Exact and stochastic methods for robustness analysis in the context of Imprecise Data Envelopment Analysis

Abstract

We consider the problem of measuring the efficiency of decision-making units with a ratio-based model. In this perspective, we introduce a framework for robustness analysis that admits both interval and ordinal performances on inputs and outputs. The proposed methodology exploits the uncertainty related to the imprecise data and all feasible input/output weight vectors delimited through linear constraints. We offer methods for verifying the robustness of three types of outcomes: efficiency scores, efficiency preference relations, and efficiency ranks. On the one hand, we formulate mathematical programming models to compute the extreme, necessary, and possible results. On the other hand, we incorporate the stochastic analysis driven by the Monte Carlo simulations to derive the probability distribution of different outcomes. The framework is implemented in R and made available on open-source software. Its use is illustrated in two case studies concerning Chinese ports or industrial robots.