主管:中国科学院
主办:中国优选法统筹法与经济数学研究会
   中国科学院科技战略咨询研究院

中国管理科学 ›› 2019, Vol. 27 ›› Issue (7): 23-34.doi: 10.16381/j.cnki.issn1003-207x.2019.07.003

• 论文 • 上一篇    下一篇

基于信息披露文本的上市公司财务困境预测:以中文年报管理层讨论与分析为样本的研究

陈艺云   

  1. 华南理工大学经济与贸易学院, 广东 广州 510006
  • 收稿日期:2018-01-03 修回日期:2018-08-10 出版日期:2019-07-20 发布日期:2019-08-01
  • 通讯作者: 陈艺云(1978-),男(汉族),湖南茶陵人,华南理工大学经济与贸易学院、金融工程研究中心副教授,硕导,博士,研究方向:金融风险管理、金融大数据分析,E-mail:royychen@scut.edu.cn. E-mail:royychen@scut.edu.cn
  • 基金资助:
    国家社会科学基金一般项目(15BJY149)

Forecasting Financial Distress of Listed Companies with Textual Content of the Information Disclosure: A Study based MD&A in Chinese Annual Reports

CHEN Yi-yun   

  1. School of Economics and Commerce, South China University of Technology, Guangzhou 510006, China
  • Received:2018-01-03 Revised:2018-08-10 Online:2019-07-20 Published:2019-08-01

摘要: 本文采用文本分析方法研究了中文年报管理层讨论与分析传递的管理层语调能否提供财务困境预测的增量信息、能否提高预测的准确性以及文本内容的信息价值,主要结论为:(1)管理层语调确实为财务困境预测提供了新的信息,能提高财务困境模型的拟合程度和预测能力;(2)管理层语调是对定量财务数据的重要补充,而且这些信息并没有在市场交易价格中得到充分反映;(3)负面语调比净语调具有更高的信息价值;(4)财经文本情感或语调的分析应以基于相关来源财经文本的情感词词典为基础,而直接引入其他领域成熟词典的效果较差。

关键词: 财务困境, 文本分析, 管理层语调

Abstract: Traditionally the prediction of financial distress is based on the quantitative information such as accounting data and market trading data, which has been proved as inefficient with a series of debt crisis after the subprime mortgage crisis. Quantitative data reflect the financial position of the company directly, while the qualitative textual content in the information diclosure reports is an important supplement, which may provide new clues for the prediction of financial distress since the wording and style may change with the financial postion of the company.
With various studies on the automatic textual analysis in the field of corporate finance, mainly on the stock market related issues, bag of words methods are applied to measure the management tone reflected in the Management Discussion & Analysis (MD&A) of Chinese annual reports, and test whether the management tone can provide additional information for financial distress prediction empirically. With different dictionaries, word segmentation tools and term weighting methods, a series of management tone variables are created, and added to various traditional financial distress prediction models.
Taking the special treatment (ST) as the symbol of financial distress, a sample of 2024 Chinese listed companies is selected.The emprical results from estimations with discrete-time hazard model, information content tests, in-sample and out-of-sample forecasting indicate:(1) management tone can provide new information for the financial distress prediction, and improve the fitness and predictive power of the financial distress prediction models; (2) management tone is an important supplement to the quantitative financial data, which have not been fully reflected in the market price; (3) the negative tone can provide more information than the net tone reflected in the textual content; (4) the tone or sentiment analysis of financial text should be based on the dictionaries created on similar text, but not the list of words from other non financial fields.

Key words: financial distress, textual analysis, management tone

中图分类号: