Introducing and modeling inefficiency contributions

Whilst Data Envelopment Analysis (DEA) is the most commonly used non‐parametric benchmarking approach, the interpretation and application of DEA results can be limited by the fact that radial improvement potentials are identified across variables. In contrast, Multi‐directional Efficiency Analysis (MEA) facilitates analysis of the nature and structure of the inefficiencies estimated relative to variable‐specific improvement potentials. This paper introduces a novel method for utilizing the additional information available in MEA. The distinguishing feature of our proposed method is that it enables analysis of differences in inefficiency patterns between subgroups. Identifying differences, in terms of which variables the inefficiency is mainly located on, can provide management or regulators with important insights. The patterns within the inefficiencies are represented by so‐called inefficiency contributions, which are defined as the relative contributions from specific variables to the overall levels of inefficiencies. A statistical model for distinguishing the inefficiency contributions between subgroups is proposed and the method is illustrated on a data set on Chinese banks.