Forecasting Federal Reserve Decisions with Text and Economic Data
Summary
The study tests whether Federal Reserve communications add useful information to forecasts of federal funds rate decisions. It compares models using economic indicators alone, text features alone, and combinations of the two, including conventional machine learning, transformer based language features, and deep learning approaches.
The strongest reported result comes from combining TF-IDF features from FOMC texts with economic indicators in an XGBoost classifier, which achieved a test AUC of 0.83. FinBERT sentiment features slightly improved ranking but classified outcomes less effectively, particularly with imbalanced classes. SHAP analysis suggests the sparse text features align more closely with policy relevant signals. These findings favor transparent hybrid models, but the excerpt gives limited detail on sample construction, validation design, or performance across different policy periods, so the result should not be treated as proof of reliable live forecasts.
Key ideas
- Combining FOMC text features with economic indicators outperformed models using either source alone.
- The best reported model used TF-IDF text features and XGBoost, with a test AUC of 0.83.
- FinBERT sentiment modestly helped ranking but performed worse for classification under class imbalance.
- SHAP analysis indicated that sparse text features were more aligned with policy relevant signals.
- The excerpt does not establish that the reported performance will persist across periods or in live trading.
Tags
Full text
# Can We Reliably Predict the Fed's Next Move? A Multi-Modal Approach to U.S. Monetary Policy Forecasting # Can We Reliably Predict the Fed's Next Move? A Multi-Modal Approach to U.S. Monetary Policy Forecasting Forecasting central bank policy decisions remains a persistent challenge for investors, financial institutions, and policymakers due to the wide-reaching impact of monetary actions. In particular, anticipating shifts in the U.S. federal funds rate is vital for risk management and trading strategies. Traditional methods relying only on structured macroeconomic indicators often fall short in capturing the forward-looking cues embedded in central bank communications. This study examines whether predictive accuracy can be enhanced by integrating structured data with unstructured textual signals from Federal Reserve communications. We adopt a multi-modal framework, comparing traditional machine learning models, transformer-based language models, and deep learning architectures in both unimodal and hybrid settings. Our results show that hybrid models consistently outperform unimodal baselines. The best performance is achieved by combining TF-IDF features of FOMC texts with economic indicators in an XGBoost classifier, reaching a test AUC of 0.83. FinBERT-based sentiment features marginally improve ranking but perform worse in classification, especially under class imbalance. SHAP analysis reveals that sparse, interpretable features align more closely with policy-relevant signals. These findings underscore the importance of integrating textual and structured signals transparently. For monetary policy forecasting, simpler hybrid models can offer both accuracy and interpretability, delivering actionable insights for researchers and decision-makers.
Shown in full with attribution under the source's licence. Licence: abstract CC0
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.