Robust ARMA Model
Robust Autoregressive Moving Average Model · Also known as: robust ARMA, outlier-robust ARMA, M-estimator ARMA, resistant ARMA estimation
The Robust ARMA model extends the classical Autoregressive Moving Average framework by replacing the sensitive least-squares loss with outlier-resistant estimation methods — typically M-estimators or median-based approaches. This protects coefficient estimates and forecasts from being distorted by additive outliers, level shifts, or innovational outliers that are common in economic and financial 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 Robust ARMA when you have a univariate, approximately stationary time series that you suspect contains outliers, level shifts, or isolated anomalous observations — situations common in macroeconomic series around recessions, financial crises, or data collection errors. It is particularly valuable when forecasting accuracy matters and you do not want a few extreme values to distort the model. Do NOT use it when your series is clearly non-stationary (difference or use ARIMA first), when the data are genuinely heavy-tailed by design (consider GARCH for volatility clustering instead), or when you have very few observations and cannot reliably detect or downweight anomalies.
Strengths & limitations
- Protects parameter estimates and forecasts from distortion caused by additive outliers and level shifts.
- Produces coefficients that represent the typical dynamics of the series rather than being driven by a handful of extreme values.
- Provides explicit information about which observations were flagged as anomalous through the robustness weights.
- Retains the interpretability and parsimony of the standard ARMA framework.
- Reduces forecast error when outliers are present but not structurally meaningful.
- More computationally demanding than classical ARMA; requires iterative algorithms that may be sensitive to starting values.
- Robust standard errors and tests are more complex and less universally implemented across software packages.
- If the series has genuine structural breaks rather than isolated outliers, robust weighting alone is insufficient — a structural break model is needed.
- Choosing the tuning constant for the loss function (e.g., Huber's k) affects the efficiency-robustness trade-off and requires judgment.
- Less suitable when the heavy tails are a true feature of the process (e.g., financial returns with volatility clustering) rather than data quality issues.
Frequently asked
How is Robust ARMA different from standard ARMA?
Standard ARMA uses least-squares or maximum likelihood estimation, which is highly sensitive to large residuals. Robust ARMA replaces that loss with a bounded function (e.g., Huber or Tukey bisquare) so that extreme observations are downweighted rather than fitted exactly. The model structure is identical; only the estimation criterion changes.
Do I still need to check for stationarity before fitting a Robust ARMA?
Yes. Robust ARMA assumes the series is stationary. If the series has a unit root, you should difference it first (producing an ARIMA framework) and then apply robust estimation to the differenced series. The presence of outliers can cause standard unit root tests to over-reject stationarity, so robust unit root tests may also be useful.
What loss functions are typically used in Robust ARMA?
The Huber loss function is the most common: it behaves like squared error for small residuals and like absolute error for large ones. The Tukey bisquare (biweight) function goes further by completely bounding the influence of very large residuals. The choice determines how aggressively outliers are downweighted.
When should I prefer a structural break model over a Robust ARMA?
If the anomalous behavior is a permanent shift in the level or trend of the series — rather than an isolated spike — a structural break model (e.g., Bai-Perron) is more appropriate. Robust ARMA handles isolated outliers well but does not explicitly model regime changes.
Is Robust ARMA available in standard econometric software?
Support varies. R packages such as RobustBase, robustarima, and tsrobprep implement robust time series estimation. Stata and EViews have limited built-in support; users often implement robust ARMA via custom iterative reweighted least squares routines.
Sources
How to cite this page
ScholarGate. (2026, June 3). Robust Autoregressive Moving Average Model. ScholarGate. https://scholargate.app/en/econometrics/robust-arma-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.
- ARIMA modelEconometrics↔ compare
- ARMA modelEconometrics↔ compare
- Robust AR modelEconometrics↔ compare
- Robust MA modelEconometrics↔ compare
- Robust OLSEconometrics↔ compare