Pooled Mean Group (PMG) Estimator
Pooled Mean Group Estimator · Also known as: PMG Estimator, Pooled Mean Group, PMG Panel Estimator, Havuzlanmış Ortalama Grup Tahmincisi
The Pooled Mean Group (PMG) estimator, introduced by Pesaran, Shin, and Smith (1999), is a panel data technique designed for dynamic heterogeneous panels where the long-run equilibrium relationship is common across groups but short-run dynamics and error variances are allowed to differ. It is particularly suited for macro-panels with moderate N and T, such as cross-country growth, energy consumption, and financial development studies.
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 PMG when working with macro-panels where both N and T are moderate to large (e.g., N = 10–60, T = 20–50), variables are non-stationary and cointegrated, and there is theoretical reason to expect a common long-run relationship (e.g., purchasing power parity, balanced growth). The method assumes: (1) variables are I(1) or cointegrated, (2) long-run coefficients are homogeneous, (3) short-run dynamics are heterogeneous. If the Hausman test rejects long-run homogeneity, prefer the Mean Group estimator. Avoid PMG with purely cross-sectional data, stationary panels, or very short T.
Strengths & limitations
- Accommodates heterogeneous short-run dynamics and error variances across groups while imposing an efficient common long-run relationship
- More efficient than the Mean Group estimator when long-run slope homogeneity holds, especially in panels with limited T
- Explicitly models error-correction dynamics, providing both long-run elasticities and group-specific adjustment speeds
- Formally testable against the MG estimator via the Hausman test, giving a principled model selection criterion
- Long-run homogeneity is a strong assumption that may not hold in heterogeneous country or firm panels; violations render PMG inconsistent
- Requires both N and T to be sufficiently large for maximum likelihood estimation to be reliable in each group
- Does not correct for cross-sectional dependence by default; in the presence of common factors, augmented versions (CS-PMG) are needed
- Sensitivity to lag order selection (p, q) in the underlying ARDL specification can affect results
Frequently asked
How does PMG differ from the Mean Group (MG) estimator?
MG estimates separate regressions for each group and averages the coefficients, allowing full heterogeneity in both short- and long-run parameters. PMG restricts the long-run coefficients to be equal across groups while still allowing heterogeneous short-run dynamics, which yields more efficient estimates when homogeneity holds but is inconsistent if it does not. The Hausman test discriminates between the two.
What panel structure is required for PMG?
PMG requires a balanced or near-balanced macro-panel where the variables are integrated of order one, I(1), and cointegrated within each group. Both N and T should be reasonably large — typically T > 20 — so that group-specific ARDL models can be estimated reliably before pooling. Very short T panels should use alternative estimators.
Can PMG handle cross-sectional dependence?
Standard PMG does not account for cross-sectional dependence arising from common factors or global shocks. In such settings, residuals across groups will be correlated, leading to size distortions. Researchers should apply cross-sectional dependence tests (e.g., Pesaran CD test) first, and if dependence is detected, use extensions such as the Common Correlated Effects (CCE) pooled estimator instead.
Sources
- Pesaran, M. H., Shin, Y., & Smith, R. P. (1999). Pooled mean group estimation of dynamic heterogeneous panels. Journal of the American Statistical Association, 94(446), 621–634. DOI: 10.1080/01621459.1999.10474156 ↗
How to cite this page
ScholarGate. (2026, June 2). Pooled Mean Group Estimator. ScholarGate. https://scholargate.app/en/econometrics/pooled-mean-group
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.
- ARDL Bounds TestEconometrics↔ compare
- Panel Fixed EffectsEconometrics↔ compare