The Research On Relative Efficiency Evaluation Models For Multi-Subsystem DMU Based On DEA

For China, a populous and relative poor country with restraints of resources, efficiency improvement of organization is one of the best way to make the comprehensive, harmonious and sustainable economic development come ture. Efficiency evaluation is one of the most important content for efficiency research, and also the basic safeguard for improving efficiency.Data envelopment analysis is regarded as one of the most attractive methods so far to estimate relative efficiency. While tranditional DEA models treat decision making units (DMUs) as"black boxes", neglecting their internal structures and the correlativity among the internal subsystems, and estimate their efficiencies using the input-output data of the whole units. As the develpoment of management theory and practice, higher requirements are needed for efficiency evaluation. For accurate efficiency information will ensure the managers act with a well-defined objective in mind and be helpful in the efficiency improvement, people desire to obtain more precise efficiency information about the whole units on the one hand, and more detailed efficiency information about their internal subsystems on the other. Therefore, it is necessary to propose new methods which meet the reality's needs for efficiency estimation, following the decomposition of technical efficiency and exploration of the subsystem efficiency.By using data envelopment analysis, this paper starts the research in view of the above problems as follows:When it comes to efficiency evaluation for DMUs with single subsystem-"black boxes", this study proposed a new method to estimate the managerial efficiency of these DMUs, based on management effectiveness and Banker-Morey model. Moreover, according to the definition of scale efficiency, this research defined the environment efficiency. And then technical efficiency is decomposed into the product of managerial efficiency, environment efficiency and scale efficiency.On efficiency evaluation for series or parallel DMUs which contain several relational subsystems, this research employ new network data envelopment analysis models based on the correlativities among subsystems and the assumption of variable return to scale, by following the analysis framework for single subsystem DMUs. The new models developed in this research can be used to estimate not only pure technical efficiency and scale efficiency, but also managerial efficiency and environment efficiency. And all these efficiency indexes for the whole DMU and its subsystems can be obtained simultaneously. The propositions of new models realize the further decomposition of technical efficiency and the acquisition of internal subsystem's efficiency information successfully. The results of theory proof and empirical analysis indicate that, for series structure, new model is superior to tranditional DEA models in identification ability and the rationality of efficiency results. Meanwhile, connection index is defined to describe effcets of the cooperation among subsystems, based on the relationship of system efficiency and those of subsystems. For parallel structure, the irrelational models assign the maximum of subsystem efficiencies to the whole DMU, while the new models assign it with a more reasonable value which is between the maximum and the minimum of subsystem efficiencies. Obviously, the results of new model is more rational than that of the irrelational models.For most network structures can be decomposed into series structure and parallel structure, this research, based on the works for the two basic structures, proposed the ideas and steps for building technical efficiency, pure technical efficiency and managerial efficiency evaluation models systematically.The works of this study will be helpful to explore the relative efficiency of multi-subsystem DMUs effectively in the future on the one hand, and will provide the managers with more reliable indexes to understand the unit's true business circumstance, work out a sound plan and an effective excitation mechanism on the other. So the research done in this paper has some theory and practical significance.