Eficiência das escolas públicas estaduais de Minas Gerais: considerações acerca da qualidade a partir da análise dos dados do SICA e do SIMAVE

Abstract

This paper develops the semi-parametrical two-stage Data Envelopment Analysis to detect the efficiency frontier of public Brazilian Minas Gerais state schools. The first stage computes the DEA-efficiency, the second one inputs the DEA results to a truncated-regression with background and wealth variables.All methods discussed here are based on recently papers of non-parametrical analysis, Simar and Wilson (1998), and semi-parametrical analysis, Simar and Wilson (2002). Further analyses concerned about outlier-detection are implemented: Cazals et ali (2002) and Sampaio de Sousa e Stosic (2005). Those new methods, based on DEA-bootstrap, have improved the robustness of analysis and made became possible an efficiency confidence intervals construction.The new data basis of school proficiency, SIMAVE, and cost-measurement, SICA, either the Educational Census of Education Ministry 2005, permitted this new application of efficiency measurement for schools. The results show inputs complementaritys, in and out school. Best located schools, where the educational inputs are better, are more efficient in probability. However, there are some good examples of efficient schools emerging from poor regions and, in general terms, all stated educational results can improve considerably if the schools get to improve their efficiency.