Beyond Normality: OGELAD Error Distribution in Energy Prices Volatility Forecasting

Abstract

luctuations in energy prices such as crude oil price poses major challenges for managing risk, predicting financial trends and shaping energy policies. Notwithstanding, traditional models for predicting volatility often assume a normal distribution, which does not account for the fat tails and asymmetry often seen in energy markets. While GARCH-family models help capture volatility clustering, their accuracy depends heavily on the choice of error distribution used – a key factor that has not been thoroughly studied in energy finance. In this study, a newly developed and three existing non-normal distributions across three GARCH models – EGARCH (1,1), TGARCH (1,1), and GJR-GARCH (1,1) are evaluated to see how well they perform with crude oil returns. All the fitted models yield parameters that are statistically significant (p < 0.05), with residual analysis confirming the elimination of autoregressive conditional heteroscedasticity. Among the different error distributions, the OGELAD distribution stands out, with the highest log-likelihood value and lowest values across key information criteria (AIC, BIC, HQIC) in every model specification. In out-of-sample forecasting over a 30-day horizon, the GJR-GARCH (1,1) model using OGELAD-distributed innovations proves to be the most accurate, outperforming alternatives. These findings support the use of the OGELAD distribution in energy market volatility modelling, as it enhances risk assessment and improves derivative pricing accuracy.

Keywords: