基于GARCH类和SV类模型的中国债券市场实证分析

发布时间：2019-04-25 18:47

【摘要】：波动性普遍存在于各种金融时间序列中,是金融研究领域一个核心的问题。目前用于研究金融时间序列波动的模型主要有两类：一类是自回归条件异方差(ARCH)模型,另外就是随机波动(Sv)模型。这两类模型在近年的实证研究中得到了进一步的发展,例如ARCH类模型的扩展模型GARCH类模型,sv模型的扩展模型厚尾sv模型等。目前我国已有一些研究用这两类模型模拟我国的金融市场,其中主要包括股票市场,期权市场。但是,对于债券市场的研究很少,因此,本文用这两类模型以及其相应的扩展模型对我国的债券市场进行模拟,并通过引入了一系列的评价指标,客观的比较了GARCH类模型和SV类模型的波动性预测能力。本文先运用TGARCH模型和EGARCH模型对国债、企业债、金融债指数收益率数据进行实证分析,然后根据MCMC方法对Sv类模型中的SV-N、SV-T、SV-MN、SV-MT SV-Leverage模型进行了贝叶斯分析。实证结果发现,国债市场、企业债市场、金融债市场都表现出一定的波动集群性、尖峰厚尾性和非对称性。本文最后引入了RMSE、MAE、LL等评价指标对本文选用的模型进行了样本外预测能力的比较。结果发现SV类模型对我国债券市场的预测能力明显强于GARCH类模型。因此,本文得到结论,SV类模型比GARCH类模型能够更好的模拟我国债券市场的波动性特征。
[Abstract]:Volatility widely exists in various financial time series, is a core issue in the field of financial research. At present, there are two main models used to study the fluctuation of financial time series: one is the autoregressive conditional heteroscedasticity (ARCH) model, the other is the stochastic fluctuation (Sv) model. These two types of models have been further developed in recent years, such as the extended model of ARCH class model, the extended model of GARCH class model, the extended model of sv model, the thick tail sv model, and so on. At present, there have been some studies using these two models to simulate the financial market of our country, including stock market and option market. However, there is little research on bond market, so this paper uses these two kinds of models and their corresponding extended models to simulate the bond market of our country, and introduces a series of evaluation indexes. The volatility prediction ability of GARCH model and SV class model is compared objectively. In this paper, the TGARCH model and EGARCH model are used to analyze the yield data of national debt, enterprise bond and financial bond index. Then, according to the MCMC method, the SV-N,SV-T,SV-MN, in the Sv class model is analyzed. Bayesian analysis is performed on the SV-MT SV-Leverage model. The empirical results show that the bond market, the corporate bond market and the financial bond market all show some volatility clustering, peak thick tail and asymmetry. At the end of this paper, RMSE,MAE,LL and other evaluation indexes are introduced to compare the out-of-sample prediction ability of the model selected in this paper. The results show that the forecasting ability of the SV model is stronger than that of the GARCH model in the bond market of our country. Therefore, this paper draws a conclusion that the SV model can simulate the volatility of China's bond market better than the GARCH model.
【学位授予单位】：南京大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：F832.51;F224

【参考文献】