CMStatistics 2017: Start Registration
View Submission - CFE
Title: Factor multivariate realized stochastic volatility model Authors:  Yuta Yamauchi - University of Tokyo (Japan) [presenting]
Yasuhiro Omori - University of Tokyo (Japan)
Abstract: Although modelling time-varying volatility and correlations of multivariate asset returns is one of most important problems in the financial risk management, it has been difficult to obtain stable inference of covariance of asset returns due to the high dimensionality of parameters in dynamic covariance structure. One major solution to reduce the number of parameters is to introduce factor structure assuming that a small number of common factors describe the dynamics of time-varying covariance matrices as discussed in the factor stochastic volatility models. We propose parsimonious modelling of multivariate asset returns based on dynamic factor stochastic volatility models with leverage effect, which allow dynamic latent factors. Firstly, to stabilize the estimation and prediction of time-varying parameters, we incorporate additional observations based on intraday asset returns and market indices. We use realized measures for covariance of asset returns based on intraday asset returns such as realized covariances, and latent factors of asset returns such as market indices. Secondly, we reduce the number of parameters of leverage effect, omitting the leverage effect between each asset and each volatility of asset. We only introduce leverage effect between each latent factor and each volatility of asset.