【作者】 江林鑫；

【导师】 郑振龙；

【作者基本信息】 厦门大学 ， 金融工程， 2009， 硕士

【副题名】一种基于Black-Litterman混合方法模型的研究

【摘要】 资产配置理论是现代投资理论中的重要组成部分。通过稳健合理的资产配置,不仅可以增加投资者利润,而且可以有效规避投资风险。资产配置策略作为基金操作的基本原则,可以解释91.5%的基金回报率波动。因此,资产配置对于机构投资者,尤其是基金管理人提高投资绩效、分散并且降低投资风险都起着非常重要的作用。近几年,随着我国基金市场的迅猛发展,国内对资产配置模型研究的关注也越来越多目前,由高盛资产管理公司正式提出的Black-Litterman资产配置模型在国际上有着广泛的应用。但由于绝大多数的金融时间序列具有尖峰、厚尾、有偏的特征,常拒绝金融资产正态分布的假设,各资产之间还有着时变相关性、聚类效应及预测误差的异方差效应等其他无法被正态分布所刻画的特征的原凶,原始的Black-Litterman模型已经不能很好地解决上述问题带来的估计缺陷。因此本文在运用原始的Black-Litterman进行优化配置的过程中,还同时引入了Copula函数和DCC多元Garch模型对这一问题进行修正。通过Copula理论的应用,抹平了投资者对未来资产收益看法的分布与市场先验隐含分布之间的差异,并放开了投资者观点与市场先验的正态分布假设;通过DCC-MVGARCH的应用,在联合衡量资产间聚类效应以及异方差效应的同时还考虑了变量间相关系数的时变特征。这两方面都是这一领域相关研究中尚未涉及的。本文选取了我国GICS行业一级指数,完成了混合方法的BLCOP-DCC模型参数估计和业绩比较。实证结果表明,在有无卖空的两种情况下,都取得了高于Markowitz以及GICS行业综合指数、上证综指的收益率表现;组合的波动性小于Markowitz以及GICS行业综合指数。在卖空限制条件下,于2008年的下跌行情中,BLCOP-DCC配置相对于Markowitz组合具有经风险调整后的更小损失。从整体来看,混合方法的BLCOP-DCC模型是一种较为理想的策略。更多还原

【Abstract】 Asset allocation is the most important part of modern investment theory. Through the stable and reasonable asset allocation, investors can not only earn more profit, but also effectively avoid investment risk. As funds’ basic operation principle, asset allocation strategy can interpret 91.5% of their return. So, for institutional investors, especially for fund managers, asset allocation plays a very important role in increasing the investment performance as well as diversification of risk. In recent years, along with the rapid development of the fund market, the research of asset allocation model is paid more and more attention.Nowadays, an asset allocation model of Black-Litterman formally proposed by Fischer Black and Robert Litterman (1991) of Goldman Sachs is widely used in the world. However most of financial time serials have the property of excess peakness, fat tails and skewness, they often reject normal distribution hypothesis. Also, there exist time varied correlation, volatility clustering and heteroscedasticity. The original Black-Litterman model cannot solve above problems properly. This paper attempts to use the original Black-Litterman model to estimate and optimize, and meanwhile we introduce the Copula methods and DCC-MVGARCH. The Copula can erase the discrepancy of investor’s view distribution and the implied market prior distribution, while the DCC model can help with the volatility clustering, heteroscedasticity and time varied correlation jointly.This paper adopts GICS industry index in China to conduct the Empirical Analysis. The conclusion shows that whether under the constraint of short sale or not, our method does achieve superior returns to Markowitz method and benchmark of GICS and SSE Composite Index. The volatility of our portfolio is smaller than other methods mentioned above. Under the constraint of short sale, our method "BLCOP-DCC" portfolio has the less loss than Markowitz in 2008. Overall, the blended method—BLCOP-DCC can be a more desirable strategy .更多还原