Structural Break ARIMA Model
Structural Break Autoregressive Integrated Moving Average Model · Also known as: ARIMA with structural breaks, break-adjusted ARIMA, piecewise ARIMA, ARIMA with regime shifts
A structural break ARIMA model extends the standard ARIMA framework by explicitly identifying and accommodating one or more abrupt shifts in the level, trend, or dynamics of a time series. Rather than forcing a single set of ARIMA parameters across the entire sample, it fits separate ARIMA specifications for each regime defined by the detected break dates.
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 ARIMA model when a time series shows visible level shifts, trend changes, or abrupt changes in volatility that cannot be captured by a single ARIMA specification across the full sample. It is especially appropriate for macroeconomic series (GDP, inflation, unemployment) spanning policy regime changes, financial series crossing crisis periods, or any setting where ignoring breaks would bias unit-root tests and yield misleading forecasts. Do not use it when the series is short (fewer than ~60 observations), because breaks consume degrees of freedom rapidly and minimum-segment constraints become binding. Also avoid it when apparent breaks are actually smooth transitions — a smooth-transition AR or Markov-switching model is more appropriate in that case.
Strengths & limitations
- Correctly identifies and isolates regime-specific dynamics, preventing one period from distorting estimates in another.
- Improves forecast accuracy when the future regime resembles the most recent sub-period rather than the full historical average.
- Avoids spurious unit-root findings caused by ignored breaks, leading to correctly specified differencing orders.
- The Bai-Perron framework provides statistically rigorous, data-driven break-date estimation rather than ad hoc choices.
- Compatible with standard ARIMA inference tools; each segment can be diagnosed and reported separately.
- Requires a sufficiently long series; each regime must contain enough observations for reliable ARIMA estimation.
- Proliferation of parameters: detecting k breaks introduces at least (k+1) × (p+d+q) parameters, inflating model complexity.
- Break-date uncertainty is not always propagated into confidence intervals for forecasts, potentially understating forecast uncertainty.
- Abrupt break assumption may be too rigid if the true shift is gradual or if multiple overlapping mechanisms drive change.
Frequently asked
How many breaks should I allow?
Start with a maximum of 5 breaks and let the Bai-Perron sequential test or BIC determine the optimal number. In practice, economic series rarely exhibit more than 2-3 genuine breaks in a 30-50 year span.
Should I test for breaks before or after differencing?
It depends: test for breaks in levels if you suspect the break affects the trend, and in first differences if you suspect it affects the growth rate. Perron (1989) tests are specifically designed for break detection in the presence of a unit root.
What is the minimum segment length?
A common rule of thumb is at least 15% of the total sample (or a minimum of 20-30 observations) per segment to ensure stable ARIMA estimation. The Bai-Perron framework allows the analyst to set this trimming parameter explicitly.
How does this differ from a Markov-switching model?
A structural break ARIMA treats breaks as one-time permanent shifts with fixed dates. A Markov-switching model allows the series to move back and forth between regimes with estimated transition probabilities. Use structural breaks for permanent changes and Markov switching for recurring cycle-like regime alternations.
Do I need to re-test for unit roots in each segment?
Yes — the integration order can differ across regimes. A series that is I(1) overall may be stationary within a shorter sub-period, so apply ADF or KPSS tests separately in each segment to choose the correct differencing order.
Sources
- Bai, J., & Perron, P. (1998). Estimating and testing linear models with multiple structural changes. Econometrica, 66(1), 47-78. DOI: 10.2307/2998540 ↗
- Perron, P. (1989). The great crash, the oil price shock, and the unit root hypothesis. Econometrica, 57(6), 1361-1401. DOI: 10.2307/1913712 ↗
How to cite this page
ScholarGate. (2026, June 3). Structural Break Autoregressive Integrated Moving Average Model. ScholarGate. https://scholargate.app/en/econometrics/structural-break-arima-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
- Bai-Perron TestEconometrics↔ compare
- Chow TestEconometrics↔ compare