Time-Varying Parameter Phillips-Perron Unit Root Test
Also known as: TVP-PP unit root test, time-varying PP test, Phillips-Perron test with time-varying parameters, TVP unit root test
The time-varying parameter PP unit root test extends the classical Phillips-Perron test by allowing the autoregressive coefficient to change over time. It detects stochastic non-stationarity in series whose persistence may shift across regimes or periods, offering more reliable inference when structural change is suspected in the data-generating process.
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-PP test when you suspect the persistence of a time series may have changed at one or more unknown points — for example, after a monetary regime switch, financial crisis, or major policy reform. It is particularly valuable when standard PP or ADF tests yield ambiguous or contradictory results that could stem from structural instability rather than genuine non-stationarity. Avoid it for very short series (fewer than 50 observations) where the Kalman filter cannot estimate the state variance reliably, and do not use it as a first-line test when there is no theoretical reason to expect parameter instability, as it introduces additional estimation complexity without benefit.
Strengths & limitations
- Allows the autoregressive coefficient to vary over time, capturing regime changes that fixed-coefficient tests cannot handle.
- Inherits the PP correction for serial correlation and heteroscedasticity without requiring a parametric error model.
- Reduces the risk of falsely retaining the unit-root null when the true process is stationary around a shifting mean or trend.
- Critical values are drawn from the same Dickey-Fuller distribution, keeping inference straightforward.
- Compatible with both demeaned and detrended specifications, mirroring the flexibility of the standard PP test.
- Requires sufficiently long series (typically at least 50-80 observations) for the Kalman filter to identify the state variance accurately.
- The choice of bandwidth for the Newey-West correction and the initial state variance can affect finite-sample performance.
- Power may be lower than structural-break unit root tests (e.g., Zivot-Andrews) when a single sharp break is the true data-generating process.
- Software support is less widespread than for standard PP or ADF tests, requiring custom coding in R or Matlab in many cases.
Frequently asked
How does the TVP-PP test differ from the standard PP test?
The standard PP test assumes the autoregressive coefficient is fixed throughout the sample and applies a Newey-West correction to handle serial correlation. The TVP-PP test allows that coefficient to evolve over time via a state-space model estimated with the Kalman filter, making it robust to changing persistence without requiring knowledge of when or how many breaks occurred.
Can I use the same critical values as the standard PP test?
Under the null hypothesis of a unit root the TVP-PP test statistic converges to the same Dickey-Fuller limiting distribution, so the standard PP critical values remain valid asymptotically. Finite-sample critical values obtained by simulation are preferable for series shorter than 100 observations.
When should I prefer the Zivot-Andrews test over the TVP-PP test?
If prior knowledge or graphical inspection strongly suggests a single, sharp structural break at an unknown date, the Zivot-Andrews test is typically more powerful. The TVP-PP test is preferable when structural change is gradual or when multiple shifts are possible and their timing is unknown.
What happens if I apply the TVP-PP test to a series with no structural change?
If the true process has constant parameters, the Kalman filter will estimate near-zero state variance and the test degenerates toward the standard PP test. The size properties remain valid, but you gain no power advantage over the simpler PP test — so use it selectively when instability is a genuine concern.
Does the TVP-PP test require stationarity of the time-varying parameter itself?
No. The state equation specifies a random walk for the autoregressive coefficient, so the parameter itself can drift without bound. The test statistic nevertheless converges to the standard Dickey-Fuller distribution under the null, a result that follows from the specific construction of the PP-type long-run variance correction.
Sources
- Phillips, P. C. B., & Perron, P. (1988). Testing for a unit root in time series regression. Biometrika, 75(2), 335-346. DOI: 10.1093/biomet/75.2.335 ↗
- Hall, S. G., & Luginbuhl, R. (1999). Modelling structural breaks in unit root tests using time-varying parameter models. Journal of Economic Dynamics and Control, 23(2), 209-231. link ↗
How to cite this page
ScholarGate. (2026, June 3). Time-Varying Parameter Phillips-Perron Unit Root Test. ScholarGate. https://scholargate.app/en/econometrics/time-varying-parameter-pp-unit-root-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.
- KPSS TestEconometrics↔ compare
- Phillips-Perron TestEconometrics↔ compare
- Zivot-Andrews TestEconometrics↔ compare