Bayesian Variable Selection of Stochastic Volatility Models in Financial Time Series

• Activity Number: 256 - Contributed Poster Presentations: Section on Bayesian Statistical Science
• Type: Contributed
• Date/Time: Monday, July 30, 2018 : 2:00 PM to 3:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Feng Chi Liu*
• Keywords: stochastic volatility model; stochastic search variable selection method; Markov chain Monte Carlo sampling; Bayesian variable selection

Abstract

Volatility modeling of financial time series has been widely investigated in financial econometrics. In empirical applications, most of volatility models are formed as a simple model, such as a GARCH(1,1) model or a stochastic volatility (SV) model with an AR(1) process. In this study, a general order of AR(p) process of SV model is considered to allow both fast and slow decay of volatility information to be obtained in the SV model. The order of AR(p) process of SV model is determined by Bayesian variable selection approach. Motivated by the stochastic search variable selection method of George & McCulloch (1993), a mixture of normal priors for the SV model parameters is proposed in this study. The estimation of model parameters and the order selection of AR(p) process are simultaneously conducted by the proposed Markov chain Monte Carlo (MCMC) sampling scheme. Simulation studies demonstrated that the proposed method can accurately select the order of SV model and estimate unknown parameters. In real data application, the results of order selection of SV model have provided new insight of volatility modeling.