Impact of an input-output specification on efficiency scores in data envelopment analysis: A banking case study

Abstract

The paper stresses the importance of making an appropriate specification of inputs and outputs in technical efficiency measurement and provides empirical evidence that this initial step of an efficiency measurement project should not be underestimated. Oriented on a case study of Slovak commercial banks for the period from 2005 to 2016, the paper explores to what extent different input-output specifications affect the comparability or congruence of technical efficiency scores in a banking application produced by four different data envelopment models differing in the efficiency measure and orientation. Building on the long-standing controversy in the banking literature about the most appropriate description of banking production, the paper compares technical efficiency scores for 9 input-output specifications of the intermediation approach, 9 specifications of the production-like approaches and 3 network integrated specifications. All these specifications were empirically applied earlier in the literature. The efficiency scores produced by different input-output specifications and models are confronted by six measures of association or dependence, and their levels are explained in a regression framework. The results attest that the choice of the input-output set is a critical judgemental input to efficiency measurement since there is vast diversity in efficiency scores of input-output sets coming from different approaches but also for input-output sets associated with the same approach. In addition, intermediation input-output specifications tend to produce higher efficiency scores than production-like specifications.