NEW YORK—On Tuesday, July 7, 2026 at 10:00 AM, the Federal Reserve Bank of New York will release a two-part Liberty Street Economics blog series based on a recent paper about bank runs. The series highlights the authors’ new database on the incidence of bank runs in U.S. history and how it can be used to improve the understanding of the nature of financial crises.
This series includes:
“Using AI to Let History Speak About Bank Runs” which describes how the authors used large language models to create the most comprehensive database of bank runs in U.S. history, drawing from historical newspapers. It describes the database’s key features, including its ability to differentiate between different types of distress episodes, plot the rate of bank runs over time, and understand the geographic dynamics of those runs.
“What Do Over 3,000 Bank Runs Teach Us About Banking Crises?” which draws on the authors’ database to understand when bank runs typically occur, how common it is for those runs to result in bank failure, and when runs will trigger full-blown banking crises. Understanding these dynamics has first order implications for how to think about financial crises and the appropriate policy responses.
