基于Black-Litterman模型下的带流动性风险测度约束的资产配置模型

【作者】 马家驹；

【导师】 李胜宏；

【作者基本信息】 浙江大学 ， 运筹与控制， 2005， 硕士

【摘要】 在资产配置的相关理论中，传统的Markowitz(1952)均值-方差模型模型奠定了数量投资理论的基石，虽然它的算法简单，但其结果的高度集中性、对输入数据的高度敏感性限制了它在实际投资中的应用。特别是其资产预期收益率单纯从历史数据中获取，而没有与投资者的主观预期相结合得到更具现实意义的投资组合。Black-Litterman模型(1990)由高盛(Goldman Sachs)的Fisher Black和Robert Litterman提出，也是基于Markowitz的均值-方差模型，但它导入了投资者对某项资产的主观预期，使得根据市场历史数据计算预期收益率和投资者的看法结合在一起，形成一个新的市场收益预期，从而使得优化结果更加稳定和准确。但是我们知道金融市场的总体风险分为两部分：一部分是由资产价格波动导致的投资者收益的不确定性，这是人们通常所认为的纯市场风险：另一部分是市场流动性风险，它是指投资者在市场交易过程中以市场中间价难以买到所要投资的资产或难以卖出手中的某种资产的头寸。流动性约束在国内市场的资产配置中有很大的影响，模型中要深入分析。而B-L模型恰恰忽视了这部分重要的风险，因此本文试图在B-L模型的基础上导入流动性风险测度约束，提出新的优化资产配置模型和方法，改进单纯基于历史收益率和历史波动率的资产组合选择模型，并对流动性风险进行限定，最终使得最优解上移，并基于中国的A股市场对模型结果进行绩效评估和实证分析，从而构建起一套完善严密的资产配置组合体系。更多还原

【Abstract】 In the theories of asset management, traditional Markowitz mean-variance model is the foundmental theory of the quantitive portfolio research. Thogh the approch is quite classical, its high sensitivity on inputs (return estimates) retricts its application of the investment activities. Espesially the forecasts of asset returns are absolutely judged by historical returns, ignoring the intestor’s own subjective forecasts of the asset .The Black-Litterman Model (1990) yields consistent asset return forecasts as a weighted combination of market equilibrium returns and the intestor’s own subjective forecasts of the asset ("views"). So the B-L Model revised returns then serve as a consistent input for mean-variance portfolio optimisation procedures. But as we know, there are two parts of the overall risk: one is pure risk resulting from moves in the mid-price, the other important part is liquidity risk arising from a trader not realizing the mid-price. While B-L Model just has focussed on the pure risk ignoring the liquidity risk. So, on the basic of B-L Model, we want to build a new reformed frame to add liquidity risk ristriction as an important component to mean-variance portfolio optimisation model. In the end we have tested the new model using the data of Chinese A shares market. The result demonstrates that the optimized portfolio is better than B-L model.更多还原