A1292
Title: Regime switching for dynamic equicorrelation
Authors: Yannick Le Pen - Universite Paris Dauphine (France) [presenting]
Arthur Thomas - Paris Dauphine University - PSL (France)
Zakaria Moussa - Universite de Nantes (France)
Abstract: The purpose is to introduce a regime-switching dynamic, inspired by Pelletier's Markov switching conditional correlation model, into Engle and Kelly's dynamic equicorrelation (DECO) model. The DECO model is suitable for a vast array of correlations, segmenting the correlation matrix into blocks with equal conditional correlations, while inter-block correlations can be unique or block-specific. Pelletier's regime-switching model, which builds on Engle's model, is limited to a smaller set of returns. The regime-switching dynamic equicorrelation (RSDECO) model is thus well-suited for modeling large conditional correlations with regime shifts. Through extensive simulations, RSDECO is demonstrated to accurately capture the true correlation levels across numerous variables. The model is exemplified by estimating daily correlations among commodity, stock, and bond returns from April 1, 2000, to December 29, 2022. Significant shifts are identified in correlation patterns, notably around September 15, 2008, and substantial alterations are observed in the correlations among the three asset classes post-2020.
