Stochastic Frontier Analysis (SFA)
Stochastic Frontier Production Function Analysis · Also known as: SFA, stochastic frontier model, stochastic production frontier, Stokastik Sınır Analizi (SFA)
Stochastic Frontier Analysis is a frontier regression model, introduced by Aigner, Lovell and Schmidt in 1977, that estimates a production, cost, or profit function while separating each unit's technical inefficiency from ordinary statistical noise. It splits the error term into a symmetric random component and a one-sided inefficiency component, producing firm- or country-level efficiency scores.
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 SFA when you want to estimate a production, cost, or profit frontier and measure how far individual firms, farms, hospitals, or countries fall short of best practice, on continuous output data with a reasonable sample (about 50 or more units). It suits cross-sectional and panel data and is appropriate when you believe part of the deviation from the frontier is genuine random noise rather than pure inefficiency. It assumes a one-sided distribution for the inefficiency term, symmetric normal noise, and a correctly specified functional form such as Cobb-Douglas or Translog. For panels, a time-varying inefficiency specification is preferred.
Strengths & limitations
- Separates genuine technical inefficiency from random statistical noise, unlike deterministic envelope methods.
- Produces interpretable firm- or country-level efficiency scores from a single estimated model.
- Applies to production, cost, and profit functions, and to both cross-sectional and panel data.
- Results depend on the assumed distribution of the inefficiency term (half-normal, truncated-normal, exponential) and on a correctly specified functional form.
- Misspecifying the frontier (e.g. wrong Cobb-Douglas vs. Translog choice) biases the efficiency scores.
- Needs a reasonable sample (about 50 or more units) for the maximum-likelihood variance decomposition to be reliable.
Frequently asked
How does SFA differ from Data Envelopment Analysis (DEA)?
DEA builds a deterministic envelope and treats every deviation from the frontier as inefficiency. SFA is statistical: it splits the deviation into a symmetric random noise term and a one-sided inefficiency term, so measurement error and bad luck are not counted as inefficiency.
Why is the error term split into two parts?
The composed error v − u lets the model distinguish ordinary statistical noise v, which can move a unit either side of the frontier, from technical inefficiency u, which can only pull a unit below it. This is the core idea behind the Aigner-Lovell-Schmidt (1977) formulation.
Which distribution should I assume for inefficiency?
Common choices are half-normal, truncated-normal, and exponential. The truncated-normal is flexible and underlies the Battese-Coelli panel model. Whatever you pick, check that the efficiency rankings are reasonably robust to the assumption.
Can I use SFA with panel data?
Yes. Panel SFA is generally preferred because it can let inefficiency vary over time. The Battese and Coelli (1995) model for technical inefficiency effects is a standard specification for this.
Sources
- Aigner, D., Lovell, C.A.K. & Schmidt, P. (1977). Formulation and Estimation of Stochastic Frontier Production Function Models. Journal of Econometrics, 6(1), 21–37. DOI: 10.1016/0304-4076(77)90052-5 ↗
- Battese, G.E. & Coelli, T.J. (1995). A Model for Technical Inefficiency Effects in a Stochastic Frontier Production Function for Panel Data. Empirical Economics, 20(2), 325–332. DOI: 10.1007/BF01205442 ↗
How to cite this page
ScholarGate. (2026, June 1). Stochastic Frontier Production Function Analysis. ScholarGate. https://scholargate.app/en/econometrics/stochastic-frontier
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.
- OLS RegressionEconometrics↔ compare
- Panel Fixed EffectsEconometrics↔ compare
- Quantile RegressionEconometrics↔ compare