Time-Varying Parameter KPSS Test
Time-Varying Parameter Kwiatkowski-Phillips-Schmidt-Shin Test · Also known as: TVP-KPSS test, time-varying KPSS stationarity test, locally stationary KPSS test, TV-KPSS
The time-varying parameter KPSS test extends the classic Kwiatkowski-Phillips-Schmidt-Shin (1992) stationarity test to settings where the deterministic or stochastic components of a series may shift over time. It tests the null hypothesis of stationarity while allowing the model's parameters to evolve, making it robust to structural instability that would otherwise distort the standard KPSS result.
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 TVP-KPSS test when you suspect that a time series contains structural breaks, gradual parameter shifts, or unconditional heteroscedasticity that would invalidate the standard KPSS test. It is particularly suitable for macroeconomic and financial series spanning long horizons where volatility regimes (e.g., the Great Moderation) or policy changes make a fixed-parameter model implausible. Do not use it as a simple drop-in replacement for KPSS when the series is short and homogeneous — the standard KPSS test has better finite-sample power in that case. Also avoid it when breaks are known and precisely dated; a Zivot-Andrews or LS unit root test with explicit break points will be more powerful.
Strengths & limitations
- Maintains correct test size under structural breaks and non-stationary volatility that cause the standard KPSS test to over-reject stationarity.
- Provides a framework that explicitly accommodates gradual parameter evolution, aligning with empirical features of long macroeconomic and financial series.
- Retains the intuitive null-of-stationarity framing of KPSS, making interpretation straightforward for applied researchers.
- Can be combined with rolling or recursive windows to visualise when and where stationarity is violated across the sample.
- Requires selecting a bandwidth or kernel for the time-varying long-run variance estimator, and results can be sensitive to this choice.
- Finite-sample power against a unit root alternative may be lower than that of ADF or Phillips-Perron tests, especially in short samples.
- The limiting distribution under the TVP null may differ from standard KPSS critical values, requiring simulation or bootstrap critical values.
- Implementation is not available in basic econometric packages; specialised R or MATLAB code is typically needed.
Frequently asked
How does the TVP-KPSS test differ from the standard KPSS test?
The standard KPSS test assumes fixed parameters and constant error variance throughout the sample. The TVP-KPSS test relaxes these assumptions by using a time-varying or heteroscedasticity-robust long-run variance estimator, so its null distribution remains valid when the series exhibits structural instability or changing volatility.
Should I use the TVP-KPSS test instead of the Zivot-Andrews test when I suspect a break?
They address different problems. Zivot-Andrews tests the unit-root null allowing for a single unknown break point. TVP-KPSS tests the stationarity null while accommodating gradual or multiple parameter shifts. If you have reason to think a sharp single break occurred, the break-point approach may be more powerful; for gradual or multiple changes, the TVP-KPSS framework is better suited.
What critical values should I use for the TVP-KPSS test?
Standard KPSS critical values (0.739 at 5% for the level-stationary case) are not valid under the TVP extension. You should use bootstrap critical values generated under the null, or rely on simulation-based tables provided in the specific TVP-KPSS paper you are implementing.
Can the TVP-KPSS test be applied to panel data?
The test is primarily designed for univariate time series. Panel extensions of KPSS-type tests exist separately (e.g., Hadri, 2000), and combining time-varying parameter corrections with panel settings is an active research area, but is not part of the standard TVP-KPSS setup.
What happens if I apply the standard KPSS test when parameters are time-varying?
The standard KPSS test becomes oversized — it rejects stationarity too often even when the series is genuinely stationary within each sub-regime. This inflated rejection rate can lead to incorrectly concluding that a series is non-stationary, which distorts subsequent modelling choices such as differencing or cointegration 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: How sure are we that economic time series have a unit root? Journal of Econometrics, 54(1-3), 159-178. DOI: 10.1016/0304-4076(92)90104-Y ↗
- Cavaliere, G., & Taylor, A. M. R. (2007). Testing for unit roots in time series models with non-stationary volatility. Journal of Econometrics, 140(2), 919-947. DOI: 10.1016/j.jeconom.2006.07.019 ↗
How to cite this page
ScholarGate. (2026, June 3). Time-Varying Parameter Kwiatkowski-Phillips-Schmidt-Shin Test. ScholarGate. https://scholargate.app/en/econometrics/time-varying-parameter-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
- Phillips-Perron TestEconometrics↔ compare