基于Black-Litterman模型的资产配置策略研究

发布时间：2018-07-09 17:37

本文选题：资产配置 + Black-Litterman模型　；参考：《山东大学》2017年硕士论文

【摘要】：随着金融产品供给丰富、机构和个人资产配置需求增加,如何进行合理有效的资产分配成为人们面临的重要议题,研究表明,资产选择和资产配置可以解释投资组合90%左右的业绩变动,自从Markowitz提出投资组合理论以后,Black和Litterman提出的将市场数据与投资人观点相结合的资产配置方法受到了广泛关注。本文运用Black和Litterman(1992)提出的Black-Litterman模型(BL模型),研究了基于2010年到2016年股票数据的行业配置策略,对前人的研究做了如下拓展:1、对BL模型逆优化阶段应该采用何种输入权重进行了理论梳理,并用Elton,Gruber,Padberg(1976)提出的资产选择模型(EGP模型)生成BL模型逆优化阶段的输入权重。2、指出了前人在设定观点方差时的一个错误。3、将回测期数延长,增加了仓位稳定性指标来评价策略。本文BL模型中的观点和观点方差一律用GARCH类模型生成,逆优化阶段所用的输入权重用EGP模型来生成,与输入权重直接采用市值权重的策略进行了比较,在大幅震荡和小幅震荡两种不同行情模式下,分别对1年的数据进行了回测,发现用EGP模型生成输入权重的策略在稳定性、收益性上均优于直接用市值权重的策略,为个人和机构投资者的资产配置提供了思路。本文分为六章,第1章阐述研究的背景,并对研究思路和创新点做了简要说明,指出了本文的研究意义。第2章对国内国外文献进行综述,简要阐述了投资组合、资产定价理论,重点梳理了前人对资产配置、Black-Litterman模型的研究,并对其研究特点进行了总结,发现前人对BL模型逆优化过程中的输入权重讨论较少,国内研究的回测期数大都很短,而且在策略评价时对策略的稳定性涉及不多,所以本文在这三方面进行了拓展。第3、4章是理论基础,从均值方差模型、资本资产定价模型开始论述,随后介绍了 Black-Litterman模型,从贝叶斯理论的角度给出了 BL模型的简要推导,并对各种输入量的确定进行了详细论述,重点介绍了前人研究中逆优化过程里的输入权重、观点及其方差的确定方法;对输入权重的确定,本文总结他人的研究方法后提出了自己的意见,并指出了前人研究中观点方差设定的错误;随后介绍了本文用来确定逆优化过程中输入权重的资产选择理论(EGP模型)。第4章介绍了本文衡量投资者观点时所用到的GARCH类模型和该模型的估计及预测。第5章是实证,这一部分首先对数据选择做了简述,对数据的基本统计性质进行了分析,其后进行BL模型输入量的求解,首先根据EGP模型进行了资产权重(作为BL模型的输入权重)的计算,随后确定各种行业指数的GARCH类模型形式,同时确定其模型系数并对模型进行检验,然后对回测期的观点和方差进行预测,最终将这些输入量代入BL模型公式,得到最终的资产配置权重。最后在两种不同行情模式下,分别对1年的数据进行了回测,统计了策略收益的各项评价指标,对策略在两种环境下的表现及适用性进行了分析,同时从权重累计变化的角度对策略的调仓特征进行了评价。本文的主要结论如下:1、对于BL模型中的输入权重,本文认为应该用以资本市场均衡为前提的模型求得,作为中性起始点。只考虑风险来构建起始输入权重这种途径还没有足够的理论支撑。2、基于因素模型的资产选择模型(EGP模型)生成的权重反映了样本期内各行业的概况,指数收益率与权重正相关;综合收益的求解过程中,模型赋予了引致均衡收益较大的权重,且后验BL收益率作为加权结果,其值处在观点和引致均衡收益中间。3、横盘小幅震荡(2016年)与大幅牛熊震荡(2015年)的环境中,用EGP模型生成的权重作为输入权重的BL模型在回测区间累计收益、收益稳定性、权重稳定性上都优于直接用市值权重作为输入权重的BL模型。
[Abstract]:With the rich supply of financial products and the increasing demand for the allocation of institutions and individual assets, how to carry out reasonable and effective allocation of assets has become an important issue for people. The research shows that asset selection and asset allocation can explain the performance changes about 90% of the portfolio. Since Markowitz put forward the portfolio theory, Black and Litterman have been proposed. The method of asset allocation which combines market data with investor views has been widely concerned. This paper uses Black and Litterman (1992) Black-Litterman model (BL model) to study the industry configuration strategy based on stock data from 2010 to 2016, and extends the previous research to the following: 1, the inverse optimization stage of BL model What kind of input weight should be used to sort out the theory and use the asset selection model (EGP model) proposed by Elton, Gruber, Padberg (1976) to generate the input weight.2 of the BL model inverse optimization stage, and points out a wrong.3 in setting the point of view variance, prolonging the number of back test period and increasing the stability index of the position to evaluate the strategy. The viewpoint and point of view variance in the BL model are generated by the GARCH model. The input weight used in the inverse optimization stage is generated by the EGP model, and is compared with the strategy of the value weighting directly using the input weight. In the two different market modes of large concussion and small amplitude concussion, the data of 1 years are measured, and EGP is found to be used. The strategy of model generating input weight is better than the strategy of directly using market value weight in stability and profit. It provides ideas for the asset allocation of individual and institutional investors. This paper is divided into six chapters. The first chapter expounds the background of the study, and gives a brief description of the research ideas and innovation points, and points out the significance of the study. The second chapter is to the country. The literature of internal and foreign countries is reviewed, and the portfolio and asset pricing theory are briefly expounded. The research on the assets allocation and Black-Litterman model has been combed, and the characteristics of the research are summarized. It is found that the predecessors have less discussion on the input weight in the inverse optimization process of the BL model, and the number of back test periods in the domestic research is mostly short, and The stability of strategy is not much involved in strategy evaluation, so this article has been expanded in these three aspects. Chapter 3,4 is the theoretical basis, from the mean variance model, capital asset pricing model, then the Black-Litterman model, from the perspective of Bayesian theory gives a brief derivation of the BL model, and a variety of input quantities. The determination is discussed in detail, and the input weight of the inverse optimization process in the previous study is introduced, the method of determining the point of view and its variance is introduced, and the input weight is determined. After summarizing the research methods of others, the author puts forward his own opinion and points out the error of the point of view variance setting in the previous study. In the fourth chapter, the fourth chapter introduces the GARCH model and the estimation and prediction of the investor's view. The fifth chapter is an empirical study. This part is a brief introduction to the data selection, analyses the basic statistical properties of the data, and then carries out the BL model. In order to solve the input, the weight of the assets (as the input weight of the BL model) is calculated according to the EGP model, and then the GARCH model of various industry indices is determined, and the model coefficient is determined and the model is tested. Then the viewpoint and the square difference of the back test period are predicted, and finally the input is replaced by the BL model public. Finally, we get the final asset allocation weight. Finally, under the two different market models, the data of 1 years are measured, the evaluation indexes of the strategy income are counted, the performance and applicability of the strategy under the two environment are analyzed. At the same time, the storehouse characteristics of the strategy are evaluated from the angle of cumulative weight change. The main conclusions of this paper are as follows: 1, for the input weight in the BL model, this paper thinks that the model should be obtained with the capital market equilibrium as the precondition, as a neutral starting point. There is not enough theory to support.2 and the right of the asset selection model based on the element model (EGP model). It reflects the general situation of each industry in the sample period, and the index return rate is positively correlated with the weight. In the process of solving the comprehensive income, the model gives the weight of the higher equilibrium income, and the posterior BL yield is the weighted result, and its value is in the middle.3 of the viewpoint and the induced equilibrium income, the horizontal shock (2016) and the big bull bear shock (2015) In the environment of the year), the BL model, which is generated by the weight of the EGP model as the input weight, has the cumulative income, the stability of the return and the stability of the weight, which is better than the BL model that directly uses the market value weight as the input weight.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F832.51

【相似文献】