Monetary policy analysis using natural language processing: Evaluating the People's Bank of China's minutes and report summary with the Taylor Rule

Abstract

This study investigates the predictive power of the PBOC's concise communication tools—meeting minutes and monetary policy report summaries—in forecasting monetary policy decisions. Existing literature primarily focuses on comprehensive monetary policy reports, often overlooking the effectiveness of brief communication forms like meeting minutes. Using Natural Language Processing (NLP) techniques and an ordered probit model within the Taylor Rule framework, we quantify economic, and inflation signals from PBOC texts between 2002Q3 and 2023Q4. Our findings reveal that economic signals from meeting minutes significantly influence policy rate changes, while inflation signals remain relatively weaker. Further comparative analysis shows that although monetary policy summaries provide balanced signals due to their comprehensive nature, meeting minutes offer stronger short-term predictive power owing to their concise format and timeliness. These results underscore the importance of balanced economic and inflation communication, enhancing our understanding of how central bank textual signals shape policy predictability and market expectations.