金融研究 | 央行货币政策文本相似度计算与可视化
本文利用金融情感词典和文本分析技术,分析中国人民银行货币政策执行报告的**文本情绪、文本相似度和文本可读性**等多维文本信息,刻画央行货币政策执行报告的文本特征,探究货币政策报告的文本信息与宏观经济和股票市场的关系。**实证研究发现,货币政策报告的文本情绪的改善会引起显著为正的股票市场价格反应, 报告文本相似度的增加会引起股票市场波动性的显著降低, 报告可读性对公布后股票市场的波动性影响不显著**。货币政策报告文本情绪还与诸多宏观经济指标显著相关。进一步研究发现,引起股票市场显著反应的是报告文本情绪中反映货币政策指引的部分,而反映宏观经济历史状态的部分对股票市场的影响不显著。本文从文本大数据分析角度证明了我国央行沟通的有效性,对国内央行沟通相关研究形成了有益补充。This paper uses text analysis techniques to analyze 71 Monetary Policy Implementation Ｒeports （ hereinafter referred to as“the reports”） of PBOC，calculates the text sentiment （ tone） ，the similarity and readability and other text indicators of the reports，and explores the relationship between these text indicators and the macro economy and the stock market． Based on the Chinese financial sentiment dictionary developed by Jiang et al． （ 2020） ，this paper uses the sentiment unit method to calculate the tone of the reports． In addition，this paper uses TF － IDF weighted cosine similarity to characterize the similarity of the reports，and uses average sentence length to characterize the readability of the reports． The paper then uses correlation analysis to examine the relationship between the tone of the reports and macroeconomic indicators such as economic growth，inflation， and interest rates． With reference to Ehrmann and Fratzscher （ 2009） ，Zhang and Hu （ 2014） ，this paper adds tone，similarity and readability to the EGAＲCH model to explore whether textual indicators of the reports affect stock market returns and the volatility on the trading day after the release． Furthermore，this paper decomposes the content of the reports into two parts： economic and financial fundamentals and central bank policy guidelines，calculates the tone of the two parts and examines their impacts on the stock market respectively．...
