Some unsolved problems in data envelopment analysis: A survey

The present study is a survey of a set of inherent characteristics and unsolved problems in current DEA models; the paper will focus on what types of local characteristics the DEA estimated technology exhibits in relation to non-jointness of a multi-output-multi-input production. No indicators are currently available telling the analysts whether or not the structure in data supports the existence of a joint technology. If the estimated CCR technology has a sparse facet structure, in the sense that few facets are ‘active’ for output vectors in certain cones of the output space, then the multi-output-multi-input CCR technology is possibly locally non-joint. Another important field of research relates to the inclusion of explicit distributional assumptions as part of the DEA model specification, and to the extension of the DEA estimation from an estimation of a deterministic frontier towards an estimation of a stochastic frontier. Some recent results will be compared. Finally, measures of scale economies from parametric models and DEA models will be compared.