Problems of using grouped DMUs for efficiency measurement: Monte Carlo experiments, empirical dimension, and a correction procedure

This paper explores the consequences for parametric and non-parametric efficiency levels and rankings when using grouped instead of individual Decision Making Units (DMU). The bias results due to the differences of the grouped DMUs frontier compared to the individual DMUs frontier. Monte Carlo experimentation is used to evaluate the empirical dimension on the estimated efficiency levels and rankings. These results are illustrated with an empirical example using a sample of German farms. The bias in ranking is found to be substantial. Finally, a correction procedure is developed to improve the results when only grouped data are available.