MODELING THE CO-VOLATILITY OF GOLD AND CRUDE OIL PRICE IN ETHIOPIA USING BAYISIAN MULTIVARIATE GARCH MODELS

Abstract

Modelling and forecasting volatility dynamics of asset returns has been an important application for asset management, portfolio analysis and risk assessment due to the simple fact that volatility has informational content and contains signals of the market information flow. The Objective of this study is modeling the co-volatility of gold and oil price using Bayesian multivariate generalized autoregressive conditional heterosedasticity (GARCH) models using the data extracted from national bank of Ethiopia for the period from 1998 to 2018. Using Bayesian method on the model parameters and prediction of future volatilities are addressed by using Markov chain Monte Carlo simulation. The visual inspection of the marginal trace plots, density estimates and autocorrelation plots along with formal tests showed good convergence of the Markov chains. The 95% credible intervals for the skewness parameter do not include 1 for oil and hence confirms asymmetry in series, whereas estimates for skewness for gold series is insignificant and contradicting the assumption of leverage effects. The estimate of the tail parameter indicates the appropriateness of heavy tail of the distribution of the returns. The estimated parameters 𝐴𝑅𝐶𝐻 (𝛼) and 𝐺𝐴𝑅𝐶𝐻 (𝛽) coefficients were significant at 95% credible intervals for each commodity returns. The CCC model hypothesis (𝛼 = 𝛽 = 0) is rejected on the basis of the marginal posterior distributions of the DCC conditional correlation parameters 𝛼 𝑎𝑛𝑑 𝛽. Therefore, it might be inaccurate or misleading to assume constant correlations (neutrality) between gold and oil price return pairs. The result of this study indicates that the dynamic conditional correlation model within skew t distributed is the most accurate for prediction of future volatility in the context of our empirical setting. The study recommends and encourages future researchers studying the forecasting performance of Bayesian multivariate GARCH models to pay particular attention to the measurement of realized volatility of this two commodity with other stock markets