Panel Phillips-Perron Unit Root Test
Also known as: Panel PP test, Phillips-Perron panel unit root, Im-Pesaran-Shin PP panel test, panel nonparametric unit root test
The Panel PP unit root test extends the nonparametric Phillips-Perron correction for serial correlation to a multi-individual panel setting. It tests the null hypothesis that all cross-sectional units contain a unit root, using a pooled or averaged PP-type statistic that is robust to heteroscedastic and serially correlated errors without requiring explicit lag selection.
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 Panel PP test when you have panel data with a moderate to large time dimension (T at least 20-25 per unit) and you want to test for unit roots without committing to a specific lag order, since the nonparametric bandwidth correction handles serial correlation automatically. It is appropriate when error variances may differ across units (cross-sectional heteroscedasticity). Do not use it when T is very short (below 15), as the nonparametric correction becomes unreliable; prefer the Panel ADF test with lag selection in that case. Also avoid it when cross-sectional dependence is strong — use second-generation tests such as CIPS (Pesaran 2007) instead, because the Panel PP statistic assumes cross-sectional independence under the null.
Strengths & limitations
- Nonparametric serial-correlation correction avoids the need to specify a lag length, reducing one source of specification error.
- Robust to heteroscedastic and autocorrelated errors within each unit.
- Combines information across all cross-sections, yielding higher power than individual time-series PP tests when N is large.
- Well-understood asymptotic distribution (standard normal) facilitates straightforward inference.
- Widely implemented in standard software (Stata, EViews, R's plm package).
- Assumes cross-sectional independence; the null distribution is distorted when units share common factors or shocks.
- Sensitive to bandwidth selection for the long-run variance estimator; different kernels or bandwidths can produce different conclusions.
- Low power when T is small (below 20) because the nonparametric correction is imprecise with short samples.
- The null of a unit root in all units is rejected even if only one unit is stationary, so the test does not tell you which units are stationary.
Frequently asked
How does the Panel PP test differ from the Panel ADF test?
Both test the same null (unit root in all units), but the Panel ADF test adds explicit lagged differences to the regression to capture serial correlation, requiring a lag-order choice. The Panel PP test instead applies a nonparametric bandwidth-based correction to the standard OLS output, so no lag-order decision is needed. In practice, when T is moderate, the two tests usually agree; when they disagree, the difference often traces to lag or bandwidth misspecification.
What should I do if my panel data has strong cross-sectional dependence?
The Panel PP test (like the Panel ADF test) is a first-generation test that assumes cross-sectional independence. With strong dependence — common in macro panels driven by global shocks — use a second-generation test such as Pesaran's (2007) CIPS test, which is explicitly designed for cross-sectionally dependent panels.
How do I choose the bandwidth for the long-run variance estimator?
The standard practice is to use an automatic bandwidth selector such as the Newey-West (Bartlett kernel) plug-in formula, which scales roughly as 4(T/100)^(2/9). Most software defaults implement this. You can also check robustness by trying one or two alternative bandwidths; if conclusions change substantially, treat results with caution.
Should I include a constant, a trend, or neither in the test regression?
Match the deterministic terms to economic theory and graphical inspection of your series. Include only a constant if the series fluctuates around a non-zero level with no trend; add a linear trend if the series appears to trend upward or downward over time. Including a spurious trend reduces power, while omitting a genuine trend inflates the rejection rate.
What happens if the Panel PP test rejects the null?
Rejection implies that at least some cross-sectional units are stationary (or trend-stationary), but the test does not identify which ones. Follow up with individual-unit PP or ADF tests to determine which series are stationary, then decide whether to model units in levels or first differences accordingly.
Sources
- Im, K. S., Pesaran, M. H., & Shin, Y. (2003). Testing for unit roots in heterogeneous panels. Journal of Econometrics, 115(1), 53-74. DOI: 10.1016/S0304-4076(03)00092-7 ↗
- 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 ↗
How to cite this page
ScholarGate. (2026, June 3). Panel Phillips-Perron Unit Root Test. ScholarGate. https://scholargate.app/en/econometrics/panel-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.
- Augmented Dickey-Fuller unit root testEconometrics↔ compare
- Panel ADF Unit Root TestEconometrics↔ compare
- Panel ARDL Bounds TestEconometrics↔ compare
- Panel Engle-Granger CointegrationEconometrics↔ compare
- Panel KPSS testEconometrics↔ compare
- Phillips-Perron unit root testEconometrics↔ compare