Probabilidade de progressão por série no Brasil: evolução, seletividade e aplicação de modelos de idade-período-coorte

Abstract

The provision of the universal access to schooling and the assurance of students permanence in school has been the object of public policies in almost all nations. Education is conceived not only as a form of individuals emancipation, but also as an important factor to promote the economic and social development. Since the 90s, Brazil has seen the implementation of many inclusive educational policies. Despite of these investments, Brazil is still far away from the average educational attainment expected to be achieved in accordance to its current level of economic prosperity. Hence, we consider that the problematic about the education quantity is still very important to the Brazilian context.This study has, thus, as focus the analysis of the access and permanence of the population at school. For this, the key concept is Grade Progression Probability (GPP). The GPP is a demography method which can be employed in the micro or macro level. In this context, this dissertation is composed by three articles. The first one deals with the concept of GPP in the macro-demography view. We present the evolution of some indicators, derived from GPP, stratified by socioeconomic attributes and we trace inter-relationships between the improvement of these indicators and the educational policies adopted since the 80s. The second article is a contribution to a large literature regarding inequality of educational opportunities in Brazil. In this article GPP concept is applied to the individual level through the estimation of school transitions models. We seek to identify the magnitude of the relationship between social origins and the progression probability on the educational scale. The third article is a methodological application, to selected grade progression probabilities in Brazil, of two alternative approaches to the identification problem in the age-period-cohort models. The aim of this exercise is to present the so-called intrinsic estimator, which emerged as an alternative to the conventional constrained generalized linear models. Literature on this subject has shown that this new estimator provides a generic solution to the estimation of age-period-cohort models, and does not require external or a priori information about data.The results of the two first articles show that there were strong advances of the educational policies adopted in Brazil, especially because they allowed a convergence between population groups, in the earlier stages of school career, on the grade progression probabilities and also the improvement in the inequality of educational opportunities. Nonetheless, we detect persisting barriers to schooling, due to the selectivity in the educational trajectory, which tends to be more pronounced in the later stages of the school career. The third article, in turn, reveals the potentiality of age-period-cohort model solution based on the intrinsic estimator, which presents excellent statistical properties. Thus, GPP projections based on this estimator may be very promising.