Avaliação da eficiência produtiva das ferrovias de carga no Brasil: uma aplicação da metodologia DEA

Abstract

The Agência Nacional de Transportes Terrestres (ANTT) monitors the performance of rail freight through various indicators, and ensures that goals are being met or not by investigating the causes of noncompliance. However, there is no evidence of a deep investigation of the level of efficiency of the railway freight transport in Brazil. Note that it was one of the arguments used in favor of privatization of the railways and is one of the principles to be pursued by the agency. The appraisal of the performance of productive units involves analysis to several processes and calculation of a wide variety of productivity measures. Either, it requires choices between an equally comprehensive range of methodologies that seek to identify the most efficient units in the production of certain goods or services. In general, these methods involve the calculation of rates of efficiencies in terms of output per unit of input (eg, cost per unit of output, profit per employee, revenue per unit of capital employed, etc.). Cooper et al. (2000) called this analysis as the search for partial productivity measures since the resulting indicators do not cover all the materials and products used in the process. For a given production process, a complete analysis of efficiency requires the inclusion of all inputs and all outputs. To ensure it, Charnes, Cooper and Rhodes (1978) presented their seminal work called Measuring the efficiency of decision-making units in which describe the fundamentals of Data Envelopment Analysis, or simply DEA. It is actually an extension of the study initially presented by Farrell (1957), which analyzed the problems of partial measures of efficiency. This work is an application of DEA to viii evaluate the productive efficiency of freight rail lines in Brazil based on data collected by ANTT from 2006 to 2009. From these data, eight models were prepared for analysis. Four was prepared considering constant returns to scale and the other considering variable returns to scale. The proposed models were solved using the package Benchmarking for R. The main hypothesis of this paper is that the large gauge railways tend to be more efficient than the metric gauge railways. The results obtained here suggest that this hypothesis should be rejected.