Labor market impacts of AI: A new measure and early evidence

Source: Anthropic Research Date: 2026-03-05 URL: https://www.anthropic.com/research/labor-market-impacts

Summary

Introduces “observed exposure” — a metric combining O*NET task data, Claude usage data from the Economic Index, and Eloundou et al. (2023) exposure assessments, weighted toward automated vs. augmentative uses. Applied to CPS employment data 2016–2024 via difference-in-differences. Finding: no systematic unemployment increase for high-exposure workers since late 2022, but suggestive evidence of slower hiring of younger workers in AI-exposed roles. Computer programmers at 75% exposure; 30% of workers at zero.

Implications

This is the economic measurement thread — building the real-time tracking framework before the disruption is visible, which is methodologically sound but also politically useful. The “no unemployment increase yet” finding gives Anthropic cover from displacement narratives while the younger-worker hiring slowdown is the real signal to watch. If that trend continues and accelerates, it reshapes the entry-level labor market for software in particular. The framework being replicable and updateable means this becomes the recurring measurement instrument — watch for quarterly updates feeding directly into policy arguments.