CORMSIS talk "A fairer assessment of DMUs in a generalised 2-stage DEA structure", by Marios Kremantzis (Southampton) Event

Event details

Data Envelopment Analysis (DEA) is a well-established approach for evaluating the performance of Decision-Making Units (DMUs). Conventional DEA uses single-stage models that ignore the processes inside a DMU. More recent research aims to address this by using 2-stage DEA models to help detect sources of inefficiency. A typical 2-stage DEA structure is serial, whereby the first stage produces intermediate measures which form the inputs to the second stage. This paper puts the spotlight on the 2-stage structure proposed by Yu & Shi (2014) with additional inputs in the second stage and part of intermediate measures as final outputs, which we will henceforth refer to as a “generalised” 2-stage DEA structure. We employ additive-based efficiency measurement and decomposition to formulate a CCR self-evaluation DEA model. The study also contributes to the literature by proposing a Goal Programming-Multiple Criteria DEA (GP-MCDEA) cross-efficiency model as an alternative secondary goal to deal with the non-unique optimal multipliers of the above-mentioned model. The CRiteria Importance Through Inter-criteria Correlation (CRITIC) method is additionally applied to delineate the final weights assigned to the individual cross-efficiencies to obtain the ultimate cross-efficiency scores; this alternatively solves the aggregation problem in cross-efficiency. Finally, three Nash bargaining game common-multipliers DEA models are introduced for a fairer assessment of the DMUs, both in the entire system and its constituent parts. An application illustrates the applicability of the proposed models, as well as that the proposed GP-MCDEA and Nash efficiency models help enhancing the weight scheme and the discriminatory power in this generalised 2-stage DEA system.