Abstract:

Data Envelopment Analysis (DEA), a popular method, has been extensively used for ranking and classiffying the decision making units. DEA, a nonparametric technique, is an alternative method to multivariate statistical methods when it is used for the data with multiple inputs and outputs. In this study, DEA’s and multivariate statistical methods’ performances are compared in ranking and classiffying the 81 provinces of Turkey with respect to 14 social-economic and demographic variables. To classify the provinces, Discriminant Analysis, mutivariate statistical method, and K-Means Cluster Analysis, nonhierarchical clustering method are used and for determining the correspondence of these methods and DEA Kappa statistics is used. To rank the provinces with respect to their development level, Principal Component Analysis and DEA are used and these methods’ ranking relationship is tested by Spearman’s rank correlation coefficient. After the applications, for both ranking and classifying. DEA gives similar results with multivariate statistical methods. To make these correspondence, a general and reliable simulation study is done and this study has given similar results. DEA provides researchers a wide usage opportunity since it does not need any assumptions, unlike the multivariate statistical methods and it has a flexibility to add new restrictions to model according to researchers need. Based on DEA’s advantages and both real data and simulation study results, it is concluded that DEA can be used instead of multivariate statistical methods.

Keywords: