价格跳跃前大中小单的行为特征和信息含量

Behavioral characteristics and informativeness of large,medium and small orders before price jumps

利用中证800指数成分股3年的分笔交易数据,将股票成交订单分成大单、中单和小单,然后分析跳跃前的交易指令与跳跃的关联性及其对跳跃的预测能力.对跳跃样本和匹配有消息发布的跳跃样本进行事件分析和回归分析,发现跳跃前有明显的、逐渐加强的、与跳跃方向相对应的异常交易,而且异常交易指令与跳跃的发生概率、方向、大小和匹配的信息有不同程度的显著关联性;进一步的预测分析也发现,在控制流动性指标后,异常交易指令可以提升模型对跳跃的预测效果.这些结果显示市场在跳跃前存在着信息泄露,且信息泄露随时间扩散.此外,事件分析、回归模型和预测效果均显示:跳跃前中单的信息含量一般高于大单和小单,说明中国股市在极端价格变化前存在隐秘交易。

Based on tick data of CSI800 index components during three years,this paper classifies the transaction orders into large,medium and small ones,and analyzes the relationship between pre-jump orders and jumps and the predictive power of pre-jump orders on jumps.Event study and regressions based on both the jump sample and information matched with the jump sample show that there are obvious and gradually strengthening abnormal trades before the jumps consistent with the jump directions,and that these abnormal trades have significant correlations with the probability,direction,size and matched information of the jumps.Further forecasting analyses show that abnormal trades can also promote the model's performance on forecasting jumps after controlling liquidity measures.These results indicate that there are information leakages before jumps in the market,and that the leakages spread over time.Besides,event study,regressions and forecasting power comparisons all show that the medium orders before jumps are more informative than large and small orders,so stealth trading exists before extreme price changes in Chinese stock markets.