Regime Switching for Dynamic EquiCorrelation
Zakaria Moussa  1, *@  , Yannick Le Pen  2@  , Arthur Thomas  2@  
1 : Laboratoire d'économie et de management de Nantes Atlantique  (LEMNA)
Nantes Université - Institut d'Administration des Entreprises - Nantes
Chemin de la Censive du Tertre - Bâtiment Erdre - BP 52231 - 44322 NANTES Cedex 3 - FRANCE -  France
2 : Université Paris Dauphine-PSL
Université Paris sciences et lettres
Place du Maréchal de Lattre de Tassigny75775 PARIS Cedex 16 -  France
* : Corresponding author

We add a regime switching dynamic, in the spirit of Pelletier (2006) markov switching conditional correlation model, to Engle and Kelly (2012) Dynamic Equicorrelation model. The DECO model decomposes the correlation matrix into blocks for which conditional correlation are equal. Correlations between blocks can be common or block-specific. Our Regime Switching Dynamic EquiCorrelation (hence RSDECO) model is therefore adequate for fitting large conditional correlations in presence of level shifts. Extensive simulations show that the RSDECO is able to reproduce the true level of correlations for a large number of variables. The estimation of daily correlations between commodity, stock and bond returns from 01/04/2000 to 12/29/2022 show significant changes in the pattern of correlations over time. The first major and significant regime change occurs around September 15, 2008. The dynamics of the correlations between the three asset classes will have become unstable after the year 2020.

2.15.0.0
2.15.0.0
2.15.0.0

Online user: 5 Privacy
Loading...