已实现波动GAS-HEAVY模型及其实证研究

引入日内高频数据计算的已实现波动,能够提高波动模型预测能力。本文将日收益和已实现波动联合建模,提出一种新的波动模型。选取尺度调整t分布和F分布作为日收益和已实现波动的分布,更为充分和灵活地捕捉厚尾性,采用得分驱动方法设定波动模型更新项,得出广义自回归得分(GAS)波动模型,提高对实际模型的逼近效率。本文对模型遍历性和平稳性进行证明,并与同类模型进行比较。蒙特卡罗模拟实验显示,在数据生成过程误设的情况下本文提出的GAS-HEAVY模型比同类模型具有更好的数据拟合效果。基于沪综指、深成指和沪深300指数2013.1至2017.4日内1分钟高频数据实证分析表明,不同损失函数的SPA检验下GAS-HEAVY模型的波动预测能力显著强于其它同类模型。本文给出的GAS-HEAVY模型为有关理论研究和市场应用提供了新的波动计量工具。

GAS-HEAVY Model for Realized Measures of Volatility and Returns

A new volatility model named GAS-HEAVY is introduced to model returns and realized measures of volatility jointly. The key features are fat-tailed distributions for returns and realized volatilities and autoregressive score-driven (GAS) model for dynamics of the latent volatility. By assuming a rescaled t distribution for daily returns and a rescaled F distributionfor realized volatility measures, the score dynamics for the latent volatility are robust to outliers and incidental large observations. Parameter restrictions are formulated to prove ergodicity and stationarity.Our simulation study shows that the new model fit data better than other alternatives. An empirical application of our model is provided to daily returns and 1-minute intraday high-frequency prices of Shanghai Composite Index,Shenzhen Component Index and CSI300 Index. The empirical evidences justify the superior volatility predictive ability of the new model in our paper.