Selection into occupations by comparative advantage amplifies wage losses in high-exposure occupations

str 8 12/31/2099 · 1 article

structural · economic · AI, labor · US

Analysis

Workers sort into occupations based on comparative advantage, meaning those in highly exposed occupations tend to be specialized precisely in the tasks being automated. This selection mechanism creates a non-monotonic relationship between occupational exposure and average wage effects: moderate exposure benefits workers (who are specialized in non-automated tasks), while high exposure harms them (who are specialized in automated tasks).

Key actors

workersoccupations

Source article

FM_AI

pdf-archive — Work, Skills and Society · — · extracted in run pdf-import-undated-1779224753682-107 · 5/19/2026, 10:48:16 PM

"As workers sort into occupations by comparative advantage, on average, incumbents of highly exposed occupations tend to be specialized in precisely the tasks being automated." [specialized in precisely the tasks being automated]

This identifies the selection mechanism that drives the non-monotonic wage effect: workers' occupational choices reveal their skill specialization, which then determines whether automation helps or harms them.

Reasoning from this article

The article formalizes this through the wage equation and occupational choice model, then validates it empirically by showing that the inverted-U shape of wage effects disappears when job transformation is removed. This is a general structural principle: in any task-based economy where workers sort by comparative advantage, automation of a task will disproportionately harm those who selected into occupations precisely because they were skilled at that task. The mechanism applies beyond LLMs to any automation shock.