美联储-通过金融媒体的视角衡量联邦公开市场委员会通信的情绪（英）-2025_51页_709kb

Summary of "Gauging the Sentiment of Federal Open Market Committee"

Authors: Shantanu Banerjee, Paul Cordova, Michiel De Pooter, Olesya V. Grishchenko
Date: July 3, 2025

This research examines how to gauge the market sentiment surrounding Federal Open Market Committee (FOMC) communications using natural language processing (NLP). The study constructs a sentiment index based on financial news articles, specifically from six major outlets, collected around FOMC meetings from May 1999 to November 2022.

Key Methodology

The authors developed topic-specific dictionaries to capture different aspects of monetary policy: conventional policy tools (e.g., policy rates), asset purchases, and forward guidance. They applied NLP techniques, including removing stop words, lemmatization, and weighting relevance by the length of articles and frequency of modifier words (e.g., "raise" or "cut"). Negations were handled by flipping word orientations. Sentiment surprises are calculated as the difference in average sentiment levels before and after FOMC announcements.

Main Findings

The sentiment index surprises provide incremental explanatory power for asset price movements in money market futures, Treasury securities, equity indexes, and currency pairs, even after accounting for standard monetary policy surprises. This holds across different economic periods, such as the post-GFC, zero-lower-bound (ZLB), and COVID-19 phases. The methodology shows higher explanatory power compared to alternative sentiment indexes, particularly due to the separate treatment of policy regimes.

Conclusion

The paper highlights that nuanced sentiment analysis from financial press articles enhances understanding of market reactions to FOMC communications, offering a robust tool for policymakers and investors.