AI exposure is associated with different productivity patterns across economists, suggesting the technology may redistribute research advantages by support and career stage, a University of Michigan job-market paper reports.
The paper, “Equalizing or Polarizing? Generative AI and Research Productivity in Economics,” was written by University of Michigan Ph.D. candidate Tianyu Jiang and co-author Yuqing (Claire) Wang. Their research draws on National Bureau of Economic Research (NBER) data covering authors’ working papers and journal publications, text-based measures of AI-assisted writing, and survey responses from 227 academic economists. The University of Michigan lists it as Jiang’s job-market paper for 2026-27.
The authors report that after ChatGPT’s release, economists without research assistants had a higher proportion of papers in leading journals in the NBER data. They also found signs of increased working-paper output and shorter publication times among researchers on smaller teams. The paper’s abstract says evidence of an average effect on working-paper output is weak, while noting substantial variation across researchers.
In separate estimates, greater AI exposure was associated with a smaller output gap between economists with and without research assistants, but a larger gap between senior and junior researchers. The authors report that junior researchers used AI more heavily and were more likely to use agentic tools. They interpret the differences as reflecting the productivity returns from AI use, not simply who adopted it.
The authors’ model treats AI as a way to ease execution work, which may help researchers with less support, while increasing the value of accumulated expertise. It allows for gaps to narrow initially and then widen if AI improves at conceptual work.
OpenAI announced ChatGPT on Nov. 30, 2022. The authors describe the NBER-record shifts as occurring after its release.
