Modelling the Covariance Dynamics of Multivariate Financial Time Series

Over the last three decades, the notion of the covariance matrix of asset returns has pervaded almost every financial aspect, and covariance forecasts are a central input to various financial applications such as asset pricing, portfolio selection, and risk management. As such, this thesis aims at establishing and verifying new methods to model and forecast the covariance matrices, thus allowing for better analyses of financial problems and improved decision-making. The thesis consists of three studies.