🤖 AI Summary
This study investigates “latent dissent”—policy disagreements among Federal Open Market Committee (FOMC) members that remain unobserved in formal voting records. Leveraging FOMC voting data and meeting minutes from 1976 to 2018, we develop a deep text model incorporating self-attention mechanisms to systematically quantify the intensity of such unexpressed disagreements for the first time. We find that latent dissent is pervasive and primarily driven by heterogeneous member expectations regarding macroeconomic indicators—especially inflation and unemployment. Its intensity exhibits significant positive correlations with both Summary of Economic Projections (SEP) forecast dispersion and monetary policy suboptimality. Moreover, financial markets respond negatively within minutes of minute releases, confirming market sensitivity to this previously unmeasured dimension of internal FOMC dynamics. By moving beyond traditional vote-based analysis, this work advances understanding of Fed decision-making processes, policy transmission, and market expectation formation, offering both novel empirical evidence and a scalable methodological framework for analyzing institutional deliberations.
📝 Abstract
Using FOMC votes and meeting transcripts from 1976-2018, we develop a deep learning model based on self-attention mechanism to quantify hidden dissent among members. Although explicit dissent is rare, we find that members often have reservations with the policy decision, and hidden dissent is mostly driven by current or predicted macroeconomic data. Additionally, hidden dissent strongly correlates with data from the Summary of Economic Projections and a measure of monetary policy sub-optimality, suggesting it reflects both divergent preferences and differing economic outlooks among members. Finally, financial markets show an immediate response to the hidden dissent disclosed through meeting minutes.
