央行文本信息及其对金融市场的预测效果研究

Research on central bank text information and its predictive power for financial markets

ES评分 0

DOI 10.12208/j.aif.20250035
刊名
Advances in International Finance
年,卷(期) 2025, 7(3)
作者
作者单位

济南大学 山东济南

摘要
随着全球金融市场的复杂程度与相互关联性显著增强,传统预测方法存在宏观数据滞后、市场情绪难量化等局限。本文以央行文本信息为研究对象,选取《中国货币政策执行报告》为代表性文本,构建面向央行文本的领域词典,系统检验其对金融市场的预测效果。在方法构建上,本文克服了通用词典在金融文本处理中的适应性不足,通过集成Word2Vec词向量技术与人工筛选机制,构建了专门针对央行政策语境的领域词典,有效解决了通用词典在金融政策文本中的语义偏差问题,并引入否定词与程度副词调整规则,构建了适用于央行语料的情绪指数测度框架。实证结果显示:央行文本信息对股票市场收益率表现出显著预测能力,尤其对创业板指的预测效果最为突出,表明成长型资产对政策情绪信号具有更高敏感性;在外汇市场方面,文本情绪信息对在岸人民币汇率的贬值压力抑制效应显著,而离岸市场因境外投资者的语言壁垒呈现预测失效。本文既为央行文本信息量化提供了方法论创新,还为政策制定者优化预期管理、投资者识别市场信号提供了实证依据。
Abstract
As global financial markets grow increasingly complex and interconnected, traditional forecasting methods face limitations such as lagging macroeconomic data and difficulties in quantifying market sentiment. This study focuses on textual information from central banks, selecting the China Monetary Policy Execution Report as a representative corpus, and constructs a domain-specific dictionary for central bank texts to systematically examine its predictive power for financial markets. Methodologically, this research addresses the inadequacy of general-purpose dictionaries in processing financial texts by integrating Word2Vec word embedding technology with manual screening mechanism. Based on semantic relevance, domain-specific vocabulary is expanded, and rules for negation words and degree adverbs are introduced to develop a sentiment indexing framework tailored to central bank language. Empirical results show that central bank textual information significantly predicts stock market returns, with the most pronounced effect observed for the ChiNext Index, indicating that growth-oriented assets are more sensitive to policy sentiment signals. In foreign exchange markets, textual sentiment significantly suppresses depreciation pressures in onshore CNY rates, whereas predictive power fails in offshore markets due to language barriers among international investors. This study not only offers methodological innovations in quantifying central bank textual information but also provides empirical support for policymakers to improve expectation management and for investors to identify market signals.
关键词
央行文本信息;领域词典;金融市场预测;Word2Vec
KeyWord
Central bank textual information; Domain-specific dictionary; Financial market forecasting; Word2Vec
基金项目
页码 28-36
  • 参考文献
  • 相关文献
  • 引用本文

段从强. 央行文本信息及其对金融市场的预测效果研究 [J]. 国际金融进展. 2025; 7; (3). 28 - 36.

  • 文献评论

相关学者

相关机构