AI linked to significant deceleration in coder employment growth
A Federal Reserve paper finds that aggregate employment growth for computer programmers has sharply decelerated since the introduction of ChatGPT in November 2022. This deceleration is attributed to an occupation-specific shock rather than broader industry trends.
AI's measurable impact on coder employment growth
A Federal Reserve paper by Leland D. Crane and Paul E. Soto investigates the impact of large language models (LLMs) on the aggregate labor market, specifically focusing on computer programming-intensive occupations.
By linking O*NET and CPS data, the study identifies a sharp deceleration in aggregate coder employment since the introduction of ChatGPT in November 2022.
This slowdown is not attributed to coders' exposure to slowing industries, but rather to an occupation-specific shock coinciding with ChatGPT's release.
The research robustly estimates that annual coder employment growth is now approximately 3 percent lower than its pre-ChatGPT rate.
While coder employment continues to expand, its growth pace is notably slower than before 2022, indicating a measurable impact of AI on this specific occupational group.
Disentangling industry and occupation shocks
The paper develops a novel control variable to separate occupation-specific impacts from industry-level shocks.
Using a within-industry/between-industry decomposition, a counterfactual employment series illustrates what coder employment would have been without an LLM-induced shock.
The study also finds that 40 percent of coders work in computer systems design and related services (NAICS 5415), emphasizing their role in contract software development.
Additionally, the authors compare generative AI exposure measures from Eloundou et al. (2024) and Handa et al. (2025), noting significant disagreement between these metrics despite both identifying coders as highly exposed occupations.