Structural Break MA Model
Moving Average Model with Structural Breaks · Also known as: MA model with structural change, broken MA model, MA with regime shift, structural break moving average
A Moving Average (MA) time series model augmented to accommodate one or more structural breaks — abrupt shifts in the mean, variance, or MA coefficients occurring at known or unknown break dates. Ignoring structural breaks in an MA process inflates forecast errors and distorts inference on the error dynamics.
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 a structural break MA model when a univariate time series shows evidence of an abrupt level shift or a change in short-run persistence — for example around a financial crisis, a policy reform, or a natural disaster — and when the primary goal is modelling the error (shock) dynamics rather than a multivariate causal system. It is appropriate when standard MA diagnostics (Ljung-Box, ARMA information criteria) improve substantially after allowing a break. Do not use it when the suspected change is gradual (prefer smooth-transition MA or time-varying parameter MA), when multiple interacting variables are central (prefer structural break VAR), or when you have fewer than roughly 50 pre- and post-break observations, since MA parameter estimates are unstable in very short segments.
Strengths & limitations
- Captures abrupt regime shifts in short-run shock dynamics that a homogeneous MA model misses.
- Reduces forecast error variance by preventing pre-break noise from contaminating post-break estimates.
- Compatible with standard MA estimation machinery once break dummies are added as regressors.
- Allows formal testing of whether the structural change is statistically significant.
- Endogenous break-date search (Zivot-Andrews, Bai-Perron) removes the need to pre-specify the break.
- Requires sufficient observations in each regime (roughly 50+) for reliable MA parameter estimation.
- Endogenous break-date search inflates size in finite samples; critical values must be corrected (Andrews 1993).
- The model is restricted to abrupt, one-time (or small-number) breaks; gradual structural change is better handled by smooth-transition or time-varying-parameter models.
- Misspecifying the order q or the number of breaks leads to biased residuals and invalid inference.
- Extending to MA models with unknown multiple breaks substantially increases computational burden.
Frequently asked
How do I choose the MA order q in the presence of a structural break?
Estimate a standard MA(q) on each regime separately and use AIC or BIC within each segment to select q. Alternatively, fit the full augmented model over a range of q values and select by information criterion; the break dummy absorbs the level shift so the AIC reflects only the error-dynamic fit.
Can I have more than one structural break?
Yes. The Bai-Perron (1998) sequential test is the standard approach for detecting multiple breaks at unknown dates. Each additional break adds a regime, reducing the observations available for estimation, so parsimony is important.
What is the difference between an additive outlier and a structural break in an MA model?
An additive outlier is a single time-period spike that does not propagate through the series; it is modelled with a one-period impulse dummy. A structural break shifts the mean or MA coefficients permanently from the break date onward; it is modelled with a level-shift or slope-change dummy. The two require different corrections and have different effects on inference.
Does a structural break affect unit-root testing of an MA process?
Yes. Ignoring a break in an MA(1) or higher-order MA process can bias the autocorrelation function and distort stationarity tests. Perron (1989) showed that unmodelled breaks make standard ADF tests incorrectly suggest a unit root. Always test for breaks before concluding about integration order.
What software can estimate a structural break MA model?
R packages strucchange and mbreaks implement Bai-Perron tests and allow break-augmented ARMA estimation. Stata's sbsingle and sbreak commands cover single and multiple break cases. EViews provides built-in break testing integrated with ARMA estimation dialogs.
Sources
- Perron, P. (1989). The great crash, the oil price shock, and the unit root hypothesis. Econometrica, 57(6), 1361–1401. DOI: 10.2307/1913712 ↗
- Zivot, E., & Andrews, D. W. K. (1992). Further evidence on the great crash, the oil-price shock, and the unit-root hypothesis. Journal of Business & Economic Statistics, 10(3), 251–270. DOI: 10.1080/07350015.1992.10509904 ↗
How to cite this page
ScholarGate. (2026, June 3). Moving Average Model with Structural Breaks. ScholarGate. https://scholargate.app/en/econometrics/structural-break-ma-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
- Moving Average ModelEconometrics↔ compare
- Structural Break AR ModelEconometrics↔ compare
- Structural Break ARIMA ModelEconometrics↔ compare
- Zivot-Andrews Structural Break TestEconometrics↔ compare