A general method of deriving the efficiencies of banks from a profit function

Abstract

Questions of whether the evolution of the financial services industry results in more efficient intermediaries, better prices and service quality for consumers, and greater bank safety and soundness cannot be answered without addressing the cost and revenue efficiencies of the industry. Most studies of the efficiency of financial institutions have used an econometric approach to measure efficiency. This paper tries to address potential econometric problems of previous efficiency studies and suggests a new technique for measuring efficiency. This method is applied to data on U.S. commercial banks from 1984 through 1989. Previous studies used one of four different approaches for estimating X-efficiencies - the econometric frontier approach, the thick frontier approach, data envelopment analysis, and the distribution free approaches. In all four cases, the econometric problem of estimating X-efficiencies is defined simply as one of distinguishing between two components of a random error term added to a cost or profit function. The authors suggest that it is possible that the actual econometric problem of estimating X-inefficiencies is not as simple as the problem of distinguishing between two random components because of the following three reasons: ( i ) the true functional forms of the cost or profit functions of the firms are usually unknown; ( ii ) explanatory variables excluded from the cost or profit function are likely to be correlated with the explanatory variables included in the function; and (iii) inconsistencies may arise if arbitrary error terms are added to a cost or profit function and their corresponding share equations. The authors use a fixed-coefficients model that allows them to address the econometric problems mentioned above. This methodology also allows for the estimation of a separate frontier for each firm as opposed to previous studies that estimate one frontier which is common to all firms. The results of the paper show that the residual which the previous studies attributed to technical inefficiency potentially included the effects of excluded variables, of inaccuracies in the specified functional forms, and of inconsistent parameter estimates. They further show that once these effects are subtracted from the residual, the measured inefficiencies are substantially reduced. The results do support some of the previous studies' conclusions - in general measured technical inefficiencies dominate allocative inefficiencies and that on average large banks are more efficient from both the technical and allocative perspectives than small and medium sized banks.