Agree to Disagree: Measuring Hidden Dissents in FOMC Meetings

Résumé 0

Based on a record of dissents on FOMC votes and transcripts of the meetings from 1976 to 2017, we develop a deep learning model based on self-attention mechanism to create a measure of disagreement for members in each meeting. While dissents are rare, we find that members often have reservations with the policy decision, and the level of disagreement is mostly driven by current or predicted macroeconomic data. Using our model to evaluate speeches made by members between meetings, we find that the informational content of speeches is low if we can only compare them to speeches made by the chair. Disagreement strongly correlates with data from the Summary of Economic Projections and a measure of monetary policy sub-optimality, suggesting that disagreement is driven by both members' different preferences and their different views about the future.