Economist Imas Identifies Missing Productivity Data for AI Job Impact
Imas calls for productivity and demand data beyond exposure metrics to predict AI labor outcomes, citing limits of O*NET, OpenAI and Anthropic studies.
University of Chicago economist Alex Imas has called for collection of data tracking AI effects on task-level productivity and resulting changes in sectoral labor demand.
MIT Technology Review reported that Silicon Valley forecasts of mass displacement cite OpenAI's December 2023 analysis of O*NET tasks launched in 1998 showing real estate agents at 28% exposure and Anthropic's February 2024 study of millions of Claude conversations mapping task overlap (MIT Technology Review, April 6 2026; OpenAI arXiv:2303.10130, 2023). Imas stated exposure metrics alone are meaningless for predicting displacement except in narrow cases of full task automation such as elevator operators or basic call triage where cost and capability thresholds are met.
Imas cited coding work where AI reduces three days of effort to one increasing output per worker and noted outcomes hinge on whether firms in competitive markets like dating apps respond with lower prices that boost demand or pocket gains with uncertain hiring effects (MIT Technology Review, April 6 2026). The article connects this to broader patterns seen in Acemoglu and Restrepo 2018 NBER work on automation that similarly required task-level data to separate displacement from productivity effects.
Original coverage in outlets such as The New York Times 2023-2025 focused primarily on exposure percentages from the same O*NET catalogue without addressing demand elasticity or measured productivity shifts per task which Imas identified as the critical gap also missed in Anthropic CEO Dario Amodei's five-year full substitution claim (MIT Technology Review, April 6 2026; Acemoglu et al. NBER, 2018).
AXIOM: Task productivity and demand elasticity data will determine if AI complements labor like computers in the 1980s or displaces it sector by sector; collection must begin immediately to inform policy.
Sources (3)
- [1]The one piece of data that could actually shed light on your job and AI(https://www.technologyreview.com/2026/04/06/1135187/the-one-piece-of-data-that-could-actually-shed-light-on-your-job-and-ai/)
- [2]GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models(https://arxiv.org/abs/2303.10130)
- [3]Automation and New Tasks: How Technology Displaces and Reinstates Labor(https://www.nber.org/papers/w25684)