Is Gold Always a Safe-Haven? Evidence from a Novel Markov-Switching Multivariate GARCH Model with Copula-Distributed Innovations
59 Pages Posted: 18 Nov 2021 Last revised: 29 Aug 2022
Date Written: August 12, 2022
We suggest a new Markov switching approach to multivariate volatility modeling for improving the dynamic assessment of financial market interdependencies. The approach takes advantage of the flexible copula multivariate GARCH (C-MGARCH) model of Lee and Long (2009), and allows state-specific higher-order dependence structures. We show some asymptotic features of the new model and motivate a two-step maximum likelihood estimator. As an empirical illustration, we consider a bivariate series of returns of a high-yield equity index (MSCI World Developed Markets) and a traditional safe-haven asset (gold), and investigate how higher-order dependencies in copula-distributed model innovations show up in observable return processes. The Markov switching generalization of the C-MGARCH model captures structural changes of heterogeneous non-linear dependence patterns among financial returns (conditional independence, flight-to-safety, crash, risk-seeking, boom), and helps to uncover flight-to-safety (i.e. safe-haven) effects. However, as a further result, we find that gold is not generally a safe-haven under changing financial market conditions. Out-of-sample density evaluations underline the merits of the suggested Markov switching model in comparison with time-invariant benchmarks (MGARCH, C-MGARCH).
Keywords: Copula, multivariate GARCH models, Markov switching, safe-haven, gold, stock market
JEL Classification: C14, C32, C51, C58, G11
Suggested Citation: Suggested Citation