Data envelopment analysis
Benchmark organisations against an observed production frontier
DEA constructs a non-parametric frontier from observed inputs and outputs and measures each unit relative to feasible combinations of peers. CCR assumes constant returns to scale; BCC permits variable returns. Input orientation asks about proportional input reduction, while output orientation asks about output expansion. Scores are relative to the selected sample and production specification, with deviations ordinarily treated as inefficiency.
Choose DEA when comparing similar banks, branches or organisations with multiple inputs and outputs and no defensible parametric production function. Comparable technologies and carefully justified measures are essential.
Strengths
- Handles multiple inputs and outputs without choosing a parametric frontier
- Identifies reference peers and potential proportional improvements
Limitations
- Sensitive to outliers and measurement error
- Too many inputs and outputs relative to units reduce discrimination
Know the boundary
A DEA-efficient unit lies on the observed frontier; it is not necessarily efficient relative to an unknown best technology.