受"维数灾祸"影响,金融收益相关矩阵往往包含噪声。从预期的角度,本文分析了金融相关噪声对马克维茨资产组合风险影响的机制。理论上讲,基于随机矩阵理论(RMT)的降维技术能够通过去除相关噪声而对投资组合风险进行控制。在证实我国股票收益相关矩阵存在大量噪声的前提下,为验证RMT去噪法用于我国股票投资组合风险控制的有效性,采用bootstrap方法,对已有三种RMT去噪法即LCPB法、PG+和KR法在我国股票投资组合构建中的应用进行了实证研究。得到两个结论,第一,这三种RMT去噪法都能通过股票收益相关矩阵去噪而带来我国股票投资组合风险的下降。且噪声越大,对组合风险的控制作用就越大。第二,KR法对组合风险的控制效果比其他两种方法更好,是一种更适合于我国股票市场的组合风险控制方法。
"Dimension disaster" makes financial return correlation matrices contain noise.From the perspective of expectation,this paper analyzes the mechanism of influence of financial correlation noise on Markowitz portfolio risks.Theoretically,dimension reduction technique based on RMT can control financial portfolio risk by removing correlation noise.On the premise of confirming the existence of noise in Chinese stock return correlation matrices,in order to test the effectiveness of RMT filtering for risk control of portfolios from Chinese stock markets,by means of bootstrapping technique,this paper carries out an empirical study using LCPB,PG and KR methods for constructing portfolios in Chinese stock markets.Two conclusions are drawn.First,it proves that all the methods can reduce risk of portfolios from Chinese stock market by filtering return correlation matrices,and that the greater the noise is,the larger the controlling effect on portfolios is.Second,the controlling effect of the KR method on portfolio risk is the largest of all the three methods,and the KR method is more suitable for risk control of Chinese stock portfolios.