A0221
Title: High-dimensional multivariate realized stochastic volatility model using characteristic factor regression
Authors: Tsunehiro Ishihara - Takasaki City University of Economics (Japan) [presenting]
Abstract: In financial econometrics, multivariate return series modeling has been widely studied. Multivariate stochastic volatility models are popular and exhibit high forecasting performance. However, estimating and forecasting with multivariate stochastic volatility models is time-consuming, especially for high dimensions. A multivariate stochastic volatility model with observed characteristic factors is proposed. Intraday information is also introduced via realized covariance. In the proposed model, high-dimensional multivariate returns are conditionally independently modeled by a univariate time-varying coefficient regression model with stochastic volatility errors. Estimation and forecasting can be performed separately for each series. Thus, computation time increases proportionally to the dimension of the returns. In addition, estimation and forecasting can be performed in parallel in each univariate model. An empirical illustrative example is presented using sector index series and market-, size-, and value-based factors, as well as their realized covariances.
