The Use of GARCH Autoregressive Models in Estimating and Forecasting the Crude Oil Volatility

Today, oil is one of the most popular commodities traded globally, due to its indispensable character and multiple properties offered to mankind. Increased attention is paid to the analysis of volatile and fluctuating trends in the overall price of this valuable energy source. Using the autoregressive conditional heteroskedasticity models such as GARCH(1,1), GARCH-M(1,1) and EGARCH(1,1), the present study has as a priority objective in estimating and predicting the volatility of the oil returns series (Brent Crude Oil return series) in the 1987-2022. The main results highlighted the preference in using the asymmetric model EGARCH (1,1) on the measurement of conditional variance, showing that Brent Crude Oil reacts over 90% to any existing market’s shock (i.e.: information, events, facts, news, etc.) in a negative manner/way. At the same time, various tests and evaluation conditions were used (ARCH-LM Test, Durbin-Waston Test, High Log likelihood, Lowest Schwarz Information Criteria) in investigating the level of performance in estimation the conditional crude oil volatility. Each GARCH (1,1) model is meeting brilliantly these conditions and acquiring the character of stability and validity in use. At the same time, performing forecast analysis on crude oil volatility in two different time periods: 1987-2022, respectively 2020-2022, it was shown that existence of the phenomenon of clustering-volatility over the time, with strong implications for the functioning mechanism of international financial markets. Fulfilling those restrictive conditions, the symmetric and parametric model GARCH-M (1,1) becomes, in our case, the most efficient model in forecasting the volatility of Brent Crude Oil return series in the analysed period.