Robust KPSS Test for Stationarity
Robust Kwiatkowski-Phillips-Schmidt-Shin Test · Also known as: Robust KPSS, outlier-robust stationarity test, robust LM stationarity test, KPSS with robustness correction
The Robust KPSS test is an extension of the classical Kwiatkowski-Phillips-Schmidt-Shin (1992) stationarity test that replaces the conventional long-run variance estimator with an outlier-robust or heteroscedasticity-robust counterpart, maintaining reliable size and power in the presence of contaminated observations, structural breaks, or non-standard error distributions.
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 Robust KPSS test when you suspect outliers, additive contamination, or a possible structural break in your time series but still need to test the stationarity null. It is particularly appropriate in financial or macroeconomic series prone to sudden level shifts and in situations where prior ADF or standard KPSS results conflict. It is not a substitute for explicit structural-break unit root tests (Zivot-Andrews, Lumsdaine-Papell) when the break date is the primary object of interest, nor is it suitable for multivariate or panel settings without additional adaptation.
Strengths & limitations
- Maintains reliable empirical size in the presence of additive outliers or heavy-tailed errors that distort the standard KPSS test.
- Null hypothesis is stationarity, making it complementary to ADF-type tests whose null is a unit root — using both together helps resolve conflicting conclusions.
- Retains the familiar KPSS framework and critical values, easing interpretation for practitioners already acquainted with the standard test.
- More powerful than the standard KPSS test when the data contain outliers that artificially inflate the conventional long-run variance.
- Critical values may differ from standard KPSS tables depending on which robust variance estimator is used; practitioners must verify the appropriate table.
- Does not explicitly identify the location or number of structural breaks, unlike dedicated break-robust tests.
- Performance depends on the choice of bandwidth and kernel for the robust long-run variance, introducing tuning sensitivity.
- For multivariate stationarity testing, separate multivariate robust extensions are required.
Frequently asked
How does the Robust KPSS differ from the standard KPSS test?
The standard KPSS uses a Newey-West long-run variance estimator that is sensitive to outliers and conditional heteroscedasticity. The robust variant replaces this with a weighted or robust kernel estimator that reduces the influence of extreme observations, yielding more reliable size and power in contaminated samples.
Should I use the level-stationary or trend-stationary version?
If economic theory or a visual inspection suggests a deterministic trend, include the trend in the regression step (trend-stationary null). If the series fluctuates around a constant mean, use only the intercept (level-stationary null). The choice affects both the test statistic and the critical values.
What if KPSS and ADF give conflicting conclusions?
Conflicting results — ADF fails to reject a unit root while KPSS rejects stationarity, or vice versa — typically signal that the series has a structural break or fractional integration. Consider a break-robust test (Zivot-Andrews) or a fractional integration test before deciding.
How do I choose the bandwidth for the robust long-run variance?
Data-driven bandwidth selection rules (such as the Andrews 1991 automatic rule or the Newey-West fixed-bandwidth rule) are standard starting points. In practice it is prudent to check sensitivity by repeating the test with several bandwidth choices.
Can I apply this test to panels or multivariate systems?
The univariate Robust KPSS test is not directly applicable to panels. Panel stationarity tests (e.g., Hadri 2000, or its robust extensions) are needed for cross-sectional data, and multivariate KPSS extensions exist for system-level stationarity testing.
Sources
- Kwiatkowski, D., Phillips, P. C. B., Schmidt, P., & Shin, Y. (1992). Testing the null hypothesis of stationarity against the alternative of a unit root. Journal of Econometrics, 54(1-3), 159-178. DOI: 10.1016/0304-4076(92)90104-Y ↗
- Hobijn, B., Franses, P. H., & Ooms, M. (2004). Generalizations of the KPSS-test for stationarity. Statistica Neerlandica, 58(4), 483-502. DOI: 10.1111/j.1467-9574.2004.00272.x ↗
How to cite this page
ScholarGate. (2026, June 3). Robust Kwiatkowski-Phillips-Schmidt-Shin Test. ScholarGate. https://scholargate.app/en/econometrics/robust-kpss-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
- KPSS TestEconometrics↔ compare