Identificação parcial do efeito das escolas privadas brasileiras

Abstract

This article aims to compare student performance in Brazilian public and private schools, from an analysis of Partial Identification or Bounds introduced by Manski (1989). In the context of strong selection bias resulting from socioeconomic factors, it is compared the estimates from the Propensity Score Matching with the ones from regression analysis, to verify whether or not the estimated limits under less restrictive assumptions are informative. Using most recent data from the Basic Education Assessment System (SAEB) of 2005 for students in the 5th year of elementary school, the results indicate a significant bias on the estimated effects by the methodologies that are rely on assumptions of ignorability and imputation. A possible explanation to this bias can be attributed to the strong selection bias related to low income level of parents of students in public schools, making it impossible to determine an acceptable counterfactual.