Breitung Panel Unit-Root Test
Also known as: Breitung Panel Unit-Root Test, Breitung (2000) Test, Breitung Nonparametric Panel Unit-Root Test, Breitung Panel Birim Kök Testi
The Breitung test, introduced by Jörg Breitung in 2000, is a nonparametric panel unit-root test designed to assess whether all cross-sectional units in a balanced panel share a common unit root. Unlike competing first-generation tests, it avoids bias-correction terms that depend on lag selection or kernel bandwidth estimation, thereby preserving local power under a homogeneous alternative. It is widely used in macroeconometrics and finance when the researcher suspects cross-sectional homogeneity in the autoregressive structure.
Key highlights
- No lag-length or kernel bandwidth selection needed, eliminating a common source of specification error.
- Asymptotically standard normal distribution simplifies inference without special critical-value tables.
- Higher local power than bias-corrected tests such as Levin-Lin-Chu under a homogeneous alternative.
- Computationally simple to implement given its reliance on straightforward OLS on transformed series.
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
Apply the Breitung test when working with balanced macro or financial panels where a homogeneous autoregressive parameter across units is a credible assumption, such as GDP growth rates or interest rates across countries under a common monetary regime. The test requires a balanced panel with moderately large T (typically T > 10) and assumes cross-sectional independence among units. It is not appropriate when heterogeneous dynamics are expected (prefer Im-Pesaran-Shin in that case) or when cross-sectional dependence is present (prefer second-generation tests such as CIPS). Always complement it with at least one other panel unit-root test before drawing policy conclusions.
Strengths & limitations
- No lag-length or kernel bandwidth selection needed, eliminating a common source of specification error.
- Asymptotically standard normal distribution simplifies inference without special critical-value tables.
- Higher local power than bias-corrected tests such as Levin-Lin-Chu under a homogeneous alternative.
- Computationally simple to implement given its reliance on straightforward OLS on transformed series.
- Assumes a common (homogeneous) autoregressive parameter, making it inappropriate for heterogeneous panels.
- Requires a balanced panel; missing observations in any cross-sectional unit invalidate the pooled regression.
- Cross-sectional independence is a maintained assumption; violations lead to size distortions.
- Consistent only under the joint limit N, T → ∞; inference may be unreliable in very short panels.
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 Breitung test differ from the Levin-Lin-Chu test?
Both impose a homogeneous autoregressive parameter and use pooled OLS, but Levin-Lin-Chu applies bias-correction terms that depend on lag augmentation and kernel estimates of long-run variance. Breitung's transformation removes nuisance parameters without these corrections, avoiding the associated variance inflation and achieving higher local power, at the cost of requiring strict cross-sectional independence.
Can the Breitung test be used with unbalanced panels?
The standard implementation requires a balanced panel because the partial-sum transformation accumulates observations across a fixed time index for each unit. Applying it naively to unbalanced data invalidates the asymptotic normality result. Researchers facing missing observations typically balance the panel by trimming or imputation before proceeding, though this introduces its own risks.
What should I do if cross-sectional dependence is suspected?
The Breitung test is a first-generation test and will exhibit severe size distortions under cross-sectional dependence. In that situation, second-generation tests—such as the Pesaran (2007) CIPS test or the Moon-Perron test—should be used instead, as they are designed to account for common factors driving cross-sectional correlation.
Sources
- 1.Breitung, J. (2000). The local power of some unit root tests for panel data. Advances in Econometrics, 15, 161–177.
You have read it. What now?
Cite this page
ScholarGate. (2026, June 2). Breitung Test. ScholarGate. https://scholargate.app/econometrics/breitung-test