Panel DF-GLS
Also known as: Panel unit-root test
Panel DF-GLS extends the Elliott, Rothenberg, and Stock (1996) GLS unit-root test to panel data, combining cross-sectional and time-series information to test whether variables contain unit roots. Introduced by Hadri and colleagues (2005), it is more powerful than standard panel unit-root tests (IPS, LLC) due to its GLS detrending approach. This test is essential for establishing stationarity before fitting cointegration or dynamic panel models.
Key highlights
- Higher power than standard panel unit-root tests via GLS detrending
- Efficiently handles trend and drift removal
- Straightforward interpretation: rejects unit-root null if p-value is small
- Implemented in most econometric software
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 Panel DF-GLS before analyzing long-run relationships or dynamic panel models. It is necessary to establish whether variables are I(0) (stationary), I(1) (integrated order 1), or higher. Particularly valuable when samples have moderate T (20-50 years) and large N (many countries/firms).
Strengths & limitations
- Higher power than standard panel unit-root tests via GLS detrending
- Efficiently handles trend and drift removal
- Straightforward interpretation: rejects unit-root null if p-value is small
- Implemented in most econometric software
- Assumes same lag length across units; heterogeneous lag lengths can bias test
- Power decreases with near-unit-root processes (rho close to 1)
- Cross-sectional dependence across units can invalidate critical values
- Requires sufficient time series length (T > 20) for reliable inference
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 do I determine lag length in Panel DF-GLS?
Use AIC or BIC on preliminary OLS regressions per unit, then take the average or maximum lag across units. Alternatively, use Newey-West bandwidth selection.
What is the null hypothesis?
Null: unit root present (series is I(1)). Alternative: stationarity (I(0)). Rejecting null means evidence for stationarity. Failing to reject means data is consistent with unit root (not conclusive evidence of unit root).
Should I include trend in the test?
Include trend if the series shows clear upward or downward drift (e.g., GDP over decades). Omit if series fluctuates around a constant level. Examine plots first.
What if units have different lag lengths?
Heterogeneous lags can reduce power. Run sensitivity analysis: test with several lag lengths and check robustness of conclusions.
Sources
- 1.Elliott, G., Rothenberg, T. J., & Stock, J. H. (1996). Efficient tests for an autoregressive unit root. Econometric Reviews, 13(4), 469-497.
- 2.Hadri, K., & Larsson, R. (2005). Testing for stationarity in heterogeneous panel data. Econometric Reviews, 24(4), 403-456.
You have read it. What now?
Cite this page
ScholarGate. (2026, June 3). Panel DF-GLS. ScholarGate. https://scholargate.app/econometrics/panel-df-gls