Journal of Banking and Finance ·Volume 91, JUN 2018

定性相似性和股价联动

作者：Travis Box (School of Business Administration, University of Mississippi)

摘要：我提出了一种基于语言共性的用于评估新闻文本中企业特定信息的定性相似性的方法。即使在考虑了双方的同期价格变动、系统性风险的共同暴露、公司流动性、价格、指数成员、文本量、总部位置、产品相似性、共享的共同基金或机构所有权、分析师共同追踪和新闻共同提及的情况下，这种新的定性相似性度量方法也可以预测未来公司之间的收益相关性。我还发现了记者独自制作的内容无法预测具有经济意义的未来联动效应。样本外测试证实了对定性相似性的了解也可以降低投资组合的风险。

关键词：信息消费；信息驱动联动效应；文本分析；相关性

Qualitative similarity and stock price comovement

Travis Box (School of Business Administration, University of Mississippi)

ABSTRACT

I introduce a method for gauging the qualitative similarity of firm-specific information based on linguistic commonality in newswire text. I show that this new qualitative similarity measure predicts future cross-firm return correlation even after accounting for the pair's contemporaneous price comovement, common exposures to systematic risk, firm liquidity, price, index membership, text volume, headquarters location, product similarity, shared mutual fund or institutional ownership, common analyst following and newswire co-mentions. I also demonstrate that content produced solely by journalists cannot predict an economically meaningful portion of future comovement. Out-of-sample tests confirm that knowledge of qualitative similarity can also reduce portfolio risk.

Keywords: Information consumption; Information-driven comovement; Textual analysis; Correlation

原文链接：

https://www.sciencedirect.com/science/article/pii/S0378426618300773

翻译：张展