基于周期GARCH过程VaR的分位回归估计

Quantile Regression Estimation for VaR of P-Garch Processes

近几年来,风险价值(VaR)已成为金融市场风险度量及风险管理的标准工具.文章用周期广义自回归条件异方差(GARCH)模型拟合金融市场数据,并应用分位回归方法得到此模型参数及条件VaR的估计,在一定条件下估计具有强相合性及渐近正态性,蒙特卡罗模拟结果表明此方法具有稳健性,且对于条件VaR的预测具有很高的准确性,沪深300指数的实证分析结果表明此方法关于VaR的预测具有非常好的效果.

Recently, Value-at-Risk （VaR） has become a standard tool for risk mea- sure and management. In this paper, we consider the periodic generalized autoregres- sive conditional heteroskedasticity （P-GARCH） process and propose a robust estimate for parameters and conditional VaR. Under some mild condition, our proposed esti- mator is strong consistent and asymptotic normal. The Monte Carlo simulation shows that the proposed estimation is robust for heavy tailed distributions and has good performance for VaR prediction. The proposed methodology for VaR prediction is also illustrated on CSI 300.