The paper summarized here is part of the fall 2025 edition of the Brookings Papers on Economic Activity, the leading conference series and journal in economics for timely, cutting-edge research about real-world policy issues. Research findings are presented in a clear and accessible style to maximize their impact on economic understanding and policymaking. The editors are Brookings Nonresident Senior Fellows Janice Eberly and Jón Steinsson.
See the fall 2025 BPEA event page to watch paper presentations and read summaries of all the papers from this edition. Submit a proposal to present at a future BPEA conference here.
A model based on the economy’s past adaptations to technological change suggests that advances in artificial intelligence (AI) will shift labor demand away from occupations once helped by technological progress, according to a paper to be discussed at the Brookings Papers on Economic Activity (BPEA) conference on September 26.
Most past technological advances, by substituting for physical labor, tended to benefit occupations with higher educational requirements, higher wages, and a greater fraction of female workers, the paper notes.
“In sharp contrast to the past two centuries, our framework suggests that AI—by substituting primarily for cognitive tasks—will shift relative demand toward occupations with lower education, lower wages, and a greater share of male workers,” write the authors—Huben Liu, Dimitris Papanikolaou, and Bryan Seegmiller of the Northwestern Kellogg School of Management, and Lawrence D.W. Schmidt of the Massachusetts Institute of Technology Sloan School of Management.
That’s because the overall increase in economic growth and productivity generated by AI will indirectly benefit workers whose jobs are not negatively affected by AI, such as jobs focused on manual or interpersonal tasks, according to the paper.
The authors used patent documents from 1850 to 2010 to identify past technological improvements. Then they used a state-of-the-art large language model with web-search capabilities (itself a form of AI) to track the extent to which innovation through the years affected task descriptions for U.S. Census Bureau occupations. They applied the insights from the past to explore how AI could affect employment trends over the next 10 to 20 years.
The paper notes that how AI affects labor demand in specific occupations will depend on the extent to which AI affects specific tasks. Bookkeepers and proofreaders, for instance, may find that AI substitutes for most of their tasks, reducing demand for their skills. Other occupations may find substitutes for only a subset of their tasks, allowing them to spend more time on other tasks and thus improve their overall effectiveness. Doctors and nurse practitioners, for instance, may be able to spend less time documenting patient visits and more time counseling patients.
Also, younger workers in occupations enhanced by AI may be better able to acquire the new skills needed to take advantage of new technologies while older workers could be left behind, the paper notes.
That suggests, Schmidt said in an interview with The Brookings Institution, the need for programs to help workers retrain, similar to federal Trade Adjustment Assistance aimed at helping workers hurt by import competition.
“The technology is here. It’s not going away. Technology always produces winners and losers. This is something that has been going on for a long time,” Schmidt said.
Authors
CITATION
Liu, Huben, Dimitris Papanikolaou, Lawrence D. W. Schmidt, and Bryan Seegmiller. 2025. “Technology and Labor Markets: Past, Present, and Future; Evidence from Two Centuries of Innovation.” BPEA Conference Draft, Fall.
-
Acknowledgements and disclosures
David Skidmore authored the summary language for this paper. Chris Miller assisted with data visualization.
The Brookings Institution is committed to quality, independence, and impact.
We are supported by a diverse array of funders. In line with our values and policies, each Brookings publication represents the sole views of its author(s).
