Abstract: The classical model of the probability of informed trading assumes that traders can make short-sell freely based on private information. This assumption, however, is violated in China's stock market due to the short-sell constraints. Therefore, the prediction of classical model may not hold in China’s stock market. Considering the status quo of the stock market in China, we develop a SC-TPIN model by incorporating two short-sell constraint variables into the classical model, and test our model using security lending stocks with bad news. Our model is well supported by the data. Base on the SCTPIN value estimated by SC-TPIN model, we develop the informed trading identification indicator group by using low frequency trading data of our sample stocks. We also develop our informed trading identification system by discerning and comparing the informed trading type of our black and white samples using support vector machine, KNN and Logit algorithm, respectively. The results show that the identification accuracy of support vector machine algorithm is as high as 89%, capable of identification effectively. In this paper, we prove that the order flow information contained in our SC-TPIN model is consistent with the actual order flow information, and the identification accuracy based on the SC-TPIN value is higher than that based on the TPIN value, showing that our SC-TPIN model can more effectively measure informed trading of stocks with bad news in China’s stock market.

Key words: short-sell constraint, PIN model, informed trading, classification and identification

摘要： 经典的测量知情交易概率的模型默认交易者可以无限制的按照私有信息进行卖空交易,而目前我国股票市场存在卖空限制,直接将经典模型应用到我国股票市场时会使测量结果出现偏差。考虑到我国股票市场现状,本文在经典的知情交易概率模型中引入两个卖空限制参数,构建了本文的SC-TPIN模型。通过对融券标的中发生利空消息的股票样本进行实证分析,证实了本文构建的SC-TPIN模型估计出的结果与实际情况相符合。本文还以SC-TPIN模型估计出的SCTPIN值为参照,基于样本股票的低频数据构建了知情交易识别指标组,并使用数据挖掘中的支持向量机算法、KNN算法及Logit模型对黑白样本的知情交易高低情况进行识别比较,构建知情交易识别体系,发现使用支持向量机算法识别全样本的正确率达到了89%,识别效果较理想。

关键词: 卖空限制, PIN模型, 知情交易, 分类识别