Agree to Disagree: Measuring Hidden Dissents in FOMC Meetings

Abstract

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 modules to create a measure of disagreement for each member in each meeting. While dissents are rare, we find that members often have reservations with the policy decision. The level of disagreement is mostly driven by current or predicted macroeconomic data at both the individual and meeting levels, while personal characteristics of the members matter only at the individual level. We also use our model to evaluate speeches made by members between meetings, and we find a weak correlation between the level of disagreement revealed in them and that of the following meeting. Finally, we find that the level of disagreement increases whenever monetary policy action is more aggressive.

Keywords: Natural language processing, disagreement, monetary policy, FOMC

JEL Classification: E52, E58, C55