基于“已实现”波动率ARFI模型和CAViaR模型的VaR预测比较研究

VaR Forecast Comparison between Realized Volatility ARFI andCAViaR Models

高频金融数据在风险价值VaR度量和预测方面的价值已经引起了学术界和业界的广泛兴趣。计算和预测VaR的方法广义上可以分为两大类:间接法和直接法,在有了高频数据后,两类方法均可行,尤其是由于高频数据导出的"已实现"波动率的出现,使得间接法有明显改进。本文将从间接法中选取基于"已实现"波动率的ARFI模型与从直接法中选取的两个CAViaR模型进行比较,采用沪深300、上证指数、深证成指的5分钟高频数据,根据多种在评价VaR预测模型表现时广泛使用的后验测试,对各模型进行实证检验,结果表明基于CAViaR模型的预测表现优于基于"已实现"波动率的ARFI模型,这对风险管理从业者有一定的参考意义。

The models for calculating and forecasting VaR can be classified into two broad categories：indirect-VaR and direct-VaR approaches.The VaR forecast performances between models taken from these two approaches respectively are compared.One is ARFI-VaR forecast model the indirect-VaR approach based on ＂realized volatility＂ obtained by high frequency data.Another two are CAViaR-based models,which are the representatives of direct-VaR approach.By the various backtests that are extensively used for VaR performance evaluation,using 5-min high frequency data of CSI 300 Index,SSE Composite Index and SZSE Component Index,the empirical evidence shows the CAViaR-based models perform better than realized volatility-based ARFI model.