AI exposure linked to higher U.S. jobs and wages in new study
A study of U.S. data through 2024 found industries more exposed to AI had stronger output, employment and wage growth, though the evidence is still early.
By Priya Nair · Economy Reporter
· 3 min read
The AI jobs debate has a new data point: a study of U.S. employment from 2017 to 2024 found that industries more exposed to AI saw higher output, employment and real wages. For investors watching the AI buildout, that suggests the technology may be doing more than helping companies cut costs, at least in the data reviewed so far.
The finding lands in the middle of a familiar argument about automation. Many workers worry that AI tools will replace white-collar roles. Others argue that cheaper, more productive technology can create more demand for work, because people and companies use more of something once it becomes easier and less expensive to produce.
That idea is known as Jevons' Paradox. The classic version comes from the Industrial Revolution: steam engines made manufacturing more efficient, but factory employment rose as machines spread. A later example often cited in the same debate is the automated teller machine. ATMs changed the work bank tellers did, but did not erase the role; tellers shifted toward customer advice and service.
Microsoft CEO Satya Nadella has applied the same logic to AI, arguing in comments reported by Barron's that AI could increase employment in software engineering and other fields touched by the technology. The mechanism is straightforward: if coding gets easier, more people and firms may do more coding, which can raise the total amount of work even if each task takes less time.
What the study found
A paper by Johnston and Makridis looked at actual U.S. employment data from 2017 through 2024 and combined it with output and productivity figures from gross domestic product releases, plus wage data from the American Community Survey. Gross domestic product, or GDP, is the broad measure of goods and services produced in an economy.
The study treated 2021 as the starting point for wider industrial enterprise AI use, followed by the spread of generative AI after ChatGPT launched in 2023. Generative AI refers to systems that can produce text, code, images or other content from prompts.
Johnston and Makridis found that a one standard deviation increase in AI exposure at a business was associated with a 7% increase in output between 2020 and 2024. A standard deviation is a statistical way to describe how far something is from the average. In this case, the difference is roughly comparable to the gap in AI exposure between a U.S. Postal Service worker and an online retail worker, according to the study description.
To measure exposure, the researchers used an index created by OpenAI and the Centre for the Governance of AI at Oxford University. The index assigns occupations a score based on how exposed their work is to AI tools.
The study also found that industries with higher AI exposure recorded stronger labor outcomes. A one standard deviation increase in AI exposure was linked to a 3.9% rise in employment and a 4.8% increase in affected employers' wage bills, according to Johnston and Makridis.
Those labor gains were smaller than the output increase. That means the study did not find that every bit of extra productivity flowed to workers through more jobs or higher pay. It did find statistically significant job growth in more AI-exposed industries.
The results are early, and the period studied ends in 2024. AI adoption is still spreading, and new data could change the picture. For now, Johnston and Makridis add evidence that AI may be following the Jevons' Paradox pattern: higher productivity paired with more work, rather than a clean swap of software for jobs.
This story draws on original reporting from Klement on Investing.
