A1455
Title: Matrix inequality constraints for vector GARCH models with spillovers: A new (mixture) formulation
Authors: Menelaos Karanasos - Brunel University (United Kingdom) [presenting]
Yongdeng Xu - Cardiff University (United Kingdom)
Alexandros Paraskevopoulos - University of Patras (Greece)
Starvoula Yfanti - Queen Mary University of London (United Kingdom)
Abstract: The purpose is to review and generalize results on the derivation of tractable non-negativity (necessary and sufficient) conditions for $N$-dimensional asymmetric power GARCH models. In practice, these constraints may not be fulfilled. To handle these cases we propose a new mixture formulation in order to eliminate some of these constraints. By using the exponential specification for some (but not all) of the conditional variables in the system we considerably reduce their dimensions. We also obtain new theoretical results about the second-moment structure and the optimal forecasts of such multivariate processes. An empirical example is included to show the effectiveness of the proposed method. We study the correlation dynamics among nineteen sectoral corporate bond indices. The time-varying cross-sector nexus is found to depend on economic fundamentals. The analysis further shows that the bond sectoral interconnectedness is highly vulnerable to crisis episodes, jeopardising the stability of the financial system.
