基于分位数回归的企业债信用风险研究

将非参数GARCH模型的方差方程取对数变换后所得的模型用于估计企业债利差波动率。针对模型误差的非对称性，利用更为稳健的分位数回归方法估计改进后的非参数可加GARCH模型。实证分析结果表明，改进后的模型对波动率的估计更为有效；分位数回归方法比最小二乘回归方法能更有效的克服模型误差的非正态影响，对异常值的敏感程度更低，是一种非常稳健的估计方法。

A study of the credit risk of corporate bonds based on quantile regression

We have established a model which replaces the variance equation of the semi-parametric GARCH model with a logarithmic transformation in order to estimate the volatility of the credit spread of corporate bonds. The new model not only ensures the independent identical distribution of model error, but also ensures that the volatility is not negative after the logarithmic inverse transform. Since the model error is asymmetric, we use quantile regression, which is a more robust method than the least squares regression method, to estimate the modified nonparametric additive GARCH model. An empirical analysis shows that the estimation of volatility is more effective with the modified model. In particular, quantile regression is more effective than the least squares regression method in overcoming the model error when the error distribution is non normal and lower on the sensitive degree of the outlier.