Scholarlink

<p>This paper investigates whether the sentiment of reports from the Central Bank of Brazil can enhance the accuracy of macroeconomic forecasts. We construct sentiment scores through textual analyses of the Copom Minutes and Inflation Report, using both a fixed lexicon dictionary and a time-varying lexicon dictionary based on machine learning. We test their predictive power for GDP and inflation and find that sentiment scores from the time-varying lexicon enhance the accuracy of prediction models. Further, these scores can explain GDP forecast errors from market expectations (Focus Survey) in real time and one quarter ahead, but not inflation forecast errors, likely due to the inherent predictability of inflation and methodological limitations.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Central Bank Narratives and Macroeconomic Forecasting: Using Textual Analysis from Machine Learning

  • 期刊论文

  • 发表日期：

    2025年6月11日

查看全文
  • Diego Pitta de Jesus,
  • Cássio da Nóbrega Besarria

摘要

This paper investigates whether the sentiment of reports from the Central Bank of Brazil can enhance the accuracy of macroeconomic forecasts. We construct sentiment scores through textual analyses of the Copom Minutes and Inflation Report, using both a fixed lexicon dictionary and a time-varying lexicon dictionary based on machine learning. We test their predictive power for GDP and inflation and find that sentiment scores from the time-varying lexicon enhance the accuracy of prediction models. Further, these scores can explain GDP forecast errors from market expectations (Focus Survey) in real time and one quarter ahead, but not inflation forecast errors, likely due to the inherent predictability of inflation and methodological limitations.