ECONDAT 2026 convenes global researchers on AI and macroeconomics
The Bank of Japan hosts the eighth ECONDAT conference in Tokyo on October 5 and 6, 2026, focusing on nontraditional data, machine learning, and natural language processing in macroeconomics. Central bank researchers and academics gather to examine AI adoption and economic forecasting.
From sentiment parsing to granular tracking
The two-day conference opened with remarks from Bank of Japan Deputy Governor Shinichi Uchida and keynote addresses by Harvard University professor Melissa Dell and the University of Michigan's Matthew Shapiro.
Across four main sessions, researchers from institutions including the Federal Reserve System, the Bank for International Settlements, and the Bank of England presented empirical work on machine learning applications.
Indrajit Mitra presented deep learning methods for firm-level input price changes, while Taejin Park analyzed macroeconomic sentiment decomposition using large language models.
Additional presentations examined high-frequency neighborhood activity and recession prediction.
Algorithmic herding and policy trade-offs
The second day focuses on the broader economic implications of artificial intelligence adoption, featuring papers on AI trading agent herding, labor market shifts from online job postings, and central bank trade-offs presented by Maryam Haghighi of the Bank of Canada.
A policy panel brings together John J. Horton from MIT, Ren Ito of Sakana AI, and Vatsala Shreeti from the BIS, before concluding remarks by Bank of Japan Executive Director Kenji Suwazono.
Beyond the experimental sandbox
Central banks are finally shifting from experimental text analysis to systemic risks from automated AI trading.
Extracting reliable policy signals from real-time granular data remains notoriously difficult.
These machine learning tools must prove their worth during real market turbulence rather than benign conditions.