In labor economics, the Mincerian wage equation treats human capital as two simple assets, years of formal schooling and years of work experience. For four decades, that framework priced a consistent bargain. College paid a reliable equity premium because universities held a near-monopoly on transmitting structured, codifiable knowledge, such as legal codes, accounting rules, syntax, corporate reporting templates, and diagnostic heuristics. Experience paid an additional compounding return because tenure accumulated tacit judgment, including client navigation, political compromise, triage under incomplete information, and the muscle memory of what breaks when production systems fail.
A new paper by Eduardo Levy Yeyati, Dario Judzik, and Juan Ignacio Barbiero (arXiv 2610.36008) uses US Current Population Survey data from 2018 through mid-2026 to track what happens to those two assets when synthetic inference floods the workplace. By comparing high-exposure professional occupations against unexposed baselines before and after the release of ChatGPT, the authors isolate a structural repricing of white-collar labor.
The first asset has collapsed in relative value. In occupations most exposed to generative models, the wage premium for each additional year of schooling fell by 0.55 log points. When foundational models make textbook knowledge, analytical syntax, and boilerplate drafting free at the point of consumption, possessing a certified memory of codified rules ceases to grant pricing power. A credential that certifies you memorized an analytical taxonomy no longer commands an economic moat if a commodity API produces identical synthesis for pennies.
The second asset split into a non-linear payoff curve. Standard human capital theory expects technological shocks to either lift or depress returns across all workers uniformly. Instead, the CPS earnings distribution shows experience rotating sharply around an eighteen-year fulcrum.
For early-career practitioners, returns to experience expanded. At three years of tenure, the wage return to an additional year on the job rose by 0.26 log points. When junior analysts combine generative tooling with raw operational exposure, they acquire practical mastery faster. The tooling automates basic syntax and literature sweeps, allowing young workers to spend their working hours testing edge cases, managing workflows, and absorbing organizational tacit knowledge.
Beyond eighteen years of tenure, the dynamic reverses. At twenty-five years of experience, the marginal return to tenure fell by 0.12 log points. Senior practitioners who built their compensation around technology-specific routines discover that their accumulated institutional capital has become obsolete. Knowing the idiosyncrasies of legacy workflow software or manual drafting conventions no longer protects compensation when the entire production pipeline migrates to generative harnesses.
The aggregate labor market does not face a uniform displacement crisis. It faces a repricing of human capital assets. Degrees function as junior equity tranches on commodified taxonomies, while domain judgment and transferable execution retain pricing power.
Top comments (0)