Partial identification of Brazilian private schools effects

Abstract

The core aim of this article is to compare student performance in Brazilian public and private schools based upon the methodology of Partial Identification. Since there is selection bias in the school choice derived from socioeconomic factors, identification restrictions in the other methodologies based upon point estimations to infer about the students’ performance differences from the two types of schools are questionable. Therefore, it is tested whether the estimated limits provided by the proposed methodology, although under less restrictive assumptions, are as much or more informative than the point estimates generated by other methodologies. Using microdata from the Basic Education Assessment System (SAEB) of 2005 for students in the 5th and 9th year of elementary school, the results indicate an overestimation of the private school effect from the most applied methodologies that rely on the assumptions of ignorability and imputation. However, the imposition of less restrictive hypotheses has kept such an effect significantly positive. These results shall come from the selection bias related to the parents’ earnings constraints which contribute to explain the school choice and make it difficult the determination of an acceptable counterfactual.