Scope
6,693 policy statements from 51 central banks, with standardized metadata and publication timestamps.
Data Construction
How MPSD is built
An advanced, reproducible pipeline combining NLP preprocessing and Large Language Model tools for standardized central bank statement collection and analysis.
Database Features
Timeframe
Continuous coverage from 1990 to 2024, capturing structural shifts in central bank communication over three decades.
Methodology
Advanced pipeline combining standard NLP preprocessing with Large Language Model (LLM) tools for aspect-based sentiment analysis and question answering.
Reproducibility
Includes the full codebase for cross-country analysis and indicator generation. Versioned releases support verification and extension of published findings.
Future Updates
Update Frequency
Data is refreshed on a quarterly or semi-annual basis to remain current with central bank publication cycles.
Geographic Expansion
The list of jurisdictions is continuously growing, with 57 countries currently in the expansion pipeline.
Next Major Update
Work is now underway to scrape and standardize central bank minutes. This will be the next major expansion of MPSD, extending the project beyond policy statements. Additional central banks will also be added, broadening the geographic coverage and supporting new research on decision-making, committee communication, and policy transmission.