Scalable inference for a full multivariate stochastic volatility model

Abstract

We introduce a multivariate stochastic volatility model that imposes no restrictions on the structure of the volatility matrix and treats all its elements as functions of latent stochastic processes. Inference is achieved via a carefully designed feasible and scalable MCMC that has quadratic, rather than cubic, computational complexity for evaluating the multivariate normal densities required. We illustrate how our model can be applied on macroeconomic applications through a stochastic volatility VAR model, comparing it to competing approaches in the literature. We also demonstrate how our approach can be applied to a large dataset containing 571 stock daily returns of Euro STOXX index.