Global scale efficiency in data envelopment analysis

Abstract In the realm of assessing scale efficiency (SE), it tends to be computed as a firm‐specific phenomenon rather than something associated with the whole shape of the frontier of the technology under evaluation. This circumstance may lead to inaccuracies in the conclusions drawn regarding the returns to scale (RTS) exhibited across the entire frontiers since they can be biased by the location of the observations. This paper addresses this gap by defining the notion of global SE measured through projecting synthetic observations randomly generated within a unit‐hypercube against the technologies under constant and variable RTS. Our approach enables the comparison of the local SE exhibited by each firm in relation to the global expected percentage of production points that display a similar type of RTS across the production frontier. The approach is illustrated through a numerical example and applied to a set of hospitals in Spain.