A study of size efficiency in US banking: Identifying banks that are too large

Maindiratta questioned the practical relevance of the most productive scale size (MPSS) and the associated concept of scale efficiency on the grounds that the observed output bundle of a firm or decision making unit (DMU), in many situations, is an assigned task and cannot be altered by the DMU. Also, in a market economy, the output–input combinations at the MPSS may not be economically viable. It would still be of interest to examine whether greater input saving can be achieved if the assigned output bundle is produced collectively by several banks, each operating efficiently, rather than individually by a single bank operating efficiently. In any specific case the optimal number of smaller banks – if an existing bank should be broken up at all – is determined within a mixed integer programming model. In this study we apply DEA, utilizing the input and output data for the years 1984–1990 from 201 banks each with assets in excess of 1 billion dollars and hence regarded as ‘large banks’ by asset-size classification. Although many banks operate in the region of diminishing returns, very few of them are found to be candidates for break up. Of those that are considered to be ‘too large’ some are large enough to be broken up into 13 smaller units or more. The mixed integer programming is also solved for the aggregate input–output vector of the banks in the sample to determine the output bundle of a representative bank for each year. Econometric analysis of findings shows that size efficiency of any observed bank declines with total assets but increases with product specialization.