【作者】 凌士勤；

【导师】 唐齐鸣；

【作者基本信息】 华中科技大学 ， 西方经济学， 2006， 博士

【摘要】 随着中国经济的发展,投资业对投资组合理论和实践方法的需求越来越大,广大投资者也需要相应的适合中国股票市场的投资理论来指导实践。因此,本文将研究的主要范围设定为包括“战略投资”(Strategic Investment)、“战术投资”(Tactical Investment)及其绩效评估和风险评估在内的投资组合理论,以求达到如下目的:在“战略投资”层面上,本研究的目的为:1)从理论上,提出基于VaR-Kelly体系的新模型以弥补现有的投资组合模型所存在的缺陷,以满足实际投资活动的需求。2)在实践中,通过对不同投资策略在中国股票市场的实证来比较各类投资策略在中国股票市场的优劣,为投资者提供相应的参考。同时,对所提出的新模型在中国股票市场上进行实证。在“战术投资”层面上,本研究的目的为:1)从理论上,研究Black-litterman模型的求解方法,以期提供更加合理的求解方法。2)在实践中,用最优权重因子法对“战术投资”策略中的信号个数、信号类别(高频信号及低频信号)进行研究,以了解信号个数及信号类别对“战术投资”绩效的影响。在绩效评估和风险评估层面上,本研究的目的为:1)从理论上,研究新的基于VaR体系的投资绩效评估方法,并与其他方法做相应的比较。2)提出新的模型对分类信息条件下的风险进行估算。在实证方法上,本文以中国A股市场的部分行业的股票数据、计算机模拟的数据及上证指数的高频数据为研究对象,采用matlab,eview等工具进行编程及计算机仿真模拟。本文所作的主要工作和有关结论如下:首先,在对“不加入安全约束条件下各类投资策略”进行了研究,由于这些传统的投资策略没有加入安全约束条件,即没有把风险考虑进去。而且,从实证的结果来看,投资绩效不是很理想,因此,我们仅仅将此类投资结果列出作为广大投资者参考之用,不推荐在实际的投资活动中采用这些方法。其次,对“加入安全约束条件下各类投资策略”进行了研究,本文提出了基于VaR及Kelly增长体系的“基于最优增长路径的增长-安全模型”。该模型相对其他的模型而言,在风险度量上采用VaR方法,较以前的以方差等指标作为风险度量的方法有了一定的改进,在收益的衡量方面,采用Kelly增长体系替代传统的收益衡量体系,并在离散条件下,用基于情景分析的方法计算其相应的结果。该模型既值得作进一步深入研究,也可作为用于投资活动的一种实践工具。在对“战术投资”的研究上,本文讨论了“最优权重比例模型(OAF模型)”和Black-Litterman模型中的部分问题。我们发现:只要信号之间的α(超过基准的超额收益率)之间的相关系数不全为零,采用最优权重因子法(OAF)中,可以显著地提高“信息比率”。同时,建议使用高频信号,并说明了采用多的信号数量对提高“信息比率”作用不大。同时,本文对Black-Litterman模型的传统求解过程进行了质疑,指出了Idzorek(2002)的求解方法的错误之处,并给出了相应的解决方法。此外,在对“投资绩效评估”的研究上,本文提出了基于VaR体系的投资绩效评估方法:“收益-VaR比率”。研究结果发现,在正态分布条件下,用夏普比率和用收益-VaR比率得到的绩效排名是一致的。然而,在非正态分布条件下,两种方法得到的结果却有可能不相同。最后,在研究分类信息条件下的风险估算问题时,本文提出了基于高频数据的分类信息混合分布GARCH类模型,考察好消息、坏消息对上证指数波动性的影响,比较了不同GARCH类模型得出的结论。同时,用GARCH模型及EGARCH模型对上证指数进行了实证研究,结果发现:基于高频数据的分类信息混合分布的EGARCH模型较基于高频数据的分类信息混合分布的GARCH模型及Craig A.Depken(1999)提出的模型更符合实际情况,能更准确的衡量分类信息对波动的影响。更多还原

【Abstract】 With the development of Chinese economy, the demand of the portfolio theory of the investment industry has growing fast. Investors need the portfolio theory suitable for the Chinese stock market to guide the activities of the investment.Therefore, this research focuses on the portfolio theory, which includes strategic investment, tactical investment and the evaluation of the performance of the investment.At the level of strategic investment, the aim of this research is:1) To put forward a new model based on the VaR-Kelly framework to fulfill the demand of the investment activities theoretically.2) To make a comparison of various strategies of the asset allocation in the Chinese stock market, which is supposed to offer the investors the corresponding references and to do empirical studies on the newly built model empirically.At the level of tactical investment, the aim of this research is:1) To study the solution method of the Black-Litterman model to find a more reasonable solution method theoretically.2) To study some topics of the Optimal Aggressiveness Factors method, such as the impact of the optimal number of signals and the classified signals on the investment performance empirically.At the level of performance evaluation, the aim of this research is:1) To study a new method of the performance evaluation based on VaR system and make comparisons to other corresponding methods theoretically.2) To put forward a new model to evaluate the risk based on the classified information.To do the empirical work, we take the data of stocks of different industries in Chinese stock market, the high-frequency data of the Index of Shanghai and the simulated data as the object to study, and use the Matlab and Eview as the program tool to do the simulating work.The main results are listed as follows: Firstly, we study the investment strategies without safety constraints. Through the empirical results, we find that the performance of investment is not ideal enough for investors to use, so we do not recommend investors to take either of the strategies into practical applications.Secondly, we study the investment strategies with safety constraints and a multi-period investment strategy has been built based on VaR and Kelly growth theory, which has advantages either on the risk control or on the criterion of capital growth.We then do the empirical and theoretical studies on OAF model and Black-Litterman model, and find that the OAF model can improve the information ratio provided that the alphas among the signals are not all zeroes. And the empirical study has shown that the high-frequency data are suggested to adopt to improve the information ratio and that the practice of increasing the number of signals may have litter help to the improvement of information ratio.We also raise suspicion with the traditional process of solving the Black-Litterman model and bring forward the corresponding solutions.Then, we develop a VaR-based measure of portfolio performance that is closely related to the widely used Sharpe ratio, formerly known as the reward-to-variability ratio. Accordingly, we refer to the corresponding VaR-based measure as the reward-to-VaR ratio. And we find that under the normal distribution assumption, using Sharpe ratio and the reward-to-VaR ratio will have the same result, whereas if the distribution is not normal, the results of the two ratios may be different.Finally, the high-frequency-data-based classified information mixture distribution GARCH model and EGARCH model, which have been formed by absorbing and borrowing the previous relevant models and based on the high-frequency-data, is more coincident with the truth when it is demonstrated in ShangZheng Index.更多还原