A Two-Staged Benchmarked Decision Support System Using DEA Profiles of Efficiency

Using the BvD BankScope (TM) database through Wharton Research Data Services (TM), we identified for 2002 through 2005 all of the US national banks listed on the NY or NASDAQ stock exchanges. This yielded for each year about 120 banks. For each year, we categorized these banks into three size groups based upon total assets. We then: (1) developed for each year using the standard CCR DEA analysis those banks that were CCR efficient, (2) using variables suggested in the literature as being important in characterizing the relative performance of banks, we developed profiles of the differences between the efficient and relatively non-efficient banks for each of the three size categories by year. This was the first stage in the DEA profiling. For the second stage, again for each year by size grouping, we: (1) calculated the Super-Efficiency [SE] scores as proposed by Andersen and Petersen (1993) for the set of CCR efficient banks, (2) developed High and Low SE groups using a median-split of these Super-Efficiency scores, and (3) profiled these SE-High and SE-Low groups. Results: we: (1) developed and illustrated a simple DEA DSS heuristic that could be used by decision makers to identify the driver variables that may be acted upon to manage their risk by moving their banks into their target efficiency group, (2) demonstrated that size is an important category variable in understanding the profiled performance of banks, (3) determined that the Super-Efficiency profiles are refinements of the CCR categorization, and (4) found that there are size-related stationarity differences among the banks which have risk implications for the various size groupings.