Indigenous and non-Indigenous proficiency gaps for out-of-school and in-school populations: A machine learning approach

Abstract

Despite the renewed emphasis in equity for SDG4, Indigenous learning gaps persist. Indigenous barriers for learning are intersectional -a combination of multi-layered and heterogeneous causes. In this paper, we use data from PISA for Development to estimate the Indigenous learning gap in Guatemala, Paraguay and Senegal for out and in school samples. We employ machine learning which allows to employ numerous controls and their interactions, accounting for intersectionality. We find that negative learning gaps remain for both samples (with some differences by level by of performance) even after controlling for around 66–217 covariates, showing the extent of Indigenous-driven inequality and discrimination.