Testing Scale Efficiency in DEA Models: A Smooth Bootstrap Approach
Hédi Essid, Pierre Ouellette, Stéphane Vigeant

Abstract
This paper discusses and implements a nonparametric statistical test procedure for organization scale efficiency. This procedure allows the practitioner to test whether the observed scale efficiency score is real or due to sampling variation. Our test procedure is based on the work of Simar and Wilson (1998, 2002). However, instead of considering a global test to give an average measure of the efficiency of a sector of activity, we consider an individual test which gives a measure of the scale efficiency for each organization. We believe that this type of analysis is the most used in empirical studies. The test is simulated in a Monte Carlo study and is shown to have the correct size in most cases. In contrast, the shape of the estimated power function is not expected. This can be explained by the “curse of dimensionality” problem which is often raised in this context.

Full Text: PDF