DEA方法的效率悖论与数据短尾现象

通过实例分析发现许多重要DEA模型给出的评价结果都可能存在效率悖论,即随着评价标准的提高,一个决策单元的效率值反而会增大.为了探究DEA效率悖论产生的原因及解决办法,首先通过实例证明一些最重要DEA模型(包括CCR模型、BCC模型、FG模型、ST模型)给出的评价结果均可能存在效率悖论.然后,从理论上解释了DEA效率悖论产生的根源在于决策单元数据的短尾现象,并给出了判定存在数据短尾现象的方法.同时,给出了一种修正的DEA模型,该模型不仅可以克服DEA效率悖论的出现,而且还能有效提高DEA方法对效率测算的准确性.最后,应用2000-2014年的中国省级经济数据对相关模型进行了比较研究.

The efficiency antinomy of the DEA method and the short tail phenomenon of data

Some efficiency antinomies can be found by example analysis when most of important DEA models are used. Namely, the higher the evaluation standards is, the larger the efficiency value of a fixed decision-making unit. In order to explore the occurrence causes and solution method of DEA efficiency antinomy, this paper firstly proves that the efficiency antinomy exists when some most important DEA models (including CCR model, BCC model, FG model and ST model) are used by examples. Then, it is explained theoretically that the origin of the DEA efficiency paradox lies in the short-tail phenomenon of data in decision making units, and the method for judging whether there is the short tail phenomenon is provided. At the same time, a revised DEA model is given, which can not only overcome the emergence of DEA efficiency paradox, but also improve the accuracy of efficiency measurement. Finally, a comparative study of the related models is given by using the provincial economic data in China from 2000 to 2014.