期货市场操纵行为判别与预警

Classify and Predict Manipulation in Futures Market

本文以我国期货市场中硬麦WT101与WT309两个疑似操纵案例为样本,应用面板数据处理技术,构建了一个多变量判别分析模型。结果表明模型能够很好地刻画出期货操纵行为的特征,模型预测正确率达到88.89%,在被操纵合约交割之前2个月,模型能持续地发出报警信号,给监管者足够的时间处理操纵风险。同时,本文应用Logistic回归分析结果表明,期现货价差与总持仓量等单变量也具有较好的预警效果。

This article addresses a multi-variable analysis model based on panel data of WT101 and WT309 contracts-two purported manipulation samples in China wheat futures market. This analysis shows how the characteristics of manipulations in the futures market can be predicted and explained by the model effectively with less than 11.11% forecasting error （88.89% reliable rate）. The model can also issue early warning 2 months in advance, which provides the regulation authority enough time to react. And the Logistic regression result shows that futures-case price basis and open interests are good indicators.