Eficácia escolar sobre o aprendizado em matemática: um estudo longitudinal sobre o efeito das escolas municipais de Belo Horizonte

Abstract

The purpose of this research was to estimate the school effect in schools under local management (municipality) in Belo Horizonte city (MG/Brazil) on mathematics learning for students between the 3rd grade (8 years old) and 9th grade (14 years old). The research is based on evidence from the Brazilian literature on students' low performance, especially those from disadvantaged social backgrounds, according to large-scale tests. International literature arguments that schools cannot compensate for social inequalities alone, but both teachers and schools have an effect on student outcomes. A constraint of most studies carried out in Brazil on the effect of schools is the use of cross-sectional data, that poor estimate the school effect. In this research, we use longitudinal data, from 2010 to 2014, from two sources: the “Censo Escolar” (Brazilian School Census) and the “Avalia BH” (large-scale Assessment System for public schools in the City of Belo Horizonte). We analyzed enrollments, proficiency in math tests, and contextual questionnaires. We point out important contributions to the school effect research with longitudinal data. Highlights: 1) we estimate the socioeconomic status (SES) of students, applying the Item Response Theory (IRT), to test whether the SES changes over the years; hypothesis that has not been confirmed. This is a relevant finding for longitudinal studies, in which this is an important control variable. 2) We analyzed missing data (attrition), a common problem in longitudinal surveys, but which can affect school effect estimates. We found evidence that differences between schools help to explain the attrition of information. 3) We use growth curves in three-level hierarchical models to estimate the school effect. It was estimated using the Intraclass Correlation Coefficient (CCI). We found school effect values below those observed in cross-sectional assessments and below expectations for developing countries. Modest effects were observed for the learning trend in promotion by grade: both the variability of the individual level and the school level explain little about learning. We conclude that Belo Horizonte public schools seem to have a homogeneous effect, better explained by the general growth trend than by the differences between schools. Effects of ethnicity, SES and age-grade distortion show that an alternative to improve learning outcomes could be the administration of public policies aimed to change these inequalities. We conclude that Belo Horizonte public schools seem to have a homogeneous effect, better explained by the general growth trend than by the differences between schools. Effects of ethnicity, SES and age-grade distortion show that an alternative to improve learning outcomes could be the administration of public policies aimed to change these inequalities.