Lumsdaine-Papell Unit-Root Test with Two Structural Breaks
Lumsdaine-Papell Unit-Root Test with Two Breaks · Also known as: LP Test, Two-Break Unit-Root Test, Double Structural Break Unit-Root Test, Lumsdaine-Papell İki Kırılmalı Birim Kök Testi
The Lumsdaine-Papell test, introduced by Robin Lumsdaine and David Papell in 1997, extends the Zivot-Andrews single-break unit-root test to allow for two simultaneous structural breaks in the intercept and/or linear trend of a time series. It is widely used in macroeconomics and finance when data are suspected to have experienced two major regime shifts — such as policy changes, financial crises, or wars — and the researcher needs to determine whether the series is nonetheless integrated of order one.
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
Apply the Lumsdaine-Papell test when theory or visual inspection suggests a time series may have experienced two discrete structural shifts in level or trend — for example, a macroeconomic aggregate spanning wars, oil shocks, or major policy reforms. The series should be sufficiently long (typically T > 100) to support reliable break-date estimation with adequate trimming. The test assumes breaks affect only the deterministic component; if breaks alter the error variance or if more than two breaks are plausible, consider the Lee-Strazicich test (which is also endogenous-break but does not spuriously over-reject) or the Bai-Perron multiple-break framework. It is inappropriate for panels or when the break locations are known a priori.
Strengths & limitations
- Extends Zivot-Andrews by simultaneously allowing two endogenous breaks, reducing omitted-break bias in integrated time series
- Three model variants (AA, AB, CC) provide flexibility to match the theoretical nature of each structural shift
- Endogenous break-date selection avoids data-mining bias associated with pre-specifying break locations
- Widely tabulated critical values and readily available software implementations (EViews, Stata, R) facilitate reproducibility
- Allows at most two structural breaks; economic histories with three or more regime shifts require alternative approaches
- Asymptotic critical values depend on the break-fraction pair, and interpolation between tabulated values introduces approximation error
- Like Zivot-Andrews, it can spuriously reject the unit-root null when breaks are present under the null (i.e., a unit-root process that itself jumps), a problem addressed by the Lee-Strazicich test
- Performance degrades in small samples (T < 80) and when breaks occur near the endpoints of the sample due to trimming constraints
Frequently asked
How does the Lumsdaine-Papell test differ from the Zivot-Andrews test?
Zivot-Andrews allows a single endogenous structural break in level and/or trend, while Lumsdaine-Papell accommodates two simultaneous breaks. When a series has actually experienced two regime shifts, the single-break test may fail to reject the unit root because one unmodelled break distorts the ADF regression. The LP test mitigates this omitted-break bias at the cost of requiring longer series and using separate, two-break-specific critical values.
Why does the Lumsdaine-Papell test sometimes reject the unit root when the Lee-Strazicich test does not?
The LP test conditions on breaks only under the alternative of stationarity, meaning the break-selection procedure can yield spurious rejection when the true process is a unit root that itself contains jumps (breaks under the null). Lee and Strazicich (2003) derived a test that explicitly allows breaks under both null and alternative, producing more conservative — but size-correct — inference. Over-rejection by LP in such scenarios is a known finite-sample issue.
How should I choose between models AA, AB, and CC?
Model AA permits two intercept shifts only and is appropriate when breaks alter the mean level but leave trend growth intact. Model CC adds trend-slope changes to intercept shifts and is suited to series where breaks permanently alter the growth rate. Model AB is a mixed form. The choice should be guided by economic theory and prior graphical analysis of the series, not by selecting the model that yields the most favourable test outcome, as the latter inflates size.
Sources
- Lumsdaine, R. L., & Papell, D. H. (1997). Multiple trend breaks and the unit-root hypothesis. Review of Economics and Statistics, 79(2), 212–218. DOI: 10.1162/003465397556791 ↗
How to cite this page
ScholarGate. (2026, June 2). Lumsdaine-Papell Unit-Root Test with Two Breaks. ScholarGate. https://scholargate.app/en/econometrics/lumsdaine-papell-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.
- Bai-Perron TestEconometrics↔ compare
- Lee-Strazicich TestEconometrics↔ compare
- Zivot-Andrews TestEconometrics↔ compare