A Nerlovian cost inefficiency two-stage DEA model for modeling banks' production process: Evidence from the Turkish banking system. (September 2020)

• Record Type: Journal Article
• Authors: Fukuyama, Hirofumi; Matousek, Roman; Tzeremes, Nickolaos G.
• Is Part Of: Omega. Volume 95(2020)
• Publication Date: 2020-09
• Subjects: Data envelopment analysis (DEA) -- Directional distance function -- Human capital -- Network DEA -- Turkish bank performance; Management -- Periodicals; 658.4005
• DOI: 10.1016/j.omega.2020.102198
• Languages: English
• ISSNs: 0305-0483

Abstract

Abstract: By developing a two-stage model the paper examines the cost inefficiency levels of Turkish banks during the period from 2007 to 2016. Using the notion of Koopmans input efficiency we deconstruct the estimated Nerlovian cost inefficiency to the sum of slack-based allocative and technical inefficiency levels. Alongside the traditional inputs, intermediates and outputs used to model banks' performance measure we additionally use their labor education quality factor as a non-discretionary input. This allows us to model how human capital factors affect a bank's revenue generation stage. Our findings suggest that cost inefficiency levels are driven mainly by a bank's ability to control its allocative inefficiency levels. The empirical evidence also shows that foreign banks perform better, suggesting that bank ownership structures matter when measuring cost performance. Finally, it is evident that the Global Financial Crisis (GFC) had a negative effect on banks' ability to minimize their cost inefficiency levels. However, the post-crisis findings demonstrate that Turkish banks have started to recover from its negative effects, mainly by improving their allocative performance. Highlights: We create a Nerlovian cost inefficiency two-stage DEA model. We model banks' production process in the presence of human capital. We evaluate the Turkish banking system over the period 2007–2016. We find a nonlinear effect of human capital on bank's inefficiency levels.