Quandt-Andrews Test for Unknown Structural Breaks
Also known as: sup-Wald Test, Andrews Breakpoint Test, Unknown Structural Break Test, Quandt Likelihood Ratio Test
The Quandt-Andrews test, formalized by Andrews (1993), detects structural breaks in regression parameters when the breakpoint date is unknown a priori. It sweeps all candidate break dates within a trimmed interior of the sample, computes a Wald (or LM/LR) statistic at each candidate, and reports the supremum of those statistics. Applied economists and time-series analysts use it to test whether coefficients remain stable across a full estimation window without needing to specify when the break occurred.
Key highlights
- Tests for structural instability without requiring a pre-specified break date, making it widely applicable in empirical economics.
- Andrews (1993) provides asymptotically valid critical values for a broad class of estimators, including OLS and GMM.
- The date that maximizes the test statistic is a consistent estimator of the true break date under the alternative.
- Accommodates heteroskedasticity and autocorrelation via HAC covariance estimation, improving robustness in many time-series settings.
Intuition
This section is available to Pro members. Upgrade to Pro
How it works
This section is available to Pro members. Upgrade to Pro
When to use it
Use the Quandt-Andrews test when economic or policy reasoning suggests a structural break may have occurred but theory does not pinpoint the exact date. The test requires a stationary or integrated regressor setting with serially uncorrelated or heteroskedasticity-and-autocorrelation-consistent errors. It applies to linear regressions, VAR systems, and generalized method-of-moments estimators. The sample must be large enough to allow meaningful trimming (at least 30–40 observations is a rough practical lower bound). Alternatives include the Chow test (when the break date is known), the Bai-Perron test (when multiple breaks are suspected), and CUSUM-based tests for cumulative parameter drift.
Strengths & limitations
- Tests for structural instability without requiring a pre-specified break date, making it widely applicable in empirical economics.
- Andrews (1993) provides asymptotically valid critical values for a broad class of estimators, including OLS and GMM.
- The date that maximizes the test statistic is a consistent estimator of the true break date under the alternative.
- Accommodates heteroskedasticity and autocorrelation via HAC covariance estimation, improving robustness in many time-series settings.
- Low power against gradual parameter drift or smooth transitions, where no single date dominates as a break point.
- Trimming removes a meaningful share of the sample from consideration, so breaks near the endpoints of the series cannot be detected.
- Critical values depend on the trimming fraction and the number of restrictions tested; using incorrect tables invalidates inference.
- The test identifies at most one break location; multiple structural changes require Bai-Perron or sequential testing procedures.
Common pitfalls
This section is available to Pro members. Upgrade to Pro
Applications
This section is available to Pro members. Upgrade to Pro
Frequently asked
How does the Quandt-Andrews test differ from the Chow test?
The Chow test requires the analyst to specify the break date before looking at the data; using it at a data-selected date inflates size. The Quandt-Andrews test explicitly accounts for the search over all candidate dates by using non-standard critical values derived by Andrews (1993), providing valid inference when the break location is unknown.
What trimming fraction should I choose?
Andrews (1993) recommends 15 percent as the default, excluding the bottom and top 15 percent of observations from the grid of candidate dates. Smaller fractions allow detection closer to the endpoints but produce larger size distortions in finite samples; the 15 percent choice balances power and size well across typical sample sizes.
Can the test handle multiple structural breaks?
No. The Quandt-Andrews test is designed to detect a single unknown break and its power diminishes in the presence of multiple breaks because the supremum statistic may be driven by the largest but not the only instability. For data suspected of containing two or more breaks, the Bai-Perron sequential testing procedure is the appropriate generalization.
Sources
- 1.Andrews, D. W. K. (1993). Tests for parameter instability and structural change with unknown change point. Econometrica, 61(4), 821–856.
You have read it. What now?
Cite this page
ScholarGate. (2026, June 2). Quandt-Andrews Test. ScholarGate. https://scholargate.app/econometrics/quandt-andrews-test