论文部分内容阅读
准确估计和预测金融资产价格波动是金融风险管理的核心问题。本文将厚尾分布纳入混合频率模型,采用上证综合指数2004-2011年滚动时间窗样本,利用SPA检验和MCS检验对低频、高频和混合频率三类波动模型预测精度进行比较。从风险价值的准确性和预测失败的损失程度角度对比三类模型的风险管理效果。通过对不同损失函数和不同时间窗口的分析发现,在10个波动预测模型中,扩展后的混合频率Realized GARCH_t在预测日波动时具有较高的预测精度,能准确计算短期风险价值,而在预测周或者月波动时,半参数NSM模型预测精度更高,风险价值更准确。
Accurately estimating and forecasting the price volatility of financial assets is the core issue of financial risk management. In this paper, the thick tail distribution is included in the mixed frequency model. The rolling time window samples of Shanghai Composite Index from 2004 to 2011 are used to compare the prediction accuracy of the three types of fluctuation models of low frequency, high frequency and mixed frequency using SPA test and MCS test. The risk management effectiveness of the three models is compared from the perspective of the accuracy of the VaR and the degree of loss of failure prediction. Through the analysis of different loss functions and different time windows, Realized GARCH_t, which is an extended mixed frequency, has a higher prediction accuracy in predicting day-to-day fluctuations and can accurately calculate short-term VaR in the 10 volatility prediction models. However, Week or monthly fluctuations, the semi-parametric NSM model has higher prediction accuracy and more accurate risk value.