基于DNS模型的我国公司债信用利差预测

利用2012~2017年我国国债、公司债、宏观经济及公司特征的月度数据，基于无风险收益率和公司债信用利差期限结构的动态Nelson-Siegel（DNS）模型，提取无风险收益率曲线和公司债信用利差曲线的水平因子、斜率和曲率，采取向量自回归模型和面板回归方法对我国公司债信用利差进行样本内和样本外预测。结果表明：我国公司债信用利差可以由无风险收益率期限结构及公司债信用利差期限结构中包含的信息进行预测；宏观经济变量和公司特征变量对我国公司债信用利差的预测贡献较小。

Forecasting credit spreads of Chinese corporate bonds with the dynamic Nelson-Siegel (DNS) model

With the recent spate of bond defaults, bond market risks have attracted more and more attention. Based on the riskless yield term structure dynamic Nelson-Siegel (DNS) model and the corporate bond credit spread term structure DNS model, this work employs the monthly data of Chinese treasury bonds, corporate bonds, macroeconomic variables as well as corporation-specific variables from 2012 to 2017 and adopts a panel regression method to conduct in sample and out of sample forecasting. The results show that future credit spreads of Chinese corporate bonds can be predicted by current riskless yield term structures and credit spread term structures, whilst macroeconomic variables and corporation-specific variables provide limited predictability.