Job Polarization

In the AI Story

The pattern was first documented in US data by Autor, Lawrence Katz, and Melissa Kearney in a landmark 2006 paper, then confirmed across Western European labor markets by Maarten Goos and Alan Manning. The phenomenon cannot be explained by standard skill-biased technological change, which predicts a smooth gradient of wage growth by education. Polarization is a task-based phenomenon: it is not that middle-educated workers fell behind, but that the specific tasks they performed — routine cognitive and routine manual — were disproportionately the ones computers could do.

The AI era threatens an extension of polarization into what Autor's recent work calls the 'hollowing of the top.' Previous automation waves commoditized middle-skill work; AI commoditizes portions of high-skill cognitive work — legal research, medical diagnosis, content production, junior-level coding. The non-routine cognitive tasks that provided the wage premium at the top are being partially absorbed by large language models, compressing the premium and extending the polarization pattern upward. This is the competitive compression Milanovic describes from a distributional angle.

The political implications of polarization have been enormous. The disappearance of middle-skill jobs contributed to the political realignment visible across Western democracies after 2010, as workers whose livelihoods depended on routine middle-skill occupations found themselves without an economic ladder. The AI extension of polarization into the professional class threatens to produce an analogous political disruption among populations that previously enjoyed economic security. The displaced expert Ehrenreich documents is polarization's newest victim.

Origin

The phenomenon was named and documented in 'The Polarization of the U.S. Labor Market' (Autor, Katz, Kearney 2006), though earlier European work by Goos and Manning had identified similar patterns. The paper remains one of the most cited in labor economics and established polarization as the dominant empirical framework for understanding technology's labor-market effects.

Key Ideas

The U-shaped distribution. Employment growth is concentrated at the top and bottom of the wage distribution, with the middle hollowing out — a pattern inconsistent with simple skill-biased technological change.

Task content predicts exposure. The routine task content of middle-skill occupations, not the education level of their workers, is what made them vulnerable to automation.

Wage polarization follows employment polarization. As middle-skill employment declined, displaced workers crowded into low-skill service jobs, compressing wages at the bottom.

AI extends the pattern upward. The same mechanism that hollowed the middle is now hollowing portions of the top, threatening populations whose economic security depended on non-routine cognitive tasks.

Debates & Critiques

Some economists, notably David Deming, have argued that polarization has already reversed since 2015, with middle-skill employment recovering in certain sectors. Others contend that AI represents such a qualitative break that the polarization framework itself requires revision. Autor's position remains that the framework's predictions continue to hold but the task frontier has moved.