波动率预测:多元GARCH模型能战胜一元GARCH模型吗?——二元DCC-GARCH模型与一元GARCH模型的PK

Volatility Forecasting:Can Multivariate GARCH Model Beat Unitary GARCH Model-PK of Binary DCC-GARCH Model and Unitary GARCH Model

波动率在不同资产(市场)之间的传导应当为预测波动率提供有用信息,使波动率预测更准确。本文考察了二元DCC-GARCH模型在预测波动率时是否优于传统的一元GARCH模型。研究发现:波动率在不同资产间的传导,可以为预测波动率提供有用信息,有助于提高预测的准确性;波动率溢出越明显,对提高预测帮助越大;日数据传递信息更及时,在预测波动率时能比周数据提供更有用的信息;实际波动率的不同度量方式会影响波动率预测准确性的评判结果,但一个占优的多元GARCH模型预测的波动率,可以在所有度量方式、评判标准下都占优。本文为进一步研究多元GARCH模型在预测波动率方面的应用提供了有益的启示。

The transmission of volatility between different assets(markets)should provide useful information for forecasting volatility and make it more accurate.This paper investigates whether the binary DCC-GARCH model is superior to the traditional unitary GARCH model in forecasting volatility.The results show that:the transmission of volatility among different assets can provide useful information for forecasting volatility and help to improve the accuracy of forecasting;the more obvious volatility spillover,the more helpful to improve forecasting;the daily data can transmit information more timely and provide more useful information than weekly data in forecasting volatility;different measurement methods of actual volatility will affect volatility forecasting result,but the volatility predicted by a dominant multivariate GARCH model can be dominant in all measurement methods and evaluation criteria.This paper provides useful enlightenment for further research on the application of multivariate GARCH model in forecasting volatility.