Robust Fixed Effects Model
Also known as: FE with robust standard errors, cluster-robust fixed effects, fixed effects with heteroscedasticity-robust SE, within estimator with robust inference
The robust fixed effects model combines the within-group estimator for panel data with variance-covariance matrices that remain valid under heteroscedasticity and within-unit error correlation. Introduced by Arellano (1987), cluster-robust standard errors paired with the fixed effects estimator are now the default approach for credible panel data inference in economics and social science.
Key highlights
- Controls for all time-invariant unobserved heterogeneity without modelling it, providing identification from within-unit variation only.
- Cluster-robust standard errors remain valid under arbitrary within-unit serial correlation and heteroscedasticity — no assumption on the error covariance structure is required.
- Consistent even when unobserved effects are correlated with the regressors, unlike the random effects estimator.
- Widely accepted as the default specification in peer-reviewed applied econometrics; output is directly interpretable as causal effects if selection on observables holds within units.
- Available in all major software packages (Stata, R plm/lfe, Python linearmodels) with a single option for cluster-robust SEs.
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 the robust fixed effects model when you have panel data with multiple observations per unit over time and you want to control for unobserved time-invariant confounders while also protecting inference against within-unit error serial correlation or heteroscedasticity. It is the standard choice in applied micro-econometrics when a Hausman test rejects random effects, or whenever the within-unit error correlation structure is unknown. Do not rely on this approach when the number of cross-sectional units G is very small (G < 20), because cluster-robust standard errors lose reliability; in those cases prefer feasible GLS, bootstrapped standard errors, or a Driscoll-Kraay estimator. If individual-specific effects are uncorrelated with the regressors and the sample is large, random effects GLS is more efficient.
Strengths & limitations
- Controls for all time-invariant unobserved heterogeneity without modelling it, providing identification from within-unit variation only.
- Cluster-robust standard errors remain valid under arbitrary within-unit serial correlation and heteroscedasticity — no assumption on the error covariance structure is required.
- Consistent even when unobserved effects are correlated with the regressors, unlike the random effects estimator.
- Widely accepted as the default specification in peer-reviewed applied econometrics; output is directly interpretable as causal effects if selection on observables holds within units.
- Available in all major software packages (Stata, R plm/lfe, Python linearmodels) with a single option for cluster-robust SEs.
- The within-group transformation eliminates all time-invariant regressors — coefficients on race, gender, country of origin, or other fixed characteristics cannot be estimated.
- Cluster-robust standard errors understate sampling uncertainty when the number of clusters G is small (typically G < 30–50), leading to spurious rejections.
- Requires a sufficiently long panel (at least T ≥ 3) for the within transformation to leave meaningful variation; very short panels (T = 2) give very few degrees of freedom.
- Does not address endogeneity arising from time-varying confounders; instrumental variables or difference-in-differences approaches are needed in those cases.
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
What is the difference between heteroscedasticity-robust and cluster-robust standard errors in a fixed effects model?
Heteroscedasticity-robust (HC) standard errors allow error variance to differ across observations but still assume independence across observations within the same unit. Cluster-robust standard errors allow both heteroscedasticity and arbitrary correlation of errors within the same cross-sectional unit over time. In panel data, within-unit serial correlation is nearly universal, so cluster-robust SEs are almost always the appropriate choice.
How do I choose between fixed effects and random effects?
The Hausman test compares the two: if unobserved unit effects are correlated with the regressors, random effects is inconsistent and fixed effects should be used. In practice, applied economists routinely default to fixed effects for credibility, and then apply cluster-robust standard errors to handle serial correlation.
What should I do if I have very few clusters (G < 30)?
Standard cluster-robust standard errors become unreliable with few clusters (G < 30). Use the wild cluster bootstrap (Cameron, Gelbach, and Miller 2008) or the CR2 bias-corrected variance estimator to obtain better-sized tests. Some researchers also consider permutation tests or analytical small-sample adjustments.
Can I include time fixed effects alongside unit fixed effects?
Yes. Two-way fixed effects (unit and time dummies) control for both time-invariant unit heterogeneity and common time shocks. With two-way FE, cluster-robust standard errors should still be computed, and you may cluster at the unit level or, if cross-sectional dependence is also a concern, use Driscoll-Kraay standard errors instead.
Does the robust fixed effects model handle endogeneity from time-varying confounders?
No. It only removes time-invariant endogeneity. If a time-varying unobserved variable is correlated with both the regressor and the outcome, estimates remain biased. Instrument variables (IV-FE) or a well-designed difference-in-differences design with parallel trends is needed in those cases.
Sources
- 1.Arellano, M. (1987). Computing robust standard errors for within-groups estimators. Oxford Bulletin of Economics and Statistics, 49(4), 431–434.
- 2.Wooldridge, J. M. (2010). Econometric Analysis of Cross Section and Panel Data (2nd ed.). MIT Press.ISBN 978-0262232586
You have read it. What now?
Cite this page
ScholarGate. (2026, June 3). Robust Fixed Effects Model. ScholarGate. https://scholargate.app/econometrics/robust-fixed-effects-model