基于Copula理论的卖空交易策略研究

发布时间：2017-12-31 18:34

本文关键词：基于Copula理论的卖空交易策略研究　出处：《东北财经大学》2013年硕士论文　论文类型：学位论文

【摘要】：20世纪70年代,量化投资开始在美国的资本市场兴起,受益于量化投资的诸多优点,量化投资很快就成为美国主流的投资模式,在众多的量化投资者中,以西蒙斯最为成功,他的大奖章基金在20年内,持续而稳定的获得了每年平均35%的净回报,并且是在扣除费用后的,这样的成绩堪比神话。而我国的量化投资还处于起步阶段,但随着我国资本市场的逐渐完善,量化投资的前景会越发明亮,同时,量化投资的发展可以减少市场上的投机行为,减少市场泡沫,反过来促进我国资本市场的发展和完善。本文正是在此背景下,欲建立一种基于Copula理论的量化投资策略模型,运用于我国的资本市场上。投资者一般习惯性做多,本文一反常态重点研究卖空方式的建立,期望使投资者在市场下跌时也能获利,帮助投资者获取超额收益。本文选取豆油指数和棕榈油指数2012年9月至12月共四个月的高频数据作为研究对象。研究对象是由市场上的主力合约加权平均计算得到的,指数能体现研究对象的连续性,所以豆油指数和棕榈油指数是非常合适的研究对象。本文所设计的投资策略主要基于Copula函数刻画的尾部相关性,即在下尾部相关性高的情况下,如果豆油指数下跌,那么棕榈油指数下跌的概率是非常高的。假如现实中棕榈油指数并没有下跌,那么在Copula函数刻画的尾部相关性下是“非正常”情况,我们相信棕榈油指数将会下跌,此时就是投资的机会。由Copula函数的定义可知,Copula函数是两随机变量的边际分布的连接函数,为保证整个投资策略的成功,需要对两组研究对象的边际分布进行精确的描述,边际分布刻画的越准确,Copula函数拟合的越准确,则可能出现的投资机会越准确。由于金融资产的高频数据的分布不具有正态性,一般有“尖峰肥尾”的性质,本文也放弃了传统的假设收益率服从正态分布的方法,转而采用核密度估计这种非参数估计的方法,以求能够更精确的刻画研究对象的边际分布情况。利用本文所研究的策略对实际数据进行测试,本文得到了非常好的累积收益率,九月至十二月的累积收益率依次是17.70%、3.45%、8.97%、-0.82%,这在当今的金融投资理论中已经属于超额收益率。交易次数分别为96、57、79、6,总次数达到了238次,属于典型的高频交易,九月至十二月四个月中单笔最大亏损依次为-1.4699%。显然亏损幅度是我们能够接受的。本文实证的结论表明本文的投资策略是具有一定的实际参考意义的。本文的创新之处主要有三点：第一,本文打破投资者的惯性思维,重点进行卖空交易；第二,创造性的基于Copula尾部相关性构造程序化交易策略,传统的资产配置思想要求资产之间的相关性要低,以分散风险,本文则是利用尾部相关性高的资产来寻找交易时机,这也是本文最突出的创新点；第三,由于数据多,计算量大,在参数拟合的过程中为得到最优的参数组合,使用遗传算法能够避免穷举算法费时费力的缺点。本文的不足之处有两点；第一,Copula函数描述下尾相关性的能力远远高于其描述上尾相关性的能力,因此,投资者可以得到的买入信号较差,因此丧失了一部分的利润,而且由于上尾相关性刻画的精度不够,反而可能降低投资收益率；第二,虽然在实证过程中并没有出现不可接受的亏损,但是过度的依赖程序化交易而考虑情况不够细致的话,一旦极特殊的情况出现,可能导致较大的亏损。
[Abstract]:In 1970s, quantitative investment began to rise in the U.S. capital market, to benefit many advantages of quantitative investment, quantitative investment soon became the mainstream mode of investment, many investors in the quantification, with Simmons's most successful, his Medallion fund in 20 years, sustained and stable won the annual average net return 35%, and after deducting expenses, this result is comparable to myth and quantitative investment in China is still in the initial stage, but with the development of China's capital market is gradually improving, quantitative investment prospects will be brighter, and at the same time, the development of quantitative investment can reduce market speculation, reduce the market bubble. In turn, promote the development and perfection of China's capital market.
Under this background, to establish a quantitative investment strategy model based on Copula theory, used in China's capital market. Investors generally used to do more, this paper focuses on the establishment of uncharacteristically selling short, so that investors expect the decline in the market can profit, help investors to obtain excess returns.
This paper selects the high frequency data of soybean oil and palm oil index index from September 2012 to December a total of four months as the research object. The research object is the main contract on the market by the weighted average calculated index, can continuously reflect the research object, and soybean oil and palm oil index index is the object of study is very appropriate.
The design of this investment strategy is mainly based on the Copula function to describe tail dependence, namely in the tail correlation is high, if the oil index fell, then the probability of palm oil fell is very high. If the palm oil index in reality did not fall, then described in the Copula function is the tail correlation non normal situation, we believe that the palm oil index would fall, this is the investment opportunity. By the definition of the Copula function, Copula function is the link function of marginal distribution of two random variables, in order to ensure the success of the investment strategy, to the marginal distribution of two groups of subjects were accurate description, marginal distribution more accurately, the Copula function fitting is more accurate, it may be more accurate investment opportunities. Due to the distribution of high-frequency data of financial assets is not normal, there are The nature of "spiking fat tail" also abandons the traditional assumption that yield is subject to normal distribution. Instead of using kernel density estimation, this method of nonparametric estimation can be used to depict the marginal distribution of research objects more accurately.
To test the actual data by using the strategy, this paper obtained the very good cumulative rate of return, from September to December the cumulative rate of return was 17.70%, 3.45%, 8.97%, -0.82%, which belongs to the excess rate of return in today's financial investment theory. The number of transactions were 96,57,79,6, the total number reached 238 second, belongs to the typical high frequency trading, September to December four months, the single largest loss in -1.4699%. obviously loss rate is acceptable to us. The empirical results show that this investment strategy is of practical significance to a certain.
The innovation of this paper has three main points: first, this paper break the inertia of thinking investors, focusing on short selling; second, Copula tail correlation program trading strategy based on the structure of creative thought, the traditional asset allocation requirements of the correlation between the assets to be low, to disperse the risk, this is the tail of the high correlation of assets to looking for the timing of the transaction, which is the most important innovation; third, because most of the data, a large amount of computation in the process parameter fitting to obtain the optimal combination of the parameters, using genetic algorithm can avoid the shortcomings of time-consuming exhaustive algorithm.
There are two shortcomings in this paper; first, Copula function to describe the ability of lower tail dependence is much higher than the upper tail dependence of the description ability, therefore, the poor buy signal that investors can get lost a part of the profits, but due to lack of characterization of upper tail dependence accuracy, but may reduce the rate of return on investment; second. Although in the empirical process and no unacceptable losses, but excessive reliance on program trading and consider the situation is not careful, once the very special circumstances, may lead to greater losses.

【学位授予单位】：东北财经大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F832.51

【参考文献】