基于配对交易的统计套利策略研究

发布时间：2018-01-17 21:09

本文关键词：基于配对交易的统计套利策略研究　出处：《东华大学》2014年硕士论文　论文类型：学位论文

【摘要】：配对交易在国外证券市场是一种被广泛运用的统计套利投资策略。它通过构建股票配对的多空头寸,赚取股票价差收敛的收益。配对交易策略的一个显著的优点在于通过对冲机制有效规避了投资的系统性风险,即使在市场整体下行的时间段里,配对交易仍然能够获得比较稳定的收益。国外对配对交易策略研究的内容大致集中在两个方面,第一是研究如何挑选出性质良好的配对以及相关的交易模型,第二是如何制定最优的交易策略使得交易的效用函数最大化。在国内,自2010年3月起,我国证券市场相继开展了融资融券业务,自此,才有了对股票的做空机制,才有了配对交易的可能。因此,国内对配对交易的研究还较少,已有的研究大多基于2010年前的模拟数据研究配对交易在国内的可行性。总结已有的研究,之前的学者在实证中对待时间参数的设置问题并不重视,研究时间参数对配对交易策略影响的并不多见。究竟配对交易形成期和交易期长度的设定是否会影响配对交易的收益,如何合理的设定配对交易的交易期长度是一个有待研究的问题。本文使用2009年到2013年沪深300成分股作为实证对象,采用最小距离法进行配对交易实证检验。并通过调整不同的配对形成期与交易期的时间长度,研究配对交易的时间参数的设置对配对交易的收益是否能够造成影响。通过将配对交易的形成期和交易期长度分别设置为1个月到12个月共进行了144次不同的组合实验,发现不同的时间参数组合对配对交易的收益具有一定的影响。之后本文提出一种如何动态设定交易期长度的方法。利用Ornstein-Uhlenbeck过程的首次穿越时间理论构造了一个描述配对交易交易期长度的模型。该模型将配对交易的交易期分为空仓期和持仓期两个阶段。两个时间阶段的长度是分别可以用首次穿越时间理论描述的随机变量。通过计算这两个随机变量和的分布,我们得到完成一次交易所需时间的分布。根据这一分布本文提出了将配对交易交易期时间长度设定为以较大概率完成一次交易所需时间的策略。时间参数作为配对交易模型的重要部分,还有待深入的研究。只有当交易策略的时间参数和交易参数都达到最优时,交易模型才能获取最大的回报。
[Abstract]:Pairing trading is a widely used statistical arbitrage investment strategy in foreign securities markets. A significant advantage of matching trading strategies is that they effectively hedge against systemic risk, even during a period of downturns in the market as a whole. The content of foreign research on pairing trading strategy is focused on two aspects. The first is how to select good matching and related transaction models. The second is how to make the optimal trading strategy to maximize the utility function of the transaction. In China, since March 2010, the securities market has launched margin financing business since then. Only with the short selling mechanism of the stock, there is the possibility of paired trading. Therefore, there is less research on the matching transaction in China. Most of the existing studies are based on the simulation data before 2010 to study the feasibility of matching transactions in China. Summarizing the existing research, previous scholars pay little attention to the setting of time parameters in the empirical. It is rare to study the effect of time parameters on the paired trading strategy. Whether the setting of the forming date and the length of the paired transaction will affect the earnings of the paired transaction is rare. How to set the length of the transaction period is a problem to be studied. This article uses the Shanghai and Shenzhen 300 constituent stock from 2009 to 2013 as the empirical object. Using the method of minimum distance to carry out the empirical test of pairing transaction, and by adjusting the time length of different pairing forming period and trading period. To study whether the setting of time parameters for paired transactions has an impact on the earnings of the paired transactions. 144 times were performed by setting the formation date and the length of the trading period of the paired transaction to 1 month to 12 months, respectively. The same combination experiment. It is found that different combinations of time parameters have a certain effect on the earnings of paired transactions. Then this paper presents a method to dynamically set the length of the transaction period. Using the first traversal time theory of the Ornstein-Uhlenbeck process, we construct a method to describe the length of the paired transaction period. This model divides the transaction period of a pair of transactions into two stages: the short period and the position period. The length of the two time stages is a random variable which can be described by the theory of first pass time respectively. The distribution of the sum of machine variables. According to the distribution of the time required to complete a transaction, this paper proposes a strategy to set the length of the trading period of a paired transaction to the time required to complete a single transaction with a high probability. As an important part of the paired trading model, the time parameter needs to be studied deeply. Only when the time parameter and the transaction parameter of the trading strategy reach the optimum, can the transaction model obtain the maximum return.
【学位授予单位】：东华大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F830.91;F224

【参考文献】