A modified super-efficiency in the range directional model

The range directional model (RDM) relaxes the assumption of non-negativity of inputs and outputs in the conventional data envelopment analysis (DEA) model with the aim of evaluating the efficiency of a decision-making unit (DMU) when some data are negative. Although the concept of super-efficiency in the RDM contributes to enhancing discriminatory power, the formulated model may lead to the infeasibility problem for some efficient DMUs. In this paper, we modify the super-efficiency RDM (SRDM) model to overcome the infeasibility problem occurring in such cases. Our method leads to a complete ranking of the DMUs with negative data for yielding valuable insights that aid decision makers to better understand the findings from a performance evaluation process. The contribution of this paper is fivefold: (1) we detect the source of infeasibility problems of SRDM in the presence of negative data, (2) the proposed model in this study yields the SRDM measures regardless of feasibility or infeasibility of the model, (3) when feasibility occurs, the modified SRDM model results in the scores that are the same as the original model, (4) we differentiate the efficient units to improve discriminatory power in SRDM, and (5) we provide two numerical examples to elucidate the details of the proposed method.