Stochastic frontier analysis
Separate modelled inefficiency from random production noise
Stochastic frontier analysis estimates a parametric production or cost frontier with a composed error: a symmetric noise term and a one-sided inefficiency component. Unlike standard DEA, it allows random shocks around the frontier. Separating noise from inefficiency requires functional-form and distributional assumptions, and production and cost frontiers use different signs and economic restrictions.
Choose this for efficiency research when measurement noise or random shocks are important and a credible production or cost function can be specified. The data must identify the frontier and its error components.
Strengths
- Distinguishes statistical noise from inefficiency under the model
- Allows formal parametric estimation and hypothesis testing
Limitations
- Efficiency rankings depend on functional and distributional assumptions
- Misspecification can be mistaken for inefficiency
Know the boundary
Estimated inefficiency is model-dependent; the statistical decomposition does not observe managerial waste directly.