Date Envelopment Analysis Models Based On Classification Of Variable Attributions

Efficiency is one of the core concepts of the economic theories and practices. Resource allocation is aimed at efficiency. The organization operation seeks for the efficiency. The economists and managers focus their attentions on the theoretical and applied studies of efficiency measurement. However, they didn't come up with methods of the efficiency evaluation. Only the operation researchers nicely resolved the problem, that is DEA.Data Envelopment Analysis (DEA) is a mathematical programming approach to assess the relative efficiencies of decision making units with multiple inputs and outputs. One of the greatest advantages of DEA is that inputs and outputs can have very different units, with no difficulty of aggregating inputs or outputs. DEA can be used to measure various efficiencies, for example, technical efficiency, scale efficiency, allocative efficiency, productivity and etc.This dissertation mainly studies the classification of variables, and DEA models based there upon. In the classic DEA model, input (or output) variables are assumed to be controllable, of singular identity, and quantifiable. But, along with the DEA study development, people discover the above-mentioned assumptions aren't realistic. While some variables are exogenous, others have "dual identity", still others are either not quantifiable or not of significance to the outputs. Therefore, variable research is one of the important contents in the DEA theories and application.This paper expands DEA variable research on basis of the existing DEA variables researches. The main innovations of dissertation are as follows: categorization of the input variables by varying criteria; the distinguishing of semi-discretionary variables as well as studies of DEA model and its property based on them; depending on whether the variables are marketable, input variable are divided into quantitative ones and attributable ones. In addition, this paper distinguishes an attribute variable and qualitative variable, and studies existent function of attribute variables. It distinguishes quasi-fixed variable and general variable according to the position of the variable. On basis of variable classification, this paper gives a detailed study on the two-stage production system DEA and the dynamic DEA. Firstly, there is the initiative way of measurement thinking. Secondly, the concept of semi-discretionary variable is introduced to build up new DEA model. In the new approach for two-stage production system DEA model, both input-oriented and the output-oriented DEA models are applied. This paper uses the actual data of china commercial banks as numerical example, and attempts to explore the commercial bank DEA efficiency evaluation method.This dissertation is divided into 7 chapters, each of whose contents are listed as follows:In Chapter 1, this paper reviews the history of DEA development and the main contents of DEA studies, and summarizes DEA study in China at present. This chapter introduces the characteristic of the DEA model on condition that factors can't be controlled, which is followed by an analysis about the meaning and significance of this study.Chapter 2 starts with an introduction of how the problem of classification of semi-discretionary variables was raised, followed by the illustration of DEA model based on semi-discretionary variables, and its basic properties. At the end of the chapter, actual data will be applied in the new model for demonstrative purposes.In Chapter 3, classification of semi-discretionary variables will be applied in the dynamic DEA model. Following a brief review of dynamic DEA studies in the past, a dynamic DEA model based on semi-discretionary variables will be put forward and tested with actual data.Chapter 4 studies another type of variables classification, namely attributive variables. Following an account of the importance of attributive variable classification, the dynamic DEA model based on attributive variables will be introduced and tested with actual data. In a sense, this chapter is an expansion of the dynamic DEA model.In Chapter 5, the author focuses on quasi-fixed variables. In addition to an introduction about the relation between quasi-fixed variables classification and the network DEA and its corresponding property, detailed introduction about the network DEA will be presented. After that, the author goes on to study the two-stage production system DEA, which is a special example of the network DEA, and concludes with a summarative analysis, pointing in particular out the innovative features of this research approach. This chapter concludes with an innovative two-stage resource-constrained production system DEA model, which is duly tested with actual data.Chapter 6, an extension of chapter 5, first reviews two-stage production system DEA models of the sequent type, which is followed by the application of semi-discretionary variables into two-stage production system DEA model, hence results in a new DEA model based on semi-discretionary and quasi-fixed variables. The new simulated model is in turn tested with actual data.In Chapter 7, the concluding section of the full dissertation summarizes some of the breakthroughs made in this dissertation, as well as the shortage of this paper and direction of further research.