DEA-based performance measurement under centralized management

Abstract

Traditional performance measurement approaches are usually characterized by a number of different limitations. Among other things, these approaches require the subjective determination of weights to aggregate a set of indicators to an overall performance score. Furthermore, traditional approaches are usually not able to incorporate additional improvement potentials that can be received from a centralized management. A performance measurement framework which can overcome these limitations is called data envelopment analysis (DEA). Against this background, this thesis provides a thorough overview of how different degrees of centralization are modeled in the current DEA literature. The systematic literature review identified 135 different approaches that assume a centralized or partially centralized management structure. A concluding discussion of the respective DEA approaches showed two fundamental research gaps. In response to this, this thesis has two fundamental objectives: The first objective is to propose a DEA-based performance measurement approach for measuring performance changes over time. The second objective is to develop another DEA-based approach for comparing the performance of management groups. In contrast to so far developed DEA-models, the here proposed approaches explicitly incorporate the respective management structure. Both DEA approaches thus developed are based on the combination of the metafrontier concept and the Malmquist index. The first approach evaluates productivity changes of operating entities over time and, hence, may indicate potential sources for performance changes. Thereby, the proposed approach preserves the individual characteristics of each local group technology. The second DEA approach proposed here uses the Malmquist index for comparing the performance of management groups. This index accounts for the existence of a central decision maker who can, e.g., undertake resource reallocations to improve the overall performance of its managed group. The applicability and usefulness of both proposed approaches is empirically shown with real-world data from KONE Corporation.