DEA-based performance measurement with a centralized view on DMUs

标准数据包络分析（Standard Data Envelopment Analysis， DEA）被广泛用于评估决策单元（Decision Making Units， DMU）的效率，该方法假设每个决策单元独立运行。然而，在集中管理的情况下，伞形组织努力提高其dmu的整体性能。我们的研究通过根据一种控制型结构修改基于dea的基准测试系统来解决这一问题，通过这种结构可以考虑到中央管理层的偏好。DFG项目（删除符号）（2015年10月- 2018年9月）旨在开发将关键控制参数纳入DEA的方法，并具有集中的测量视图，而项目（删除符号）（2018年10月- 2020年9月）的目的是为划分为不同群体的dmu和分权程度不同的组织设计激励系统。我们各自的研究结果开辟了另一条研究途径，即将环境变量纳入集中的测量视图。一个设计良好的方法来解释这些变量不仅可以加强我们正在开发的概念框架，而且还可以产生创新的模型。它们是我们最终研究目标的焦点：考虑环境变量-设计具有集中测量视图的激励系统。当采用基于边界的基准方法来运行激励系统以提高绩效时，考虑dmu环境的异质性是特别重要的：虽然这样的系统迫使表现不佳的单位节省成本，例如，表现良好的单位可能因表现更好而获得奖励。当然，这些单位可能会努力声称，它们对自己的表现不负有全部责任，因为它们的表现在某种程度上取决于它们无法控制的因素（如环境变量）。在这种背景下，我们的目标是设计一个考虑到环境变量的新的激励制度。将在巴西的电力传输部门显示其有用性。有关的管理机构在制定适当的奖励制度方面面临着严重的问题，这个制度应该用来管理这个部门。基于这个例子，我们的第一个目标是展示如何从一般的委托代理理论和特别是基于边界的激励监管的角度来建模这样的监管系统。为了考虑环境因素，巴西监管机构采用了所谓的第二阶段程序：在第一阶段计算的DEA效率得分在第二阶段的一组环境变量上进行回归。然而，监管机构的程序存在许多严重的缺陷，例如，样本量小。因此，我们研究的第二个目标是令人信服地将环境变量纳入绩效分析。

Standard Data Envelopment Analysis (DEA) has widely been used to evaluate the efficiency of Decision Making Units (DMUs) on the assumption that each DMU operates independently. Under the scenario of a centralized management, however, an umbrella organization strives for improving the overall performance of its DMUs. Our research addresses this scenario by modifying DEA-based benchmarking systems according to a controlling-type structure by which the central management’s preferences can be taken into account. While DFG project (sign removed) (10/2015–09/2018) served to develop approaches to incorporate crucial controlling parameters into DEA with a centralized view of measurement, the purpose of project (sign removed)(10/2018–09/2020) is to design systems of incentives for DMUs partitioned into distinct groups and for organizations with different degrees of decentralization.Our respective findings have opened up one more research avenue that refers to the incorporation of environmental variables into the centralized view of measurement. A well-designed approach to account for such variables not only can enhance our conceptual frameworks being developed but also gives rise to innovative models. They are in the focus of our final research objective: Accounting for environmental variables – Design of a system of incentives with a centralized view of measurement.Taking the heterogeneity of the DMUs’ environments into account is of particular importance when frontier-based benchmarking approaches are used to run incentive systems for improving performance: While such a system obliges poor-performing units to save costs, e.g., good-performing units may receive rewards to perform better. Naturally, the units may strive to claim that they are not fully responsible for their performance that would to some degree depend on factors (i.e. environmental variables) beyond their control.Against this background, we aim at designing a new system of incentives that takes environmental variables into account. Its usefulness will be shown with regard to the electricity transmission sector in Brazil. The respective regulatory agency faces severe problems regarding the development of an appropriate incentive system by which this sector should be regulated. Based on this example, our first goal is to show how such regulatory systems can be modeled from both the perspective of principal-agent theory in general and frontier-based incentive regulation in particular. To account for environmental factors, the Brazilian regulator applies the so-called second-stage procedure: The DEA efficiency scores computed in a first stage are regressed on a set of environmental variables in a second stage. However, the regulator’s procedure suffers from a number of serious shortcomings, cause by, e.g., the small sample size. Hence, the second goal of our research is to convincingly incorporate environmental variables into the performance analysis.