Abstract: By splitting the large commissioned orders and tracking the average market transaction prices, the traditional VWAP(volume weighted average price)strategies can minimize the market impact costs. Meanwhile, forecasting accurately the distribution of the intraday volume is the key to use the VWAP strategies to split orders. Investigating in detail the modeling methods and forecasting trading volume of the existing improved VWAP strategies, we find that the periodic structure of stock’s intraday volume is difficult to be separated completely, and the existing models are limited by the sample and difficult to apply to most stocks. From the perspective of the relationship between stocks’ volume and the trends of market volume, we deduce the equation of the stock volume and market volume, construct individual stocks’ factor loading on the market volume effects, and then decompose the intraday volume into two parts to forecast the trading volume distribution: one part describing the volume changes affected by the market; the second part reflecting the specific volume changes of the individual stock. Moreover, we build a dynamic VWAP strategy. The empirical results show that our trading volume decomposition model can effectively separate the periodic structure of the stocks’ intraday volume. Compared with the traditional VWAP strategies, our improved dynamic VWAP strategy has not only broader applicability but also smaller tracking error, and reduces the market impact costs more effectively.

Key words: algorithmic trading, VWAP strategy, trading volume decomposition, periodic structure of stock’s intraday volume

摘要： 传统VWAP(交易量加权平均价格)策略通过拆分大额委托订单,跟踪市场成交均价,达到最小化冲击成本的目的,而准确预测成交量日内分布是运用VWAP策略的关键。通过详细考察现有的改进VWAP策略中成交量预测模型的建模方式和预测结果,发现由于无法分离成交量日内周期结构,现有模型样本依赖性较大且难以适用于多数股票。因此,本文从个股与市场成交量变化趋势的关系角度出发,推导个股成交量与市场趋势的关系,通过构造个股成交量关于市场因素的因子载荷,将日内成交量分解为市场共同部分和个股特殊部分,预测成交量日内分布并构建动态VWAP策略。实证结果表明新的成交量分解模型可以有效分离个股的成交量日内周期结构,在此基础上构造的改进VWAP策略不仅具有较为广泛的适用性,且跟踪误差减少幅度比现阶段同类型的改进VWAP策略更大,能更好的降低市场冲击成本。

关键词: 算法交易, VWAP策略, 成交量分解, 日内周期结构