中国物理B ›› 2020, Vol. 29 ›› Issue (4): 48901-048901.doi: 10.1088/1674-1056/ab7740

• INTERDISCIPLINARY PHYSICS AND RELATED AREAS OF SCIENCE AND TECHNOLOGY • 上一篇    下一篇

Network correlation between investor's herding behavior and overconfidence behavior

Mao Zhang(张昴), Yi-Ming Wang(王一鸣)   

  1. 1 School of Economics, Peking University, Beijing 100871, China;
    2 Key Laboratory of Mathematical Economics and Quantitative Finance, Peking University, Beijing 100871, China
  • 收稿日期:2019-12-02 修回日期:2020-01-13 出版日期:2020-04-05 发布日期:2020-04-05
  • 通讯作者: Mao Zhang E-mail:zhangmao@pku.edu.cn
  • 基金资助:
    Project supported by the Youth Program of the National Social Science Foundation of China (Grant No. 18CJY057).

Network correlation between investor's herding behavior and overconfidence behavior

Mao Zhang(张昴)1, Yi-Ming Wang(王一鸣)1,2   

  1. 1 School of Economics, Peking University, Beijing 100871, China;
    2 Key Laboratory of Mathematical Economics and Quantitative Finance, Peking University, Beijing 100871, China
  • Received:2019-12-02 Revised:2020-01-13 Online:2020-04-05 Published:2020-04-05
  • Contact: Mao Zhang E-mail:zhangmao@pku.edu.cn
  • Supported by:
    Project supported by the Youth Program of the National Social Science Foundation of China (Grant No. 18CJY057).

摘要: It is generally accepted that herding behavior and overconfidence behavior are unrelated or even mutually exclusive. However, these behaviors can both lead to some similar market anomalies, such as excessive trading volume and volatility in the stock market. Due to the limitation of traditional time series analysis, we try to study whether there exists network relevance between the investor's herding behavior and overconfidence behavior based on the complex network method. Since the investor's herding behavior is based on market trends and overconfidence behavior is based on past performance, we convert the time series data of market trends into a market network and the time series data of the investor's past judgments into an investor network. Then, we update these networks as new information arrives at the market and show the weighted in-degrees of the nodes in the market network and the investor network can represent the herding degree and the confidence degree of the investor, respectively. Using stock transaction data of Microsoft, US S&P 500 stock index, and China Hushen 300 stock index, we update the two networks and find that there exists a high similarity of network topological properties and a significant correlation of node parameter sequences between the market network and the investor network. Finally, we theoretically derive and conclude that the investor's herding degree and confidence degree are highly related to each other when there is a clear market trend.

关键词: complex network, time series, herding behavior, overconfidence behavior

Abstract: It is generally accepted that herding behavior and overconfidence behavior are unrelated or even mutually exclusive. However, these behaviors can both lead to some similar market anomalies, such as excessive trading volume and volatility in the stock market. Due to the limitation of traditional time series analysis, we try to study whether there exists network relevance between the investor's herding behavior and overconfidence behavior based on the complex network method. Since the investor's herding behavior is based on market trends and overconfidence behavior is based on past performance, we convert the time series data of market trends into a market network and the time series data of the investor's past judgments into an investor network. Then, we update these networks as new information arrives at the market and show the weighted in-degrees of the nodes in the market network and the investor network can represent the herding degree and the confidence degree of the investor, respectively. Using stock transaction data of Microsoft, US S&P 500 stock index, and China Hushen 300 stock index, we update the two networks and find that there exists a high similarity of network topological properties and a significant correlation of node parameter sequences between the market network and the investor network. Finally, we theoretically derive and conclude that the investor's herding degree and confidence degree are highly related to each other when there is a clear market trend.

Key words: complex network, time series, herding behavior, overconfidence behavior

中图分类号:  (Economics; econophysics, financial markets, business and management)

  • 89.65.Gh
64.60.aq (Networks)