发布者:党政办公室(科研) 时间:2025-10-31
This study contributes to understanding the tone in press conferences held by the President of the European Central Bank (ECB) after a Governing Council meeting. We use a Large Language Model (LLM) for sentiment analysis, specifically focusing on the financial context using the finBERT model. We derive two types of sentiment indices, for the whole press conference, but also individually for its introductory statement as well as for its Q&A part. We find that the tone in the introductory part of the conferences is related to macro-events such as crises, while the Q&A portion is connected to both shocks and presidential periods. We also identify a strong link between sentiment and several macroeconomic variables. Variables related to inflation and industrial production have a significant, but differing impact on our polarity and subjectivity indexes. Our results contribute to the understanding of the tone in the ECB’s communication and highlight the potential of large language models to unveil sentiment contained in central banks’ narrative.