Structural Break ARDL Bounds Test
Structural Break Autoregressive Distributed Lag Bounds Test · Also known as: SB-ARDL bounds test, ARDL bounds test with structural break, Fourier ARDL bounds test, break-augmented bounds testing
The structural break ARDL bounds test extends the Pesaran, Shin and Smith (2001) bounds testing framework to accommodate one or more structural breaks in the long-run relationship between time-series variables. By incorporating break dummies or smooth Fourier terms into the ARDL error-correction equation, it allows researchers to test for cointegration even when the data have experienced shifts in intercept or slope caused by policy changes, crises, or regime switches.
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 ARDL bounds test when you suspect or detect that the long-run relationship between I(0) or I(1) variables (or a mixture) shifted at one or more known or unknown points in the sample — for example, after a financial crisis, a policy reform, or a commodity price shock. It is preferable to plain ARDL bounds testing whenever residual diagnostics or visual inspection suggest parameter instability. It is not appropriate when variables are I(2), when breaks are too frequent relative to the sample size (leaving too few observations per regime), or when the true break date is entirely unknown and the sample is short (fewer than 40–50 observations), as power deteriorates sharply in such cases.
Strengths & limitations
- Accommodates structural breaks while preserving the main advantage of ARDL: no requirement that all variables share the same order of integration.
- Can be applied with a relatively small sample compared to multivariate cointegration tests such as Johansen.
- Captures both known (exogenous) break dates and endogenously detected breaks in a single regression framework.
- Fourier-based variants can approximate multiple smooth, gradual breaks without specifying exact break dates.
- The error-correction model directly delivers the short-run dynamics and break-adjusted long-run equilibrium in one step.
- Power declines sharply when the sample is small or when breaks occur near the beginning or end of the sample, leaving few observations in one regime.
- When breaks are endogenously detected, the F-test critical values from PSS (2001) may be distorted; bootstrap critical values are more reliable but add complexity.
- The method does not handle I(2) variables; a prior unit-root analysis with break-robust tests (e.g., Zivot-Andrews, Lee-Strazicich) is required.
- Selecting the wrong break date or omitting a break can bias the long-run coefficients and invalidate the test conclusion.
Frequently asked
Can I use PSS (2001) critical values when I detect the break endogenously?
Not reliably. Endogenous break detection introduces a pre-testing problem that distorts the asymptotic distribution of the F-statistic. Bootstrap critical values tailored to your sample size and break fraction are safer and are increasingly standard in published work.
How many breaks can I include?
Most applications include one or two breaks. With each additional break, you lose degrees of freedom and power. Practical guidance is to have at least 20–25 observations in each regime; with fewer, inference is unreliable.
What if I do not know the exact break date?
Use a Bai-Perron multiple break test or the Zivot-Andrews unit-root test to identify candidate break dates before running the augmented ARDL. Alternatively, replace dummies with Fourier sine and cosine terms, which approximate unknown smooth breaks without specifying exact dates.
Is this the same as the Fourier ARDL bounds test?
Related but not identical. The Fourier ARDL bounds test (Enders & Jones 2016) is a specific variant that uses low-frequency trigonometric terms to capture gradual or multiple smooth breaks. The general structural break ARDL category also includes sharp break dummies for abrupt regime changes.
Do I still need to test for unit roots before the bounds test?
Yes. You must confirm that no variable is I(2). Use break-robust unit-root tests such as Zivot-Andrews or Lee-Strazicich, since standard ADF or PP tests have low power in the presence of structural breaks and may falsely indicate I(2) behaviour.
Sources
- Pesaran, M. H., Shin, Y., & Smith, R. J. (2001). Bounds testing approaches to the analysis of level relationships. Journal of Applied Econometrics, 16(3), 289–326. DOI: 10.1002/jae.616 ↗
- Enders, W., & Jones, P. (2016). Grain prices, oil prices, and multiple smooth breaks in a VAR. Studies in Nonlinear Dynamics and Econometrics, 20(4), 399–419. DOI: 10.1515/snde-2014-0101 ↗
How to cite this page
ScholarGate. (2026, June 3). Structural Break Autoregressive Distributed Lag Bounds Test. ScholarGate. https://scholargate.app/en/econometrics/structural-break-ardl-bounds-test
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.
- ARDL Bounds TestEconometrics↔ compare
- Engle-Granger Cointegration TestEconometrics↔ compare
- Fourier ARDL Bounds TestEconometrics↔ compare
- Nonlinear ARDLEconometrics↔ compare
- Structural break VECMEconometrics↔ compare
- Zivot-Andrews Structural Break TestEconometrics↔ compare