Economy
LLM analysis may provide an early signal of shifts in global monetary policy
A Federal Reserve staff analysis compares an LLM-based measure of central-bank communication with a traditional semi-structural estimate of monetary-policy stance. Using speeches from 119 countries between 1996 and 2025, it finds that both measures move around major crises, but the communication measure often turns earlier and varies more sharply. The authors interpret this as possible forward-looking information in speeches, not proof of predictive power. The global measure is weighted toward countries that speak more frequently, especially major economies.
Editorial summary · Publisher article
What matters
- The study scores speeches with both a fixed lexicon and Claude Haiku 4.5, producing materially different magnitudes after the global financial crisis.
- LLM-based communication measures often shifted before semi-structural estimates around the financial crisis, COVID-19 and the post-pandemic tightening cycle.
- The communication-weighted global indicator overrepresents speech-active and systemically important central banks.
- The authors present the LLM measure as a complement to macroeconomic models, not a replacement.
Different signals of policy stance
The LLM measure captures policymakers’ stated interpretation of incoming information, risks and intended direction. The semi-structural model infers stance from inflation, output, unemployment, interest rates and long-term yields, so divergence between them can reflect forward guidance or changing expectations.
Important weighting and interpretation limits
Global scores are weighted by speech frequency rather than equally across countries. The authors also stress that the apparent lead of communication varies by country and episode, and that the analysis does not establish a universal predictive relationship.
Context & caveats
- This is a Federal Reserve staff analysis presenting the authors’ views; its historical associations do not establish that LLM scores will reliably forecast future policy.
Why this made the edition
The Federal Reserve staff note provides a detailed, transparent comparison of LLM-based central-bank communication measures with semi-structural monetary-policy estimates, including methodology, limitations and historical results.
Source material