From Text to Quantified Insights: A Large-Scale LLM Analysis of Central Bank Communication
IMF Working Papers, June 6, 2025
Source details
- Canonical URL
- From Text to Quantified Insights: A Large-Scale LLM Analysis of Central Bank Communication
Other formats
Bibliographic details
- Authors: Thiago Christiano Silva, Kei Moriya, Romain M Veyrune
- Published: June 6, 2025
- Series: IMF Working Papers
- DOI: https://doi.org/10.5089/9798229013802.001
Overview
- Paper introduces a classification framework to analyze central bank communications across four dimensions: topic, communication stance, sentiment, and audience.
- Uses a fine-tuned large language model trained on central bank documents to classify individual sentences and convert policy language into systematic, quantifiable metrics.
- Applied to a multilingual dataset of 74,882 documents from 169 central banks spanning 1884 to 2025.
- Presents what the authors describe as the most comprehensive empirical analysis of central bank communication to date.
Methodology
- Classification framework dimensions: topic, communication stance, sentiment, audience.
- Model approach: fine-tuned large language model trained on central bank documents to label individual sentences.
- Dataset scale and coverage:
- 74,882 documents
- 169 central banks
- Time span: 1884 to 2025
Key Findings
- Monetary policy communication changes significantly with inflation targeting:
- Backward-looking exchange rate discussions give way to forward-looking statements on inflation, interest rates, and economic conditions.
- Directional signals:
- Development of a directional communication index capturing signals about future policy rate changes and unconventional measures, including forward guidance and balance sheet operations.
- This unified signal helps explain future movements in market rates.
- Audience tailoring quantified:
- First systematic quantification showing central banks tailor messages to audiences.
- Audience-specific risk communication has remained stable for decades, suggesting a structural and deliberate tone.
- Typical patterns by audience:
- Neutral, fact-based language with financial markets
- Confidence-building language with the public
- Emphasis on risks when communicating to governments
Directional Communication Index
- Captures signals about:
- Future policy rate changes
- Unconventional measures, including forward guidance and balance sheet operations
- Empirical role:
- The unified signal helps explain future movements in market rates
Audience-Specific Communication
- Systematic quantification shows consistent audience tailoring:
- Stability: Audience-specific risk communication stable for decades
- Tone by audience:
- Financial markets: neutral, fact-based language
- Public: confidence-building language
- Government: highlighting risks
Crisis Dynamics and Predictive Power
- Communication shifts markedly during crises:
- Confidence-building rises in communication to the financial sector and government
- Risk signaling increases for other audiences during crises
- Forward-looking risk communication predicts future market volatility:
- Demonstrates that central bank language influences both monetary and financial stability channels
Policy Implications and Interpretation
- Communication as active policy tool:
- Findings provide novel evidence that communication is an active policy tool for steering expectations and shaping economic and financial conditions
- Importance of measuring language:
- Quantified, directional, and audience-specific signals can inform assessments of expected policy actions and potential financial stability implications
Content in this bundle
- Working Paper