CIPS Test — Cross-sectionally Augmented IPS Panel Unit-Root Test
Also known as: Pesaran CIPS Test, Cross-Sectionally Augmented IPS, Second-Generation Panel Unit-Root Test, CIPS Birim Kök Testi
The CIPS test, introduced by Pesaran (2007), is a second-generation panel unit-root test designed for panels in which the cross-sectional units share unobserved common factors that induce cross-section dependence. By augmenting each individual ADF regression with cross-sectional averages and their lags, the CIPS test accounts for this dependence and produces reliable inference where first-generation tests such as the original IPS test break down. It is widely applied in macroeconomic and finance panels where shocks propagate across countries or regions.
Key highlights
- Explicitly accounts for cross-section dependence through cross-sectional augmentation, correcting the severe size distortions of first-generation tests such as LLC or IPS under dependence.
- Computationally simple: requires only OLS estimation of N augmented ADF regressions followed by averaging, with no need to estimate factor models directly.
- Valid under both stationary and non-stationary common factors, making it robust to a wide range of data-generating processes.
- Tabulated critical values covering a wide grid of (N, T) combinations are provided by Pesaran (2007), facilitating straightforward inference.
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 CIPS test when you have a balanced or nearly balanced panel and prior evidence or theory suggests cross-section dependence — for instance, panels of countries, regions, or firms exposed to common macroeconomic shocks. The test assumes a single-factor structure for the error cross-section dependence and works best when T is moderate to large relative to N. It should not replace, but rather complement, diagnostic tests for cross-section dependence (e.g., Pesaran's CD test). If multiple common factors are suspected, the PANIC approach may be preferable.
Strengths & limitations
- Explicitly accounts for cross-section dependence through cross-sectional augmentation, correcting the severe size distortions of first-generation tests such as LLC or IPS under dependence.
- Computationally simple: requires only OLS estimation of N augmented ADF regressions followed by averaging, with no need to estimate factor models directly.
- Valid under both stationary and non-stationary common factors, making it robust to a wide range of data-generating processes.
- Tabulated critical values covering a wide grid of (N, T) combinations are provided by Pesaran (2007), facilitating straightforward inference.
- The single-factor cross-section dependence assumption may be violated in panels driven by multiple common shocks, reducing test accuracy.
- Critical values are tabulated for specific (N, T) combinations; interpolation is required for non-tabulated panel dimensions.
- The test has low power when only a small fraction of units are stationary (the alternative hypothesis is heterogeneous), though this is common to all IPS-type tests.
- Requires a balanced or near-balanced panel; large proportions of missing observations complicate the computation of cross-sectional averages.
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 CIPS test differ from the original IPS test?
The IPS test averages individual ADF t-statistics from regressions that contain only unit-specific regressors, making it invalid under cross-section dependence. The CIPS test augments each individual ADF regression with cross-sectional averages and their lags to proxy for common factors, restoring size control under dependence while retaining the simple averaging structure of IPS.
Do I need to estimate a factor model before running the CIPS test?
No. That is the key practical advantage of the CIPS approach. The cross-sectional mean of the observed panel series acts as a proxy for the latent common factor, so no separate factor extraction step (such as principal components) is required. You only need to run N augmented ADF regressions by OLS and then average the resulting t-statistics.
What should I do if the CIPS test and a first-generation test give conflicting results?
When a CD test or similar diagnostic confirms cross-section dependence, you should trust the CIPS result over the first-generation test, because the latter is known to be oversized under dependence. If cross-section dependence is absent, both tests should agree asymptotically; any discrepancy is likely a finite-sample artifact that may warrant checking lag selection and deterministic specification.
Sources
- 1.Pesaran, M. H. (2007). A simple panel unit root test in the presence of cross-section dependence. Journal of Applied Econometrics, 22(2), 265–312.
You have read it. What now?
Cite this page
ScholarGate. (2026, June 2). CIPS Test. ScholarGate. https://scholargate.app/econometrics/cips-test