基于高频交易数据的“羊群行为”测度模型  

A Measurement Model of Herd Behaviour Based on High Frequency Data

作　　者：朱菲菲[1] 唐涯[1] 徐建国[2,3] 李宏泰 Zhu Feifei;Tang Ya;Xu Jianguo;Li Hongtai(Guanghua School of Management, Peking University;National School of Development, Peking University;Institute of Digital Finance, Peking University;Zhangzhou Municipal People's Government Office)

机构地区：[1]北京大学光华管理学院 [2]北京大学国家发展研究院 [3]北京大学数字金融研究中心 [4]漳州市人民政府办公室

出　　处：《数量经济技术经济研究》2019年第8期129-146,共18页Journal of Quantitative & Technological Economics

基　　金：国家自然科学基金项目(71472006、71772004)的资助

摘　　要：研究目标:对“羊群行为”测度进行理论建模和实证检验。研究方法:通过结构化方法构建拓展的“羊群行为”测度模型,使用递归方法和极大似然估计方法对参数进行估计,并基于中国和美国市场的高频交易数据对“羊群行为”程度进行测量。研究发现:拓展的“羊群行为”测度模型不仅可以对“羊群行为”的短期脆弱性特征加以刻画,而且可以较好地区分“伪羊群行为”和“基于信息的羊群行为”。拓展模型下,知情交易者交易大额单的比例显著高于噪声交易者,而且发生“羊群行为”的知情交易者占比也更高。研究创新:创新性地考虑“交易单顺序”和“交易单规模”,并使用高频数据对“羊群行为”程度进行度量。研究价值:更加准确地识别和度量“羊群行为”,提高金融市场定价效率。Research Objectives:Estimating the herding behavior using structural model and empirical analysis.Research Methods:Construct a structural model of herding behavior,estimate model s parameters using recursive method and maximum likelihood estimation.Besides,test the model empirically by using the high frequency data of China and American stock market.Research Findings:The extend structural model not only can capture the short-lived and fragile characteristics of herding,but also distinguish“pseudo herding”and“informational herding”.Under the new model,the probability of“large orders by informed traders”is much higher than the probability of“larger orders by noise traders”,and more informed traders are participated in herding.Research Innovations:By taking“the order of transaction”and“transaction size”into consideration,we can capture herding behavior more accurately.Research Value:Accurately identify and measure herding behavior and improve the price discovery function of the stock market.

关 键 词：结构模型 高频交易数据 交易单顺序 交易单规模 

分 类 号：F831[经济管理—金融学]