Study On The Theory Of DEA-Based Resource Allocation And Its Application

Efficiency evaluation methodology can be classified as parametric and nonparametric. The parametric methods require explicit specification of production function before evaluation, while the DEA (Data envelopment analysis), as nonparametric method, doesn’t need any a priori information such as production function. It takes advantage of production possibility set. In general, the econometric models estimate the mean efficiency, while the DEA characterize the extremal relation. Therefore, both mothods deserve merits, and are important in the application and theory of efficiency evaluation.At present, the application field of DEA grows larger and larger. It has been extended from the traditional fuction as an ex post evaluation of the past performance to ex ante function as a planning tool. The former function is diagnostical, and the latter function is predictive. In the current thesis, DEA is studied from the perspective of ex ante function aiming at extending its predictive ability. In particular, it mainly studies the application of DEA for resource allocation purpose.The thesis consists of six chapters, the topics of which are illustrated as follows:Chapter1firstly reviews the basics in the field of efficiency evaluation. Then, it summarizes the main topics under research to date in the DEA literatue. In the end, associated with the application of DEA to the resource allocation and target setting topic, the significance of the thesis is highlighted. The chapter concludes with the structure of the thesis.Chapter2studies FDH model for parallel production system, and the related properties. The proposed model is then applied to the research of traditional assignment problem in operational research by modeling it as a parallel production system. In doing so, this paper firstly extends the DEA methodologies both in theory and application; secondly, the paper broadens the application of the traditional assignment problem in operational research, i.e., this paper extends the traditional assignment problem to cases with multiple inputs and outputs from a single criterion such as cost or profit. Finally, the proposed model is able to aid decision makers in assigning jobs with resource constraints together with target setting.Chapter3aims at the problem with a particular parallel production system that the variable returns to scale production possibility set (PPS) of decision maki" (DMUs) can not be directly characterized by the convex hull of PPS’s of Sub-DMUs (SDMUs). This chapter studies the efficiency assessment models, the relevant properties, and its resource configuration implication for parallel production system. Then it proposes two different definitions of technical efficiencies:the first is termed as DMU internal technical efficiency and the other is DMU technical efficiency. The ratio of the two indicators is used to reflect the performance of the internal resource allocation of DMU. Finally, a simulation experiment is used to study the proposed models deeply.Chapter4presents a multi-objective programming to aid the decision makers in resource allocation and targets setting. Five criteria, namely feasibility criterion, effectiveness criterion, efficiency criterion, equality criterion, and minimal variation criterion are proposed and incorporated into the formation of resource allocation and target setting scheme. The proposed approach is based on network Data Envelopment Analysis (NDEA). Hence it can be used to generate the optimal distribution of resources among the internal components of each DMU. In this way, it enables decision makers to link the resource allocation and target setting scheme at a DMU level to that at an organization level. In addition, the proposed methodology offers decision makers an approach to exploring the tradeoffs among effectiveness, efficiency, equity, and allocation cost in a consistent way. Finally, simulated data of10banks under the administration of a centralized office are used to illustrate the approach.Chapter5proposes a new DEA model to study efficiency of parallel system. In addition, it takes into account the AR restrictions and resource allocation of the parallel system. Not only can this method obtain reasonable solutions to the proposed DEA model, but also provide the managers with inner-information relative to resource allocation among the sub-systems within each DMU, and this really aids them in allocating resources efficiently in the real producing processes. Finally, to illustrate the advantages of the new DEA model, the paper presented an example of application to all countries that had participated in the last Summer and Winter Olympic Games.Chapter6summarizes all the work of this thesis and presents some useful directions for future researches.