returns for non-completers all suggest that conventional human capital estimates overstate the true causal return to schooling.

Data Points: College wage premium: 83% - Average college graduates’ earnings compared with high school graduates in 2011. College wage premium circa 1970: 35%–40% - Approximate college-high school earnings gap around 1970. Return to attending college without graduating: 10% - Earnings premium for college attendance absent completion. Five-year graduation rate for four-year college: 55% - Approximate share of students finishing within five years. List price tuition for main four-year state universities: about $5,000 per year - Estimated average actual tuition after aid and adjustments. Average list price at state universities: $10,000–$12,000 for four years - Rough total sticker price cited for four-year public universities. Marriage cost-of-living savings: about 35% - Approximate reduction in living expenses from marriage, used as an analogy for selection effects. Initial measured return reduction after controlling for IQ: about 30% - Correcting for pre-existing ability lowers estimated college payoff substantially. Kaplan’s rough estimate of genuine gain: about 55% - His best guess for the share of the observed college premium that is truly causal. Signaling share of educational payoff: about 80% - Kaplan’s stated view of the importance of signaling relative to human capital. Labor economist mainstream view of signaling: 5%–10% - Kaplan’s characterization of the standard academic position. Low-ability students’ graduation chance: around 10% - For students in the bottom quarter of their high school class, according to Kaplan’s cited estimates. Bachelor’s degree earnings effect in poor countries: around 10% per year - Reported international return to education in low-income countries with weak school quality. Minimum wage literature challenge: most workers unaffected - Kaplan notes that many workers are not directly affected, complicating causal estimation.

Pivotal Quotes: "I think that there’s some truth in it, but it’s greatly exaggerated." — Brian Kaplan: Kaplan’s response to the common view that higher technological complexity explains the rising college premium. "You need to distinguish between college graduates being more valuable and college itself being more valuable." — Brian Kaplan: A key distinction in his critique of the standard human-capital explanation. "The heart of the signaling model is saying that actually there’s a big negative externality of education, that every time you get more, you’re making everyone else who didn’t jump through the same number of hoops look worse." — Brian Kaplan: Kaplan’s policy conclusion about education as an arms race.

Implications: Listeners should be cautious about interpreting average college returns as causal. The episode suggests completion and credentialing matter more than attendance, and that expanding subsidies may intensify a costly signaling race rather than raise social productivity.