金融资产交互相关的噪音干扰

The Noise to the Correlations Between Financial Returns

本文应用并扩展随机矩阵理论的检验方法,实证检验资产组合协方差矩阵的噪音干扰。实证发现,经验估计的协方差矩阵存在较高程度的噪音干扰。表现在,经验协方差矩阵的特征值较好地吻合随机矩阵的理论分布,77.53%的特征值落在随机矩阵的理论取值范围内,并具有随机矩阵的普适性质。在过滤噪音后,资产组合的风险估计偏误得到显著降低,最小方差组合在评价期的风险水平显著降低。

The noise to the correlations between stocks returns are tested by using the RMT method.77.53% of the eigenvalues of the sample covariance matrix are found to fall within the RMT bounds and agree with the universal properties predicted by RMT-implying a large degree of noise.The noise content of the covariance matrix is filtered by applying Mean-value,and Zero-value schemes.With the filtered matrix,the ex post risk of the minimum variance portfolio is reduced significantly.