Abstract:In this study, we analyze documents published by central banks using text mining techniques and propose a method to evaluate the policy tone of central banks. Since the monetary policies of major central banks have a broad impact on financial market trends, the pricing of risky assets, and the real economy, market participants are attempting to more accurately capture changes in the outlook for central banks' future monetary policies. Since the published documents are also an important tool for the central bank to communicate with the market, they are meticulously elaborated on grammatical syntax and wording, and investors are urged to read more accurately about the central bank's policy stance. Sentiment analysis on central bank documents has long been carried out, but it has been difficult to interpret the meaning of the documents accurately and to explicitly capture even the intentional change in nuance. This study attempts to evaluate the implication of the zero-shot text classification method for an unknown economic environment using the same model. We compare the tone of the statements, minutes, press conference transcripts of FOMC meetings, and the Fed officials' (chair, vice chair, and Governors) speeches. In addition, the minutes of the FOMC meetings were subjected to a phase analysis of changes in each policy stance since 1971.
What problem does this paper attempt to address?
The problem this paper attempts to address is the analysis of documents released by central banks using a textual entailment model to assess the policy tone of central banks. Specifically, the researchers aim to more accurately interpret the communication intentions of the Federal Reserve (Fed) through zero-shot classification methods, especially in unknown economic environments.
### Main Issues:
1. **Accurately Interpreting the Meaning of Central Bank Documents**: Traditional sentiment analysis methods struggle to accurately explain subtle changes in documents, especially intentional semantic changes. The method proposed in this paper aims to capture these changes more precisely.
2. **Evaluating the Effectiveness of Different Data Sources**: Researchers compared the effectiveness of different data sources, such as FOMC meeting statements, meeting minutes, press conference records, and speeches by Fed officials, in conveying policy intentions.
3. **Long-term Trend Analysis**: By analyzing FOMC meeting minutes from 1971 to the present, researchers hope to reveal trends in the Fed's policy tone and its consistency with actual policy changes.
### Research Background:
- **Importance**: The monetary policies of major central banks have a broad impact on financial market trends, risk asset pricing, and the real economy. Therefore, market participants need a better understanding of future changes in central bank monetary policy.
- **Data Sources**: The study used various documents, including FOMC meeting statements, meeting minutes, press conference records, and speeches by Fed officials.
- **Technical Methods**: A zero-shot classification textual entailment model was used to assess the policy tone in the documents.
### Research Objectives:
- **Improve Interpretation Accuracy**: Using natural language processing techniques to accurately read the intentions of document authors and speakers, becoming the most accurate communication receiver.
- **Long-term Trend Analysis**: Analyze FOMC meeting minutes from 1971 to the present, classifying the characteristics of the Fed's policy tone during periods of rising and falling interest rates.
### Research Methods:
1. **Data Acquisition and Preprocessing**: Various documents were obtained from the Fed's website and preprocessed, such as removing footers and annotations, and splitting documents into sentence units.
2. **Topic Classification**: Using a zero-shot classification textual entailment model to classify sentences into three topics: "inflation," "employment growth," and "economic growth."
3. **Hypothesis Statement Setting**: Based on keywords extracted from past FOMC statements, hypothesis statements were set for each category, balancing expressions in both upward and downward directions.
4. **Entailment Judgment**: Entailment judgment was performed on sentences in each category, scores were calculated, and the number of sentences in each document that matched the hypothesis statements was counted.
5. **Stance Scoring**: Stance scores for each category were calculated by comparing the differences in the number of sentences in the upward and downward directions to assess the policy tone.
### Results:
- **Short-term Analysis (December 2018 to March 2022)**: The results show differences in the policy tone conveyed by different data sources regarding inflation, employment growth, and economic growth. Press conference records showed early awareness of increasing inflationary pressures.
- **Long-term Analysis (February 1971 to March 2022)**: Long-term analysis of FOMC meeting minutes revealed that the Fed's communication was consistent with actual policy changes, especially during periods of rising and falling interest rates.
### Conclusion:
- **Effectiveness of the Model**: The zero-shot classification textual entailment model can effectively assess the policy tone in central bank documents and shows consistency across different data sources.
- **Future Work**: Further optimization of data preprocessing methods to improve the accuracy and practicality of the model.
Through these methods, researchers hope to provide market participants with more accurate tools for interpreting central bank policies.