Title: Modelling latent network stochastic volatility spillovers
Authors: Laurent Pauwels - University of Sydney (Australia) [presenting]
Michael McAleer - ()
Manabu Asai - Soka University (Japan)
Abstract: Volatility spillovers and linkages of financial portfolios are modelled by using novel latent network stochastic volatility (NetSV) models that capture the latent linkages across the financial assets. Financial theory points to latent information linkages as the origin of volatility spillovers. These linkages are created from common information, which impacts the expectations of financial traders, and also information spillovers that arise from their hedging behaviour. The theory is extended to network and stochastic volatility models, which are assumed to be random and latent. The networks provide an interpretable and tractable model of the volatility linkages. New Bayesian algorithms are developed to identify latent random networks within the context of multivariate stochastic volatility models. Upon identifying the network, the volatility spillover effects across markets and dynamic optimal hedge ratios can be estimated and tested statistically. The latent network approach reduces the estimation burden of the spillover effects that are typically encountered in multivariate volatility models with high-dimensional parameters.