Andersen and Petersen (1993) presented an extension of the basic DEA methodology, called modified DEA, which has the desirable feature of ranking not only the inefficient DMUs, but the e ficient ones as well. However, when their basic approach is extended to the cases of variable returns to scale and non-discretionary inputs, several conceptual problems arise. This paper addresses these problems, and illustrates the proposed extensions to the modified DEA method using data from a major U.S. bank.