Can We Reliably Predict the Fed's Next Move? A Multi-Modal Approach to U.S. Monetary Policy Forecasting

2025-06-28

AI Summary

Traditional macroeconomic indicators struggle to capture policy shift signals embedded in the Federal Reserve’s forward guidance, limiting interest rate prediction accuracy. This paper proposes a multimodal forecasting framework that integrates central bank textual data (FOMC statements) with structured economic indicators, yielding an interpretable hybrid model. Methodologically, we systematically compare unimodal and multimodal architectures—including TF-IDF, FinBERT, XGBoost, and Transformer-based models—and employ SHAP for feature attribution analysis. Results show that the TF-IDF + economic indicators + XGBoost model achieves an AUC of 0.83 on the test set, significantly outperforming unimodal baselines; SHAP analysis confirms that its feature importance aligns with actual monetary policy logic, validating the incremental explanatory power of textual signals for interest rate movements. To our knowledge, this is the first study to demonstrate that lightweight, interpretable multimodal models achieve high predictive performance, transparency, and practical feasibility in monetary policy forecasting.

Technology Category

• Machine Learning: Multimodal Learning
• Natural Language Processing: Language Grounding & Multi-modal NLP
• Computer Vision: Multi-modal Vision

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

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.

Problem

• Predicting U.S. federal funds rate shifts accurately
• Enhancing forecasts with multi-modal data integration
• Balancing model accuracy and interpretability for policy insights

Innovation

• Multi-modal framework combining structured and unstructured data
• Hybrid models outperform unimodal baselines in forecasting
• TF-IDF features with economic indicators enhance accuracy