Vector Error Correction Model with Structural Breaks (SB-VECM)
Vector Error Correction Model with Structural Breaks · Also known as: SB-VECM, VECM with regime shifts, cointegration model with structural breaks, break-augmented VECM
The Structural Break VECM extends the standard Vector Error Correction Model to allow the cointegrating relationships, adjustment speeds, or short-run dynamics to shift at one or more known or estimated break dates. It preserves the long-run equilibrium framework of the VECM while explicitly modelling regime changes caused by policy shifts, crises, or institutional changes.
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 VECM when you have two or more I(1) time series that share a long-run equilibrium but the sample contains a known or suspected major regime change — for example a financial crisis, a monetary policy shift, a currency reform, or a commodity price collapse — that is likely to alter the cointegrating relationship or adjustment dynamics. It is especially appropriate when a standard VECM yields unstable parameters or poor residual properties, and when Gregory-Hansen or similar tests reject cointegration under the null of no break but support cointegration once a break is allowed. Do not use it when there is no genuine cointegration (even with breaks), when the break date is purely data-mined without economic justification, or when the sample is too short (fewer than roughly 60 observations per regime) to estimate separate regimes reliably.
Strengths & limitations
- Preserves the long-run cointegration framework while accommodating major economic regime changes.
- Avoids spurious rejection of cointegration that arises when structural breaks are ignored in standard Johansen tests.
- Provides separate estimates of adjustment speeds and long-run coefficients for each regime, enabling richer economic interpretation.
- Gregory-Hansen and Johansen-Mosconi-Nielsen procedures are well-established with published critical values.
- Consistent with impulse response and forecast error variance decomposition analysis within each regime.
- Requires a sufficient number of observations in each regime; short regimes yield imprecise estimates.
- Endogenous break-date search raises pre-test concerns: inference conditional on an estimated break date must use appropriate critical values.
- The number of parameters grows rapidly with the number of breaks, variables, and lags, increasing the risk of over-fitting.
- Assumes breaks are abrupt (instantaneous); gradual or smooth transitions are better handled by smooth-transition VECM specifications.
- Software implementations vary in the break-augmented Johansen critical values they use, requiring careful verification.
Frequently asked
How does SB-VECM differ from a standard VECM?
A standard VECM assumes the cointegrating vector and adjustment coefficients are constant throughout the sample. The SB-VECM allows these parameters to change at one or more break dates, so the long-run equilibrium and the speed of return to it can differ across regimes.
How do I choose the break date?
Breaks can be pre-specified from economic knowledge (e.g., the 2008 financial crisis) or estimated endogenously. Endogenous estimation typically uses the Gregory-Hansen supremum-based test, which searches over all candidate break dates and selects the one that minimises the residual variance or maximises the test statistic. When estimated endogenously, inference must use the supremum critical values rather than conventional ones.
What if I suspect more than one structural break?
Multiple breaks can be accommodated by including additional dummy variables for each regime and applying the Bai-Perron sequential procedure or the Hatemi-J (2008) two-break cointegration test. Each additional break reduces the effective sample within a regime, so parsimony is important.
Can I still perform impulse response analysis after SB-VECM estimation?
Yes, but impulse responses should be computed and reported separately for each regime, since the short-run dynamics and adjustment speeds may differ. Generalized impulse responses (Pesaran and Shin 1998) are preferred when the shock ordering is uncertain.
What is the minimum sample size recommended?
A common rule of thumb is at least 60 observations per regime after break-date trimming, giving roughly 120 or more total observations for a single break. Smaller sub-samples produce unreliable estimates of the cointegrating vector and adjustment coefficients.
Sources
- Gregory, A. W., & Hansen, B. E. (1996). Residual-based tests for cointegration in models with regime shifts. Journal of Econometrics, 70(1), 99–126. DOI: 10.1016/0304-4076(69)41685-7 ↗
- Johansen, S., Mosconi, R., & Nielsen, B. (2000). Cointegration analysis in the presence of structural breaks in the deterministic trend. Econometrics Journal, 3(2), 216–249. DOI: 10.1111/1368-423X.00047 ↗
How to cite this page
ScholarGate. (2026, June 3). Vector Error Correction Model with Structural Breaks. ScholarGate. https://scholargate.app/en/econometrics/structural-break-vecm
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.
- Nonlinear VECMEconometrics↔ compare
- Structural break Johansen cointegrationEconometrics↔ compare
- Structural Break VAR ModelEconometrics↔ compare
- Vector Error Correction ModelEconometrics↔ compare
- Zivot-Andrews Structural Break TestEconometrics↔ compare