Zivot-Andrews Unit-Root Test with One Structural Break
Also known as: ZA Test, Zivot-Andrews Break Test, Endogenous Break Unit-Root Test, Zivot-Andrews Birim Kök Testi
The Zivot-Andrews (ZA) test, introduced by Eric Zivot and Donald Andrews in 1992, is a sequential unit-root test that allows for a single structural break at an unknown date. It extends the augmented Dickey-Fuller framework by endogenously selecting the break point that provides the strongest evidence against the unit-root null hypothesis, making it particularly useful for macroeconomic and financial time series that may have been disrupted by events such as policy changes, financial crises, or supply 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 Zivot-Andrews test when your time series is suspected to contain a single structural break at an unknown date and you need to distinguish between a unit root and a trend-stationary process with a break. It is appropriate for macroeconomic aggregates (GDP, inflation, exchange rates) and financial series spanning episodes of policy reform or market disruption. The test assumes at most one break; if two or more breaks are likely, prefer the Lumsdaine-Papell or Lee-Strazicich tests. The sample should be large enough (typically n ≥ 50) to obtain reliable asymptotic approximations, and the lag length should be chosen by information criteria.
Strengths & limitations
- Endogenous break-date selection eliminates the need for prior knowledge of when the structural break occurred.
- Higher power than the standard ADF test in the presence of a genuine structural break.
- Three model variants (level, slope, and combined break) provide flexibility to match the data-generating process.
- Asymptotic critical values account for the search over break dates, maintaining correct size.
- Restricted to a single structural break; multiple breaks require alternative procedures.
- The test can still over-reject the null when breaks are ignored and the true number of breaks exceeds one.
- Asymptotic critical values are derived under specific distributional assumptions and may not perform well in small samples.
- Selecting the wrong model (A, B, or C) can distort both size and power.
Frequently asked
How does the Zivot-Andrews test differ from the standard ADF test?
The ADF test assumes no structural breaks and loses power when a break is present. The Zivot-Andrews test explicitly incorporates a one-time break in the null and alternative hypotheses and searches for the break date endogenously. It uses more negative critical values to account for this search, preserving correct test size while gaining power against trend-stationary alternatives with a single break.
Which model (A, B, or C) should I use in practice?
Model A is appropriate when theory or inspection of the data suggests a sudden level shift with no change in trend slope. Model B is suited to a change in the growth rate without a level jump. Model C is the most general and is recommended when the nature of the break is uncertain, though it uses the most parameters and therefore has slightly less power than the more restricted models when the true break is of a simpler type.
What should I do if I suspect more than one structural break?
The Zivot-Andrews test is designed for a single break, so applying it to a series with multiple breaks risks both size distortion and misleading break-date identification. In that case, use the Lumsdaine-Papell test (two endogenous breaks) or the Lee-Strazicich LM-based test (one or two breaks), which are explicitly designed for multiple-break environments and maintain better size properties.
Sources
- 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 2). Zivot-Andrews Unit-Root Test with One Structural Break. ScholarGate. https://scholargate.app/en/econometrics/zivot-andrews-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.
- Augmented Dickey-Fuller TestEconometrics↔ compare
- Lee-Strazicich TestEconometrics↔ compare
- Lumsdaine-Papell TestEconometrics↔ compare