Augmented Mean Group (AMG) Estimator
Also known as: AMG estimator, augmented mean group, Artırılmış Ortalama Grup Tahmincisi (AMG)
The Augmented Mean Group estimator, developed by Eberhardt and Teal (2010), is a panel data method for estimating heterogeneous slope coefficients in the presence of cross-sectional dependence. It approximates the unobserved common dynamic process driving all units and folds it into unit-by-unit regressions, then averages the results.
Key highlights
- Handles cross-sectional dependence by explicitly modelling an unobserved common dynamic process.
- Allows fully heterogeneous slope coefficients across units rather than imposing a single pooled slope.
- Provides a transparent, two-stage alternative to the CCE Mean Group estimator.
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 AMG with macro panels where the time dimension T and the cross-section N are both reasonably large (at least about 50 observations overall, and avoid it when T is below 10), the slope relationship plausibly differs across units, and there is cross-sectional dependence (units share common shocks) — ideally confirmed beforehand with a Pesaran CD test. It is a natural alternative to the CCE Mean Group estimator for this setting.
Strengths & limitations
- Handles cross-sectional dependence by explicitly modelling an unobserved common dynamic process.
- Allows fully heterogeneous slope coefficients across units rather than imposing a single pooled slope.
- Provides a transparent, two-stage alternative to the CCE Mean Group estimator.
- Unreliable when the time dimension is short (T < 10); the unit-by-unit regressions and the common process estimate become noisy.
- Requires reasonably large T and N, so it is unsuited to short or narrow panels.
- Assumes the common factor structure is adequately proxied by the common dynamic process; misspecification of that process biases results.
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 AMG differ from the CCEMG estimator?
Both target heterogeneous slopes under cross-sectional dependence. CCEMG controls for common factors by adding cross-sectional averages of the variables to each unit's regression, whereas AMG explicitly extracts a common dynamic process from pooled year dummies and augments each unit's regression with it. AMG is offered as a direct alternative to CCEMG.
What is the 'common dynamic process' in AMG?
It is a single time-varying series, recovered from the estimated year dummies of a first-stage pooled regression, that proxies the unobserved common factors shared by all units. Adding it to each unit-level regression is the 'augmentation' that gives the method its name.
Why does the time dimension T matter so much?
AMG fits a separate regression for every unit and estimates a common process over time, so it needs enough time periods to do this reliably. When T is below about 10 the estimator is considered untrustworthy, and reasonably large T and N are required overall.
Do I need to test for cross-sectional dependence first?
Yes. AMG is built for panels where units share common shocks, so you should confirm that cross-sectional dependence is actually present, typically with a Pesaran CD test, before relying on the method.
Sources
- 1.Eberhardt, M. & Teal, F. (2010). Productivity Analysis in Global Manufacturing Production. Economics Series Working Papers, No. 515, University of Oxford.
- 2.Bond, S. & Eberhardt, M. (2013). Accounting for Unobserved Heterogeneity in Panel Time Series Models. Nuffield College Discussion Paper.
You have read it. What now?
Cite this page
ScholarGate. (2026, June 1). Augmented Mean Group Estimator. ScholarGate. https://scholargate.app/econometrics/amg-estimator