Structural Break AR Model
Autoregressive Model with Structural Breaks · Also known as: AR model with structural change, breakpoint AR model, piecewise autoregressive model, AR model with regime shifts
The structural break AR model extends the standard autoregressive framework by allowing the intercept and autoregressive coefficients to shift at one or more unknown break dates. Each regime between consecutive break points is governed by its own AR parameters, capturing abrupt changes in the dynamics of a time series caused by crises, policy shifts, or other shocks.
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 structural break AR model when economic or empirical reasoning suggests that the dynamics of a univariate time series changed at one or more points — for example after a financial crisis, a monetary policy regime change, or a major institutional reform. It is appropriate when a standard AR model leaves systematic patterns in residuals around a suspected break date, or when unit-root tests that allow for a break (Zivot-Andrews) reveal a significant breakpoint. Do not use it when no substantive reason for a break exists, when the sample is very short (fewer than roughly 60-80 observations, since each segment needs enough data to estimate AR coefficients), or when the breaks are gradual rather than abrupt (smooth-transition AR models are better suited for gradual change).
Strengths & limitations
- Captures abrupt, permanent shifts in time-series dynamics that a constant-parameter AR model would miss.
- Allows the number and location of breaks to be unknown and estimated from the data using well-established procedures (Bai-Perron).
- Each regime is estimated by simple OLS, keeping the method computationally transparent and interpretable.
- Segment-specific coefficient tests reveal whether intercepts, persistence, or both changed at the break.
- Compatible with standard AR model selection criteria (AIC/BIC) applied within each regime.
- Requires a minimum segment length (often 10-15% of the sample) to identify breaks reliably; short series or many breaks exhaust this requirement quickly.
- Assumes breaks are instantaneous; gradual regime changes are poorly captured and may inflate the estimated number of breaks.
- Sequential break tests lose power when errors are non-normal or heteroscedastic unless HAC corrections are applied.
- Break-date estimates carry uncertainty; treating estimated break dates as known in subsequent inference understates standard errors.
Frequently asked
How do I choose the number of structural breaks?
Use a sequential testing strategy: apply the sup-F test to determine whether at least one break exists, then use UDmax or WDmax for the maximum number, and finally confirm with the BIC or the Liu-Wu-Zidek (LWZ) criterion. Economic priors about known events should also guide the choice.
Does the structural break AR model require the series to be stationary?
Each segment should be stationary conditional on its regime parameters; the model is typically applied to stationary or trend-stationary series. If the overall series is I(1), consider whether differencing or a structural break unit-root test (Zivot-Andrews) is more appropriate before fitting a break AR model.
What is the minimum segment length requirement?
Bai and Perron recommend a trimming parameter of 10-15% of the total sample as the minimum allowable segment length. For a series of 100 observations this means each segment must contain at least 10-15 data points; in practice 20+ per segment is preferable for reliable AR coefficient estimates.
How is this different from a Markov-switching AR model?
A Markov-switching AR model allows regimes to recur probabilistically at every period, so the series can switch back and forth between states. The structural break AR model assumes breaks are permanent and the series never returns to a previous regime, making it more appropriate when a break reflects a one-time policy or institutional change.
Can I use this model for forecasting?
Yes, but forecasting accuracy depends on whether future data remain in the last identified regime. If the most recent break is well in the past and the regime appears stable, segment-specific AR forecasts often outperform a full-sample AR that ignores the break. For very recent or anticipated breaks, scenario-based forecasting across possible regimes is advisable.
Sources
- Bai, J., & Perron, P. (2003). Computation and analysis of multiple structural change models. Journal of Applied Econometrics, 18(1), 1-22. DOI: 10.1002/jae.659 ↗
- 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). Autoregressive Model with Structural Breaks. ScholarGate. https://scholargate.app/en/econometrics/structural-break-ar-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.
- Augmented Dickey-Fuller unit root testEconometrics↔ compare
- Autoregressive modelEconometrics↔ compare
- Structural Break ARIMA ModelEconometrics↔ compare
- Structural Break VAR ModelEconometrics↔ compare
- Structural break VECMEconometrics↔ compare
- Zivot-Andrews Structural Break TestEconometrics↔ compare