Robust efficiency evaluation of ban k branch performance: the case of Banco do Brasil

Abstract

We estimate DEA technical efficiency for 3663 branches of the Banco do Brasil, by combining Bootstrap and Jackknife techniques. Moreover, bank efficiency is affected by many factors outside the control of the bank branches such as socio-economic environment. To account for these factors, in the second stage we use regression methods. Our first stage results reveal that outliers (super efficient branches) have a huge number of clients, and/or deal mainly on the agribusiness sector. In the second stage, quantile regression results show that efficiency is positively related to the size of the branch and to per capita income of the clients. We also show that there is no evidence that competition boosts efficiency, as indicated by the literature on bank branch efficiency. Finally a good workplace environment clearly contributes to increase the branch efficiency.