Title: Statistical inference § parametric approximation of non-parametric frontier: the case of Tunisian banking sector

Authors: Boutheina Bannour; Moez Labidi

Addresses: Applied Economics and Simulation (AES), Faculty of Economics and Management at Mahdia, University of Monastir, Avenue Habib Bourguiba, Hiboun Mahdia, 5111, Tunisia ' Applied Economics and Simulation (AES), Faculty of Economics and Management at Mahdia, University of Monastir, Sidi Messaoud – Mahdia, 5111, Tunisia

Abstract: The aim of this article is to present a statistical inference for a frontier non-parametric model. We find that DEA and parametric approaches to estimate the efficiency and productivity share a common weakness: the inability to determine a statistical accuracy of the results. In the case of parametric approach and due to the highly nonlinear combination, the efficiency scores are calculated from global estimates. As far as DEA is concerned and because of the non-parametric aspect of the method in use, the distribution of efficiency measures is neither known nor specified. In the same vein, the absence of an indicator of statistical significance undermines the reliability and usefulness of the results. Consequently, we simulated the efficiency scores obtained by estimating the DEA model (input-oriented and variable returns to scale) of 20 Tunisian banks using the bootstrap method.

Keywords: bootstrap; mathematical programming; non-parametric frontier; parametric approximation; operational research; Tunisia; banks; banking industry; statistical inference; modelling; DEA; data envelopment analysis; simulation.

DOI: 10.1504/IJMOR.2015.071272

International Journal of Mathematics in Operational Research, 2015 Vol.7 No.5, pp.485 - 518

Received: 02 Jan 2013
Accepted: 26 Sep 2013
Published online: 19 Aug 2015 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article