量化投资：从行为金融到高频交易

发布时间：2018-05-26 10:47

本文选题：行为金融 + 动量效应　；参考：《华东师范大学》2013年博士论文

【摘要】：量化投资作为一种新的投资方法,在海外的发展已有30多年的历史,由于投资业绩稳定,其所占市场规模和份额不断扩大。在国内,量化投资虽仍处于起步阶段,但其发展迅速,所受市场的重视程度也超过其他投资方法。交易策略是量化投资中的核心问题。一项投资是否成功,其效果如何,很大程度上由交易策略决定。鉴于投资策略的重要性,本文尝试研究行为金融和高频交易中的的几种交易策略,并讨论了它们在实际投资中的应用。在行为金融中,我们重点讨论动量策略的相关问题(可参考Wang,S.and Zheng, W.A.[93]).在实际金融市场中,有很多频繁出现的金融现象无法用传统金融给出解释,人们把这些金融现象称为金融异象。在这些金融异象中,动量效应、反转效应是两类重要的现象。关于动量效应,已经有大量学者研究了它的存在性及来源。在关于它的来源中,Hong-Stein模型是典型的将动量和反转效应组合在一起的模型。在Hong-Stein模型下,市场能产生动量效应及反转效应。然而,在其能解释动量效应的同时,该模型也易于产生两种极端情形。一方面,基本均衡价格函数下的收益率序列中的自相关性过于显著,使得价格变动方向可以用最临近时期的价格变动方向预测；另一方面,在推广的均衡模型中,收益率序列的自相关性过于不明显,这使得收益率序列几乎成为相互独立序列。对于这两种情形,我们构造了相应的检验统计量,并在各国家指数上进行了检验。检验结果标明,这两种极端情形均被拒绝。此外,我们还检验了传统动量策略在A股市场及商品期货市场上的有效性。检验结果表明,在A股市场上,动量策略不能产生显著正收益；而在商品期货市场上,动量策略却能带来显著的正收益。基于对Hong-Stein模型的讨论,我们构建了随机持有期动量策略,并重新在A股市场及商品期货市场上检验其有效性。检验结果表明,在A股市场及商品期货市场上,随机持有期动量策略均能产生显著的正收益。经过风险分析,我们发现策略的收益无法用相关的风险因子进行解释,这也说明了中国证券市场非弱有效。在对动量策略的研究中,我们发现,作为一种低频交易策略,虽然动量策略可以容纳大量的资金在数量众多的标的资产进行投资,也可以产生显著的正收益,但动量策略本身有无法克服的问题。这些问题主要包括：持仓时间长,占用资金成本高；在持仓过程中,账面收益易出现大幅波动,其中较大的收益回撤会对投资者的心理产生很大的压力；长时间持仓也使得收益率序列本身波动率加大,这使得策略的夏普比率不高。此外,随着交易所公布的数据越来越详细,高频数据会提供低频数据之外的很多信息,比如市场微观结构的信息,而这些信息在低频交易策略中并没有考虑到。因此,这启发我们研究高频交易策略,以克服低频交易中的不足。在高频交易中,本文讨论了基于一类技术指标(抛物反转指标,]Parabolic SAR)的交易策略的构建,并讨论了它的应用(可参考Wang,S. and Zheng, W.A.[94])技术分析是高频交易策略中的重要的组成部分,而技术指标是技术分析的基础。技术指标多以价格为基础,而资产定价是数理金融的核心问题之一,因此,在数理金融的框架下讨论基于技术指标的交易策略是恰当的。Parabolic SAR是一类重要的趋势性技术指标,因其具备加速追赶趋势的特点而与其他趋势类指标相异。根据SAR的定义,每个时间点上SAR的取值都依赖于初始值,这对于指标计算是繁杂的。本文中,我们讨论截断SAR指标的新定义,并由此得到新的SAR指标。因为SAR指标同样围绕价格走势波动,因此我们构造了价格序列与SAR之差作为新指标Xt。在Black Scholes模型下,我们证明了Xt的平稳性。模拟数据及基于模拟数据的检验同样支持指标平稳性的结论。基于Xt,我们构造出本文的交易策略。利用中国A股市场的高频数据,我们对该策略的有效性进行检验。在对策略有效性的检验中,我们采用了传统t检验对策略收益的显著性进行检验。对于在GARCH类模型下策略有效性的检验,我们采用残差bootstrap方法。检验结果表明,策略可以产生显著的买入收益和卖出收益。由于现阶段仍然是高频交易的快速发展阶段,高频交易也在为投资机构带来巨大的收益,因此,许多高频交易策略都处于保密状态,外界无法知道策略的具体内容。本文则尝试研究了一类具体的交易策略,并且经在实际投资部门的检验,策略也具备有效性。因此,不论从实际操作层面还是理论研究层面,本文对于高频交易的研究都很有意义。
[Abstract]:As a new investment method, quantitative investment has been developing abroad for more than 30 years. As the investment performance is stable, the scale and share of the market is expanding. In China, although the quantitative investment is still in its infancy, its development is rapid and the value of the market is more than the other investment methods. The trading strategy is quantified. At the core of the capital, the success of an investment is largely determined by the transaction strategy. In view of the importance of the investment strategy, this paper tries to study several trading strategies in behavioral finance and high frequency transactions, and discusses their use in real investment.
In behavioral finance, we focus on the related issues of momentum strategy (reference Wang, S.and Zheng, W.A.[93]). In real financial markets, many frequent financial phenomena can not be explained by traditional finance. These financial phenomena are called financial anomalies. In these financial anomalies, the momentum effect and reversal effect are two A large number of scholars have studied the existence and source of the momentum effect. In its source, the Hong-Stein model is a typical model that combines momentum and reversal effect. In the Hong-Stein model, the market can produce momentum effect and reversal effect. However, the momentum effect can be explained in the market. At the same time, the model is also easy to produce two extreme cases. On the one hand, the autocorrelation in the return sequence of the basic equilibrium price function is too obvious, making the price change direction can be predicted by the direction of the price change in the nearest period; on the other hand, the autocorrelation of the return sequence is too unknown in the extended equilibrium model. Obviously, this makes the return sequence almost independent sequence. For these two cases, we construct the corresponding test statistics and test the index in each country. The results indicate that these two extreme cases are rejected. In addition, we also test the traditional momentum strategy in the A stock market and the commodity futures market. The results show that momentum strategy can not produce significant positive returns in the A stock market, while momentum strategy can bring significant positive returns in the commodity futures market. Based on the discussion of the Hong-Stein model, we construct a random holding momentum strategy and check its effectiveness in the new A share market and commodity futures market. The results show that in the A stock market and the commodity futures market, the momentum strategy of the random holding period can produce significant positive returns. After the risk analysis, we find that the earnings of the strategy can not be explained by the related risk factors, which also shows that the Chinese securities market is not weak and effective.
In the study of momentum strategy, we find that as a low-frequency trading strategy, although momentum strategy can accommodate a large amount of funds to invest in a large number of standard assets, the momentum strategy can produce significant positive returns, but the momentum strategy itself has an insurmountable problem. These include the long holding time and the appropriation of funds. The cost is high; in the process of holding, there is a large fluctuation in the book income, of which a large return will have a great pressure on the investor's psychology; long holding also increases the volatility of the return sequence itself, which makes the SHARP ratio of the strategy not high. The data will provide a lot of information outside the low frequency data, such as the information of the market microstructures, which are not considered in the low-frequency trading strategy. Therefore, it inspires us to study high frequency trading strategies to overcome the shortcomings of low frequency transactions.
In the high frequency transaction, this paper discusses the construction of a transaction strategy based on a class of technical indicators (]Parabolic SAR), and discusses its application (reference Wang, S. and Zheng, W.A.[94]) is an important component of the high frequency trading strategy, and the technical index is the basis of technical analysis. Based on price, asset pricing is one of the core problems of mathematical finance. Therefore, under the framework of mathematical finance, the discussion of the trading strategy based on technical indicators is an appropriate.Parabolic SAR as an important trend technical indicator, which is different from other trend indicators because of its characteristics of accelerating the trend of catching up. According to the definition of SAR, The value of SAR at each time point is dependent on the initial value, which is complicated for the index calculation. In this paper, we discuss the new definition of the truncated SAR index and thus get a new SAR index. Because the SAR index also revolves in the fluctuation of the price trend, we construct the difference between the price sequence and SAR as the new index Xt. in the Black Scholes model We prove the smoothness of the Xt. The simulation data and the test based on the analog data also support the conclusion of the stability of the index. Based on Xt, we construct the trading strategy in this paper. We use the high frequency data of the Chinese A share market to test the effectiveness of the strategy. In the test of the strategy, we use the traditional t test. Test for the significance of policy gains. For the test of policy effectiveness under the GARCH class model, we use the residual bootstrap method. The results show that the strategy can produce significant buy returns and sell returns. As the current phase is still the rapid development stage of high frequency transactions, high frequency transactions are also brought to investment institutions. As a result, many high frequency trading strategies are in a state of secrecy, and the outside world can not know the specific content of the strategy. This paper tries to study a specific kind of transaction strategy, and the strategy is also valid in the actual investment department. The study of high frequency trading is of great significance.
【学位授予单位】：华东师范大学
【学位级别】：博士
【学位授予年份】：2013
【分类号】：F224;F830.59

【参考文献】