Letting AI agents write econometric code

CEPR has published material titled “Letting AI agents write econometric code,” examining the use of artificial intelligence agents in a core part of empirical economics research. Econometric code is used to clean data, estimate statistical models and reproduce analytical results, making its quality important for research reliability.

The topic reflects the growing use of generative AI tools in technical workflows. AI agents can be designed to carry out multi-step tasks, including writing and revising software code, rather than only responding to individual prompts. In econometrics, such systems could affect how researchers develop, document and check empirical analysis.

The discussion is relevant to economists, policymakers and institutions that rely on quantitative evidence. Potential applications must be considered alongside established research practices such as transparent methods, reproducible code, data governance and human review. CEPR, the Centre for Economic Policy Research, is an international network focused on economic research and policy discussion.