Robust EGARCH Model
Robust Exponential Generalized Autoregressive Conditional Heteroscedasticity Model · Also known as: Robust EGARCH model, outlier-robust EGARCH, robust exponential GARCH, REGARCH
Robust EGARCH extends Nelson's (1991) Exponential GARCH model by replacing standard quasi-maximum likelihood estimation with outlier-resistant procedures — typically bounded-influence or M-estimation — so that a small fraction of extreme observations or data errors cannot distort the estimated volatility dynamics or the leverage effect.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use Robust EGARCH when modelling financial return volatility in series known to contain outliers, data errors, or structural jumps that are not the main focus of analysis — for example daily equity returns around crisis events, commodity prices with recording errors, or cryptocurrency returns with exchange glitches. It is especially appropriate when the leverage effect is economically important and standard EGARCH estimates appear unstable across subsamples. Do not use Robust EGARCH if every large shock is genuinely informative (e.g., scheduled macro announcements), because downweighting real tail events will underestimate true tail risk. Also avoid it when the sample is too small (fewer than ~300 observations) to identify both the asymmetry and the robust tuning constant reliably.
Strengths & limitations
- Inherits EGARCH's log-variance specification, ensuring positive conditional variance without inequality constraints on parameters.
- Captures leverage asymmetry (negative returns raising volatility more than positive returns of equal magnitude) robustly.
- Limits the influence of isolated outliers or data errors on all subsequent volatility forecasts through the persistence chain.
- Allows detection of genuine volatility clusters while preventing a single bad data point from inflating persistence estimates.
- Estimated parameters are consistent and asymptotically normal under mild contamination, unlike standard QMLE under heavy contamination.
- Requires choosing a tuning constant for the influence function; wrong choice either leaves outliers unaddressed or over-downweights legitimate tail events.
- Iterative robust estimation is computationally heavier than standard QMLE and can face convergence difficulties in high-order EGARCH specifications.
- Theory for robust confidence intervals and hypothesis tests in EGARCH is less developed than for GARCH, complicating formal inference.
- Not suitable when outliers are structural breaks — a Markov-switching or structural-break GARCH is more appropriate in that case.
- Short samples (fewer than ~300 observations) may not support reliable joint identification of leverage, persistence, and the robust tuning constant.
Frequently asked
How does Robust EGARCH differ from standard EGARCH?
Standard EGARCH maximises the Gaussian or Student-t quasi-log-likelihood, giving each observation equal influence. Robust EGARCH replaces or augments this with a bounded-influence objective so that extreme standardised residuals are downweighted, preventing outliers from distorting the estimated leverage coefficient and persistence parameters.
How do I choose the tuning constant for the influence function?
The tuning constant controls the trade-off between robustness and efficiency. A common approach is to set it so that the estimator achieves 95% asymptotic efficiency under the assumed clean-data distribution (e.g., normal or Student-t). Simulation studies calibrated to the expected contamination rate of your series provide a more principled choice.
Can I use information criteria (AIC/BIC) to compare Robust EGARCH with standard EGARCH?
Standard AIC/BIC formulas based on the log-likelihood are not directly comparable between robust and non-robust estimators because the objectives differ. Use simulation-based model comparison or predictive accuracy measures (e.g., out-of-sample RMSE of volatility forecasts) instead.
When should I prefer Robust TGARCH over Robust EGARCH?
Choose Robust TGARCH (GJR-GARCH) when you prefer a model in variance (rather than log-variance) space and are comfortable imposing non-negativity constraints. Choose Robust EGARCH when you need the leverage effect to be modelled in log-variance, avoiding parameter constraints, especially for series with very persistent volatility.
Does robust estimation eliminate the need for outlier pre-processing?
Not entirely. Robust EGARCH reduces sensitivity to isolated outliers but is not designed for large clusters of outliers or regime changes. If the series contains a known contamination period, pre-screening with outlier detection or using a Markov-switching model alongside the robust estimator may be more appropriate.
Sources
- Muler, N., & Yohai, V. J. (2008). Robust estimates for GARCH models. Journal of Statistical Planning and Inference, 138(10), 2918–2940. DOI: 10.1016/j.jspi.2007.11.003 ↗
- Nelson, D. B. (1991). Conditional heteroskedasticity in asset returns: A new approach. Econometrica, 59(2), 347–370. DOI: 10.2307/2938260 ↗
How to cite this page
ScholarGate. (2026, June 3). Robust Exponential Generalized Autoregressive Conditional Heteroscedasticity Model. ScholarGate. https://scholargate.app/en/econometrics/robust-egarch
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- DCC-GARCH modelEconometrics↔ compare
- EGARCH modelEconometrics↔ compare
- GARCH ModelEconometrics↔ compare
- Robust GARCH modelEconometrics↔ compare
- Robust TGARCHEconometrics↔ compare
- TGARCH modelEconometrics↔ compare