Super-Efficiency Data Envelopment Analysis

Inefficient DMUs receive scores identical to their standard DEA scores, which are below 1.0. The procedure only changes outcomes for units that were on the efficient frontier. Removing an inefficient unit from the reference set does not alter the feasible region in a way that changes its score, so standard and super-efficiency results coincide for inefficient observations.

Under variable returns to scale, the convexity constraint requires that intensity weights sum to one. When the evaluated unit is removed and its output levels lie beyond what any convex combination of remaining DMUs can produce, the output constraint becomes infeasible. This is most common for units with extreme scale characteristics, such as the very smallest or very largest DMU in the sample.

Cross-efficiency DEA evaluates each DMU using the optimal weights derived from every other DMU's LP, then averages those peer-evaluated scores. Super-efficiency DEA uses each DMU's own LP but excludes the unit from its reference set. Cross-efficiency avoids infeasibility and often yields more stable rankings, while super-efficiency preserves the original LP structure and is more directly interpretable as a frontier distance measure.