A discussion on the robustness of conditional heteroskedasticity models: Simulation evidence and applications of the crude oil returns

This paper discusses the inherent robustness of the generalized autoregressive score (GAS) model, such that consistent estimators can still be obtained when large outliers do exist. A recent example includes the historical-writing negative price of the West Texas Intermediate (WTI) crude oil future in early-2020. Via simulation studies, we demonstrate that GAS can produce much more robust estimates than the popular GARCH model. Empirical analysis using the WTI returns over 2017-2020 further supports the superiority of GAS over GARCH models.

Automated summary

This paper discusses the inherent robustness of the generalized autoregressive score (GAS) model, demonstrating that consistent estimators can still be obtained even in the presence of large outliers. Simulation studies and empirical analysis confirm the superiority of GAS over GARCH models.