Evaluation of the technical efficiency of Brazilian municipalities in public education

Abstract

This paper estimates technical efficiency scores using data envelopment analysis (DEA), in order to understand what factors explain the variance in return rates across municipalities in primary public education. Four features of the study are noteworthy: i) it considers a wide range of inputs and outputs; ii) it includes municipalities from all over Brazil, in contrast to previous estimates with uncertain external validity; iii) it carefully examines the influence of extreme observations, estimates the bias intrinsic to the nonparametric method, and obtains coefficients and standard-errors in the econometric model which exhibit better statistical properties; and iv) it also implements two group comparison tests to clarify the importance of some policies. Some of the new results are as follows: i) the average inefficiency dropped from 24% to 12% after eliminating the most influential data points, but the final estimate is 17% due to the bias correction, which is still much lower than in previous studies; ii) the midwest is the most inefficient region at the local level, which is particularly clear after the scarcity of non-discretionary inputs in the northeast is accounted for; and iii) the dissemination among schools of programs for tutoring and reduction of dropout or repetition rates, as well as the existence of a municipal plan of education and public consortium in this area, is associated with higher levels of technical efficiency.