EcoSta 2017: Start Registration
View Submission - EcoSta2017
A0368
Title: Leverage, asymmetry and heavy tails in high-dimensional factor stochastic volatility models Authors:  Mengheng Li - Vrije Universiteit Amsterdam (Netherlands)
Marcel Scharth - The University of Sydney Business School (Australia) [presenting]
Abstract: It is known that time-varying volatility and leverage effects are often observed in financial time series which is believed to be asymmetrically distributed with heavy tails. The rich literature studying various forms of univariate stochastic volatility model tends to confirm such empirical findings. Yet the literature focusing on high dimensional stochastic volatility models lacks a corresponding general modelling framework and efficient estimation method due to curse of dimensionality. The aim is to propose a flexible factor stochastic volatility model with leverage based on generalized hyperbolic skew Student's t-error to model asymmetry and heavy tails. With shrinkage, the model leads to different parsimonious forms, and thus is able to disengage leverage effects and skewness in idiosyncratic noise from those in factors. A highly efficient Markov chain Monte Carlo estimation procedure which uses efficient importance sampling to exploit the Gaussian mixture representation of the error distribution is proposed to analyze the univariate version of the model. Multivariate extension is achieved with marginalization of factors and boils down to many univariate series which can be estimated in parallel. We assess the performance of our proposed method via a Monte Carlo study with both univariate and multivariate simulated data. Finally we apply our model to a equally weighted portfolio consisting of stocks from S\&P/ASX50.