Data envelopment analysis with imprecise data revisited

Wang et al. (2005) proposed a pair of data envelopment analysis (DEA) models to deal with the efficiency assessment of decision-making units (DMU) in the presence of interval input/output data. Their approach was developed with reference to an earlier approach proposed by Despotis and Smirlis (2002) for the same problem. Given that the input/output data are given as interval numbers, the efficiency scores are interval measures as well. In such a setting, both approaches provide lower and upper bounds for the efficiency scores. Wang et al. (2005) claim that the lower and upper bounds calculated in Despotis and Smirlis (2002) are incorrect. Then, they present different models to calculate the true bounds. In this paper, we counter-argue their claim and we show that the Despotis and Smirlis bounds are correct and those provided in Wang et al. are estimated in a manner that they fail to satisfy an obvious property that they should possess. We illustrate our arguments with a counterexample that was originally used in Wang et. al (2005).