摘要
针对我国债券市场,选取2005年6月至2010年6月的AAA级企业债和国债月度交易数据分别构建基于SV模型的利率期限结构,然后利用遗传算法求解来拟合较为精确的企业债和国债的即期利率曲线,据此计算信用价差.在对我国企业债按不同的期限进行时间序列分析后,发现各信用价差序列均是一阶单整序列,呈现自回归和移动平均的特征,信用价差ARMA模型的拟合优度相比多元线性回归模型有提高;信用价差VAR模型能很好的拟合多个期限企业债信用价差序列之间的动态相关关系;脉冲响应分析表明,各期限信用价差序列对其他期限信用价差序列的冲击在前10期会造成较为剧烈的波动,但幅度较低,且冲击不具有较长的持续效应.实证结果在一定程度上为投资者和监管者提供了决策依据.
Selecting the trading information bonds in China＇ s bond market on a monthly of corporate bonds with the credit rating of AAA and treasury basis, and solving the Svensson （1995） model of term structure of interest rates through genetic algorithm, the paper gives more exact yield curves of both types of bonds and derives the credit spreads. After time series analysis of China＇s bonds with different terms of maturity, it can be concluded that all the credit spread series have one unit root and show the characteristics of autoregressive and moving average. The R-squared of ARMA model gives better fitness than the multiple linear regression model. The VAR model can fit the dynamic relationship among credit spreads well. Impulse response analysis implies that the shock from other series of credit spreads to one sequence of credit spread will result in severe fluctuations with low magnitude in the first 10 terms and the impact does not have long and continuous effect. The empirical results provide the basis of decisions for investo~ and re^mlatnr~ tn ~nm~ ~Yt^nt
出处
《管理科学学报》
CSSCI
北大核心
2014年第3期37-48,共12页
Journal of Management Sciences in China
基金
国家自然科学基金资助项目(71171012
71071010
71371023)
