Evaluation of Brazilian electricity distributors using a Network DEA model with shared inputs

Data Envelopment Analysis (DEA) is a methodology widely used to assess the efficiency of electricity distribution companies. Most papers use the network length as one of the variables in the modeling: some of them as an input, since it accounts for the total electricity distributed, and others as an output, since it consists in the product of the investments made and the geographic dispersion, thus showing different dimensions of the same problem. In order to deal with this situation, this work presents an innovative way to deal with this variable by using a Network DEA model with two stages. In our proposed approach, the network length has been set to be both the output of the first stage (whose input is a rate of the operating costs) and an input for the second stage (whose outputs are number of consumers and total consumption). In addition, factors for sharing the operating costs between the stages are taken into account. For the sake of completeness, two Network DEA models are introduced, namely relational and additive approaches. Finally, the results of both models are presented, compared and discussed.