Further results on AI and labor market reallocation

A new NBER paper reveals how artificial intelligence is reshaping the U.S. labor market. By merging data on worker demographics and labor force participation from the Current Population Survey and the Job Openings and Labor Turnover Survey with AI-related metrics from OpenAI and Lightcast, researchers constructed measures of labor force stocks, worker movements, and job market tightness based on AI exposure and adoption levels.

Since the launch of large language models, workers heavily reliant on AI have faced more significant drops in their ability to secure jobs and switch careers compared to other workers. However, they have shown a notable uptick in the frequency of within-job activity changes. The benefits of AI on nominal wage growth and overall work hours have weakened since the emergence of LLMs, indicating that AI-driven labor market reallocation involves within-firm task reshuffling and the restructuring of labor inputs, along with reduced demand for workers most exposed to AI.

To measure the impact of LLMs on labor market reallocation, the authors developed individual-level indices for "hiring potential" and "job switching feasibility." They found that job-finding prospects have become much more uneven since 2023, primarily driven by AI factors. The natural rate of unemployment—estimated by accounting for AI-related inflows and outflows in unemployment rates and incorporating a theoretical model of structural unemployment—has risen by around 0.1-0.2 percentage points since the introduction of LLMs. However, this estimation comes with significant uncertainty.