资产组合协方差矩阵的信息结构

摘要

本文应用随机矩阵理论及方法解析资产组合协方差矩阵的信息结构。实证发现，我国股票组合的协方差矩阵存在市场因素、行业因素的信息结构。最大特征值携带着反映市场因素的信息，影响程度非常高，是股票资产间相关性的主导因素。最大特征值偏离 RMT 上界的倍数远远高于其他市场。偏离 RMT 上界的特征值，携带着反映行业因素的信息，影响同一行业或者主营业务相同的公司，但随着特征值与 RMT 上界靠近，信息特征减弱。%We use the methods of random matrix theory(RMT) to investigate the information structure of the covariance matrix between stock returns. The largest eigenvalue are found to represent the market information. Chinese stock portfolio has a particularly high value of largest eigenvalue, which is about 175 times larger than the RMT upper bound. Market information is the dominate factor determining the correlations between stock returns. The other eigenvalues deviating from the RMT upper bound represent information affecting stocks belonging to similar or related industries.