Enhancing Regime Shift Detection Using Unstructured Data: A Study on the Treasury Market

Researchers developed a hybrid framework combining large language models with statistical analysis to detect regime shifts in financial markets by analyzing Federal Reserve communications alongside Treasury market data. The approach achieved 82% accuracy in identifying monetary policy regime changes, outperforming traditional data-only methods and detecting shifts on the same day they occur.

This research addresses a critical gap in financial market analysis: the difficulty of reliably detecting regime shifts when they occur. Traditional detection methods rely exclusively on historical price and macroeconomic data, which are noisy and highly correlated, making it challenging to identify structural breaks in real time. The study demonstrates that central bank communications—FOMC minutes in this case—contain leading indicators of regime changes that often precede their manifestation in market prices.

The framework's innovation lies in its bidirectional validation approach. Large language models analyze policy text to identify potential regime-shift candidates, while these suggestions are statistically validated against multivariate time series using bootstrap likelihood-ratio tests. Conversely, data-driven detectors' findings are cross-checked against textual evidence, creating a complementary system where each method validates the other.

For market participants, this has significant implications. The ability to detect monetary policy regime shifts with same-day latency and high accuracy provides informational advantages for traders, portfolio managers, and risk analysts. Early detection enables more timely rebalancing and hedging decisions. The methodology also improves interpretability—combining quantitative evidence with policy text makes regime identification less of a black box.

The consistent outperformance over pure data-driven baselines across four different detector types suggests the approach is robust and generalizable. Future applications could extend beyond Treasury markets to equity, commodities, and cryptocurrency markets, where policy communications similarly drive regime transitions. The framework essentially weaponizes central bank transparency, converting their public communications into actionable market intelligence.

• →LLM-enhanced regime detection achieves 82% F1 score, outperforming traditional time-series-only methods
• →The framework detects monetary policy regime shifts on the same day they occur, providing real-time market signals
• →Text analysis of FOMC communications reveals policy intent before market prices reflect structural changes
• →Bidirectional validation between LLM text analysis and VAR statistical tests creates a robust, interpretable detection system
• →The detector-agnostic design enables integration with existing regime-detection algorithms across asset classes