Robust ARCH Model
Robust Autoregressive Conditional Heteroscedasticity Model · Also known as: robust ARCH, outlier-robust ARCH, heavy-tailed ARCH, robust conditional volatility model
The Robust ARCH model extends the classical Autoregressive Conditional Heteroscedasticity framework by replacing the standard maximum-likelihood estimator with robust alternatives that downweight or eliminate the influence of outliers. This makes volatility estimates resistant to extreme observations that frequently contaminate financial and macroeconomic time series.
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 the Robust ARCH model when your time series contains outliers, heavy tails, or irregular spikes that standard ARCH or GARCH estimators would absorb into the variance path and distort for many subsequent periods. It is particularly appropriate for daily or intraday financial returns, exchange rates, or commodity prices where data quality is imperfect or flash events occur. Avoid it when the series is clean and well-behaved, as the robust estimator sacrifices some efficiency relative to maximum likelihood under normality. It is also less warranted when the primary goal is comparison with a large existing literature that uses standard GARCH, since robust estimates may not be directly comparable.
Strengths & limitations
- Resistant to isolated outliers and heavy-tailed innovations that would distort standard ARCH variance estimates.
- Produces stable conditional volatility paths that reflect genuine market volatility rather than a single anomalous day.
- Retains the ARCH interpretive framework — coefficients and the persistence structure remain meaningful.
- Sandwich standard errors remain valid under departures from the assumed innovation distribution.
- Applicable to the same types of high-frequency financial and macroeconomic time series as classical ARCH.
- Less efficient than maximum-likelihood ARCH when the data are truly normally distributed and outlier-free.
- The choice of tuning constant c involves a bias-variance trade-off; there is no universally optimal value.
- Iterative robust estimation is computationally more demanding than closed-form or one-step ARCH fitting.
- Software support is narrower than for classical GARCH; implementation often requires specialised R packages or custom code.
- Inference results may not be directly comparable to a broader GARCH/ARCH literature that uses standard MLE.
Frequently asked
How does Robust ARCH differ from simply using a GARCH model with t-distributed errors?
A GARCH with t-distributed errors accommodates heavy-tailed innovations globally but still gives influential weight to every large observation in the score. Robust ARCH actively limits the contribution of any single outlier through a bounded score function, providing stronger protection against isolated spikes regardless of the assumed marginal distribution.
What is the tuning constant and how do I choose it?
The tuning constant c determines how sharply observations are downweighted. A common default in M-estimation is c = 1.345 (for a Huber estimator targeting 95% efficiency under normality). For financial returns with frequent outliers, a lower c offers more robustness at the cost of efficiency. Sensitivity analysis across a range of c values is recommended.
Can I test for ARCH effects using the robust model?
Yes. After fitting the robust mean equation, a robust Lagrange multiplier test on the squared robust residuals assesses whether conditional heteroscedasticity is present. The test is analogous to Engle's original ARCH-LM test but uses robust score contributions to avoid size distortions caused by outliers.
Is Robust ARCH available in standard statistical software?
Standard packages like R's rugarch or Python's arch do not natively implement robust ARCH, but the robustarima or robGARCH packages in R, and custom M-estimation wrappers, provide the functionality. Alternatively, iteratively reweighted ARCH can be programmed directly using the base optimisation routines.
When would you prefer Robust ARCH over just removing outliers and fitting standard ARCH?
Outlier removal is subjective and can induce its own biases, especially when it is unclear whether an extreme observation reflects genuine market behaviour or a data error. Robust ARCH lets all observations remain in the sample while automatically limiting the influence of extreme values, making results reproducible and less dependent on ad hoc deletion decisions.
Sources
- Engle, R. F. (1982). Autoregressive conditional heteroscedasticity with estimates of the variance of United Kingdom inflation. Econometrica, 50(4), 987–1007. DOI: 10.2307/1912773 ↗
- Iqbal, F. (2013). Robust estimation for the ARCH models. Revista Colombiana de Estadística, 36(1), 41–56. link ↗
How to cite this page
ScholarGate. (2026, June 3). Robust Autoregressive Conditional Heteroscedasticity Model. ScholarGate. https://scholargate.app/en/econometrics/robust-arch-model
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.
- ARCH modelEconometrics↔ compare
- EGARCH modelEconometrics↔ compare
- GARCH ModelEconometrics↔ compare
- Quantile RegressionEconometrics↔ compare
- Robust RegressionStatistics↔ compare
- Stochastic Volatility ModelFinance↔ compare