Analysis of Conditional Distributions in Asymmetric MSGARCH Models for Volatility Forecasting: A Focus on Crude Oil Prices

Idowu, A. O. & Adesanya, C. O., Volume 6 Issue 1, July 2025 Pages 57-69, Published: 2025-05-05

Abstract

Volatility is an aspect of financial time series and a key factor in decision-making. This research uses the Markov Switching Generalized Autoregressive Conditional Heteroscedasticity (MS-GARCH) approach. The study aims to monitor the volatility conditions using asymmetric single regime switching dynamics, which deal with the non-linearity and heteroscedastic nature of the fluctuation in the price of crude oil. This research also explores the effects of conditional heteroscedasticity in evaluating crude oil price volatility. To analyze crude oil price, several distributions were compared based on maximum likelihood function used in estimating the model, two single regimes that monitor the volatility in the data on crude oil price, the GARCH model, generalized Error Distribution (GED), the student-t, and the Gaussian distribution, which were also used in this research. It is affirmed that, out of all the distributions compared, only GED exhibits the best fit in the analysis of the crude oil data set. The results show that among the compared distributions, Generalized Error Distribution exhibits the best fit, with a log-likelihood value of -1542.37, which performs better than Student-t (-1559.22) and Gaussian (-1574.89), respectively. Furthermore, the estimated GARCH parameters in the distribution signify volatility clustering and regime switching effects, enumerating the effectiveness of the model in capturing constant shifts in the prices of crude oil due to market volatility. The findings also reveal that the robustness of MS-GARCH approach in the price of crude oil provides an important insight for policymakers, financial analysts, and energy economists