Panel ARDL Bounds Test
Panel Autoregressive Distributed Lag Bounds Testing Approach · Also known as: Panel ARDL, Panel bounds testing, Panel ARDL cointegration, Panel PSS bounds test
The Panel ARDL Bounds Test extends the Pesaran, Shin and Smith (2001) bounds testing procedure to panel data, allowing researchers to test for long-run cointegrating relationships between variables without requiring all series to be integrated of the same order. It is widely used in macro-panel studies where variables may be I(0), I(1), or a mixture of both.
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 ARDL Bounds Test when you have a balanced or mildly unbalanced macro panel (moderate T, moderate N) and the integration orders of your variables are uncertain or mixed I(0)/I(1). It is especially suitable for growth, energy, trade, and finance studies where theory predicts a long-run equilibrium. Do not use it when T is very short (fewer than 15–20 time periods per unit) because the asymptotic critical values require sufficient time-series variation. Avoid it when cross-sectional dependence is strong and uncorrected, as this inflates test size; in that case, use second-generation panel unit root and cointegration procedures first.
Strengths & limitations
- Does not require pre-testing for unit roots or a priori knowledge of integration orders, handling I(0), I(1), or mixed series.
- Pooled mean-group estimation produces efficient and interpretable long-run homogeneity alongside heterogeneous short-run dynamics.
- Jointly estimates short-run dynamics and long-run equilibrium in a single coherent error-correction framework.
- More powerful than single-equation ARDL when panel dimension N is moderate, by pooling cross-sectional information.
- Flexible lag structure selection (AIC, BIC) reduces misspecification risk.
- Requires a sufficiently long time dimension (T >= 15–20 per unit); with very short T, critical values are unreliable.
- PMG's long-run slope homogeneity constraint may be inappropriate if true slopes differ substantially across units; a Hausman test of MG vs PMG should be performed.
- Standard critical values assume cross-sectional independence; uncorrected cross-sectional dependence distorts test size.
- Cannot handle I(2) variables — series must be tested and confirmed to be at most I(1) before proceeding.
Frequently asked
What is the difference between PMG and MG estimators in Panel ARDL?
The pooled mean-group (PMG) estimator restricts long-run slope coefficients to be equal across panel units while allowing short-run dynamics and error variances to differ; it is more efficient if homogeneity holds. The mean-group (MG) estimator allows full heterogeneity in all parameters. A Hausman test discriminates between them: if PMG restrictions are valid, PMG is preferred for efficiency.
What if my F-statistic falls between the lower and upper critical value bounds?
The result is inconclusive. You must determine the integration orders of your variables (e.g. via panel unit root tests). If all variables are confirmed I(1), compare against the upper-bound critical value; if all are I(0), use the lower bound.
Can Panel ARDL handle I(2) variables?
No. The bounds testing framework is derived under the assumption that variables are at most I(1). Including I(2) variables invalidates the asymptotic critical values. Always confirm that no variable is I(2) before applying the test.
How should I choose lag lengths p and q?
Select lag lengths individually for each cross-sectional unit using AIC or BIC, subject to a maximum lag chosen based on data frequency (e.g. max 4 for annual data, max 12 for monthly). Ensure residuals are free of serial correlation after selection.
Does Panel ARDL require a balanced panel?
The estimators can accommodate mildly unbalanced panels, but the time dimension T must be large enough for each unit to support ARDL estimation. Very uneven T across units can affect pooled inference.
Sources
- Pesaran, M. H., Shin, Y., & Smith, R. J. (2001). Bounds testing approaches to the analysis of level relationships. Journal of Applied Econometrics, 16(3), 289–326. DOI: 10.1002/jae.616 ↗
- Pesaran, M. H., & Pesaran, B. (1997). Working with Microfit 4.0: Interactive Econometric Analysis. Oxford University Press. link ↗
How to cite this page
ScholarGate. (2026, June 3). Panel Autoregressive Distributed Lag Bounds Testing Approach. ScholarGate. https://scholargate.app/en/econometrics/panel-ardl-bounds-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.
- Nonlinear ARDLEconometrics↔ compare
- Panel Engle-Granger CointegrationEconometrics↔ compare
- Panel Granger CausalityEconometrics↔ compare
- Panel Johansen CointegrationEconometrics↔ compare
- Panel NARDLEconometrics↔ compare
- Panel VECMEconometrics↔ compare