Linear Programming and Econometric Methods for Bank Efficiency Evaluation: an Empirical Comparison Based on a Panel of Italian Banks

Abstract

Research on bank efficiency has developed in two separate streams: i) econometric studies, aimed at improving OLS estimates with an asymmetric structure for the residuals; ii) Data Envelopment Analysis, a linear programming model for multiple-input/multiple-output firms. These two branches of literature have developed quickly, but separately; now, the lack of correspondence between the rankings dictated by the two approaches suggests more comparative research. In this paper the two techniques have been tested on a common panel of 270 Italian banks, and this parallel application has suggested the following: i) econometric and linear programming results do not differ dramatically, when based on the same data and conceptual framework; ii) when differences arise, they can be explained by going back to the intrinsic features of the models. Moreover, some findings on Italian banks may be of interest also to the international reader: i) the efficiency scores show a high variance; ii) the banking system is split in two, between northern and southern banks; iii) there is a direct (rather than inverse) relationship between productive efficiency and asset quality; iv) the efficiency of Italian banks did not increase over the period 1988–1992.

I wish to thank Maria Luisa Di Battista, Mario Comana, Marco Oriani, Agnese Sironi, Emmanuel Thanassolulis, Anna Torriero and an anonymous Referee for their precious comments; Marco Corbellini (AssBank, Milan) and Tim Coelli (New England University, Armidale), for providing me, respectively, with the “Bilbank” database and the Frontier 2.0 software, used for ML estimates; Gianfranco Gambarelli for his constant, warm encouragement. This said, I am the only responsible for all remaining errors.