Distorção idade-escolaridade: uma análise a partir de modelos de contagem

Abstract

A great deal of efforts has been placed on the study of how education affects people’s lives, as well as what are the determinants of a good education. This paper aims to collaborate with the debate on education dealing with the determinants of the age-grade distortion, defined as the difference in the expected grade of a student of certain age and his actual grade. Using an econometric analysis, with 2013 PNAD database, it assess how personal and familiar features, such as gender, race, dwelling with one’s mother, maternal schooling or household’s per capita income, is related with the emergence and magnitude of age-grade distortion. Specially, this work employs count data models, such as Poisson and Negative Binomial, and also models that work with excessive zeros, such as Zero-inflated Poisson and Zero Inflated Negative Binomial. Among its results, it finds that as age advances the absence of any grade repetition is related to a smaller probability of one ever comes to happen in the future.