TARGET SETTING IN THE PRESENCE OF NEGATIVE DATA: AN APPLICATION IN BANKING

Conventional DEA models assume non-negative data. However, in real-world applications there many occasions in which input/output variables can take negative variables. Especially in some cases both of positive and negative data simultaneously appear in inputs/outputs data. In paper we will illustrate the approach proposed in Kazemi Matin and Azizi (2011) in an empirical application in Banking. To do this end, first we will give a review and also comparison of the most recent of approaches that introduced in the literature to deal with negative data. Afterwards with some numerical examples we will show that the presented approach may fail in setting efficient target units when positive input/output levels are preferred in benchmarking. A new additive DEA model is next introduced to overcome this issue. An empirical application of the proposed model in performance measurement and benchmarking of 61 Islamic bank branches is also used to illustrate the approach.

Keywords:

Data envelopment analysis , Negative data , Target setting ,