Human Capital Repricing in the AI Economy

In the AI Story

Becker's framework transformed labor economics by treating workers as capital assets whose value could be increased through investment. The insight was powerful: it explained wage premiums for education, justified public spending on schooling, and provided the conceptual foundation for the postwar expansion of higher education. The model distinguished general human capital (skills transferable across employers) from specific human capital (skills valuable only within a particular firm), predicting that workers would invest in general capital while firms would invest in specific capital. The framework assumed that the skills certified by credentials—the ability to perform complex cognitive tasks that most people could not—would remain scarce for the decades following their acquisition.

AI breaks the scarcity assumption for a specific class of skills: trainable, rule-based, execution-oriented cognitive tasks. Learning Python was a valuable human capital investment in 2015 because competent Python developers were scarce relative to demand, and the skill could not be easily replicated. A decade of coding experience built mental models, debugging intuition, and architectural judgment that commanded market premiums. When AI writes Python competently from natural-language descriptions, the marginal return on learning Python collapses—not to zero, but toward the much lower return on 'being able to evaluate whether this Python code is correct,' a task requiring less investment and less scarcity. The ten-year experience premium compresses because the execution component that consumed most of the ten years is now provided by the tool.

The educational institutions built on the Beckerian model face a brutal adjustment. A four-year computer science degree that costs $200,000 was economically rational when it certified skills commanding $150,000 starting salaries with predictable career trajectories. When the market reprices those skills—reducing the premium on implementation while raising the premium on judgment—the degree's value proposition strains. The network access, credential signal, and socialization retain value, but the core justification—that the execution skills taught will produce a lifetime earnings premium sufficient to repay tuition—becomes harder to sustain. Enrollment in CS programs is already softening at some institutions as prospective students perform their own marginal calculations and increasingly conclude the ROI is uncertain.

Cowen's prescription for education is characteristically unsentimental: restructure around the scarce input. Stop optimizing for execution-skill certification that the market is repricing. Start optimizing for judgment development, integrative capacity, the ability to ask valuable questions. This requires pedagogical methods institutions struggle to scale: case-based learning, apprenticeships, cross-disciplinary exposure, high-friction engagement with ambiguous problems. The judgment that markets will pay for is built through experience, failure, and consequence—conditions that standardized education systematically removes in favor of controlled, low-stakes practice. The institution that solves the judgment-development problem at scale will capture the human capital market of the AI age. The institutions that cling to execution-focused curricula will produce graduates holding credentials the market has repriced.

Origin

Gary Becker developed human capital theory across the 1960s, formalizing it in his 1964 book Human Capital, which argued that individuals and societies should invest in education and training as they invest in physical capital. The framework won Becker the 1992 Nobel Prize and became the dominant lens for understanding education's economic value. Cowen's repricing thesis builds on Becker while recognizing that Becker's model assumed a static relationship between skills and scarcity. When a general-purpose technology disrupts that relationship—making previously scarce skills abundant—the human capital framework requires fundamental revision. Cowen's innovation is recognizing that the revision is not merely quantitative (lower returns) but qualitative (different skills command premiums), and that the institutions built on the old model will resist the revision long past the point where the market has already repriced.

Key Ideas

Execution skills face commodity pricing. Any trainable, rule-based cognitive skill AI can perform loses its scarcity premium—not worthless, but worth far less than the investment it required.

Judgment skills resist commoditization. The capacity to evaluate, integrate, decide under uncertainty, and originate questions remains scarce and commands rising premiums because it resists both training and automation.

Educational institutions face a value-proposition crisis. Degrees certifying execution skills are depreciating assets; institutions must restructure around judgment development or watch enrollment and outcomes diverge.

The repricing is faster than institutional adaptation. Markets reprice in months; universities restructure curricula in years—the lag is where students and workers bear uncompensated transition costs.