Purpose – The aim of this paper is to call the attention of public entities for the econometric frontier models which allows benchmarking the tax offices accurately. In the context of the European Maastricht, the European countries face public receipts shortage and to focus in the management of the tax agencies in order to maximize receipts. The procedure adopted in this paper allows benchmarking the tax agencies accurately. Design/methodology/approach – The paper adopts the stochastic frontier model to benchmark the tax offices. This model is presently very popular in benchmarking exercises, based on its statistical framework. Findings – The paper finds that the tax offices analysed vary along the sample and along the period. Research limitations/implications – The homogeneity of the tax offices and the short data set are the main limitations of the paper. Relative to the first, since they are managed by the same entity and perform the same task we can claim that they are similar, relative to the second critic, we cannot overcome it because of the secrecy related to this units. Practical implication – The efficiency rankings are due to the way the tax offices handle costs and outputs. The way they handle the combination of output and inputs is explained in the literature by the principal-agent relationship, by collective-action problem, due to asymmetric information between different tax offices and X-inefficiency. The public policy has to address these problems in order to upgrade the efficiency of the tax offices. Originality/value – This is the first paper adopting the stochastic frontier model in tax offices.