我国上市商业银行股票日收益风险价值研究——基于AR、HAR和MIDAS模型的分析

Value at Risk Analysis of Daily Returns of Listed Commercial Banks in China——Based on the AR,HAR and MIDAS Models

本文选取了上市商业银行中的四家大型商业银行和四家股份制银行作为研究对象,基于2011年到2012年的5分钟高频数据进行分析,运用了AR、HAR和MIDAS模型预测股票日收益风险价值VaR,并以均方根误差(RMSE)、标记损失比较和正态分布为基准的三种方法,结合递归方案和固定方案进行预测效果评估。结果显示,在预测银行日收益风险方面HAR和MIDAS预测效果要优于AR模型。而就前两者而言,HAR对波动率较大的股份制银行预测效果要好于MIDAS模型,MIDAS模型对波动率较小的四家大型商业银行的预测效果要好于HAR模型。

This article analyzes the 5 minutes high-frequency data of four major state-owned banks and four joint-stock banks between 2011 and 2012 using AR, HAR and MIDAS models to forecast their daily Value at Risk （VaR）. Then it assesses the forecast by using Root-mean-square error（ RMSE）, tick loss comparison and normal distribution as benchmarks and combines them with fixed scheme and recursive scheme. The results show that the accuracy of HAR and MIDAS models are better than AR model. The accuracy of HAR model for joint-stock banks with larger volatility is better than MIDAS model, and MIDAS model predict better than HAR model for four major state-owned banks with smaller volatility.