随机波动率模型的参数估计及对中国股市的实证

基于有效重要性抽样(EIS)技巧，提出极大似然(ML)方法估计了四种不同收益分布假定的随机波动率(SV)模型的参数. 以上证综合指数和深证成份指数为例，实证检验了不同收益分布假定的SV模型的性能，分析适合我国股票市场的SV模型及收益分布. 实证结果表明，与正态分布、学生t-分布和广义误差分布(GED)假定的SV模型相比，具有“有偏”和“尖峰厚尾”特征的有偏学生t-分布假定的SV (SVSKt)模型能够更好地描述中国股票市场的波动性.

Estimation of stochastic volatility models: An empirical study of China's stock market

Based on efficient importance sampling (EIS) technique, we propose the maximum likelihood (ML) method to estimate the stochastic volatility (SV) models based on four different return distributions. Taking Shanghai Stock Exchange composite and Shenzhen Stock Exchange component indices as an example, we empirically test the performance of the SV models based on different return distributions, and aim to find the appropriate return distribution for China's stock market. Empirical results demonstrate that the SV model based on the skew student's t-distribution (SVSKt model) which can account for skewed and peaked and heavy-tailed returns provides significant improvement in model fit over the SV models based on the normal, the student's t and the generalized error distributions (GED).