Is data envelopment analysis a suitable tool for performance measurement and benchmarking in non-production contexts?

After 40 years of research with thousands of application-oriented scientific papers, empirical evidence that data envelopment analysis (DEA) has really improved the practice of performance measurement and benchmarking in real-life non-production contexts is rare. The main reason for this deficit may be that DEA is founded on the concepts of production theory such as production possibility set or returns to scale. These concepts can hardly be applied to pure multiple-criteria evaluation problems, which are often attempted to be solved using DEA. This paper systematically investigates strengths and weaknesses of DEA in the exemplary case of welfare evaluation using real data on 27 countries of the European Union. We analyze and explain the differences in the results of various frequently used DEA models for two different, but strongly connected sets of welfare indicators, thereby demonstrating the pitfalls, which often arise in the application of DEA, as well as some approaches for avoiding them.