Volatility Modelling of Crude oil Prices Using a Five-States Figarch-Hidden Markov Model Frame Work
Aliyu Mansur
Ibrahim Lawal Kane
Abdulhameed Ado Osi
Huzaifa Abdurrahman
Hussaini Abubakar
Usman Abubakar
Abstract
Nigerian crude oil prices suffer from long memory, volatility, and varying volatility levels including extremely low, low, moderate, high, and extremely high. These characteristics directly increase market risk for the Nigerian economy. Studies model these levels into two levels low and high. However, the volatility levels can fall into other levels that include; extremely low, moderate, and extremely high. Therefore, this study aims to model these features using a newly developed hybrid time series model 5-States-FIGARCH-HMM. The data for this study was accessed documented record of central bank of Nigeria. The data were recorded monthly from 1990-2025 daily. The study employed ADF and KPSS tests to check for stationarity and non-unit roots was found. The study employed GPH test to test for presence of long memory in the time series and the long memory was found. Stationarity was achieved through fractional differencing. The study employed ACF and PACF plots to estimate the orders of AR and MA models respectively. The study found that 5-States-FIGARCH (1, 2)-HMM was the best model with least MAE, MSE, and RMSE when compared with FIGARCH and HMM models. The forecast results indicates rapid volatility stabilization and reduced tail risks, with a higher probability of low to moderate volatility regimes, implying more predictable pricing and reduced hedging requirements. Overall, the findings reveal strong regime persistence under calm market conditions, which may support macroeconomic stability toward early 2026.
References