Data envelopment analysis and assurance region profit ratios and sensitivity of 100 large U.S. banks

Several important DEA/AR concepts were applied here to banking for the first time. This application includes classification, sensitivity, uniqueness, linked cones (LCs), and profit ratios. Notably, large bank behavior seems to be explained better by profit ratios than by relative efficiency. Measures of DEA efficiency, AR efficiency, and LC profit ratios were made for a bank panel of the US's 100 largest banks in asset size from 1986 to 1991. High levels of inefficiency were found, as in previous studies. Classification of the DEA efficiency measures identified the inefficient decision-making units (DMUs) with some positive primal slacks. Sensitivity analysis of the DEA efficiency measures showed that the extreme-efficient classification was generally relatively insensitive to errors in the data. The ARs eliminated (i) 44% to 60% of the DEA-extreme-efficient DMUs and (ii) all of the banks with unprofitable actual profit ratios. Some statistical analyses highlight the superiority of the LC profit ratios, relative to the AR efficiency measures.