Ensaios em economia da educação com dados do PISA

Abstract

This thesis is composed of three essays with PISA data. The first, “Early Childhood and its Impact on Student Performance in PISA”, focuses on the discussion of early childhood education and its impacts on future performance. For this purpose, data from Brazil in PISA 2018 was used. were the Propensity Score Matching (PSM), the Quantile Treatment Effects (QTE), the Propensity Score Generalized (PSG) and the sensitivity analysis by Rosenbaum (2002). later or who did not enter, positively influences the performance in PISA. In a stratified way, these results remain, in part, between the medians or the best performances. When exposing students to different levels of education, heterogeneous effects were observed on the performance. The marginal return on performance increases with the increase in schooling in the initial years, that is, up to 2 years in mathematics and 3 years in science and reading. Sensitivity analysis de Rosenbaum (2002) indicated that the results are robust to unobservable variables. The second essay, “Do Empathy and Awareness of Bullying Affect PISA Performance?”, investigates how empathy and awareness of bullying impacts academic performance. For that, data from Brazil was used in the 2018 PISA and the technique used was Double/Debiased Machine Learning. Through the results, it was inferred that students who are more empathetic and/or aware of bullying present better performance in the three PISA competencies. These results are maintained even when making several cuts in the data, namely, by location size, by type of school (public or private) and by gender. The performance increase is more pronounced when comparing students who live in regions with more than 1 million inhabitants or who are girls. This accentuation varies between the types of schools, in which it is conditioned to the form of measurement and the type of skill considered. The results are robust, given that Oster's (2019) sensitivity analysis did not find problems with omission of variables. The third essay, “Unconditional Quantile Decomposition of Gender Performance Gaps”, focuses on the debate on inequality between boys and girls, comparing the differences in the performances of Brazilian students with those of the OECD in the 2018 PISA exams. distribution and decompose the grade, the methodology of Firpo, Lemieux and Fortin (2018) was used. From the difference in points of the performance distribution, it can be concluded that there is inequality in the three evaluated competences, regardless of the performance level. In general, boys perform better than girls in math and science, while girls perform better in reading. In addition, it was inferred that inequality is greater in Brazil in mathematics and science, regardless of performance, while in reading, inequality is greater in the OECD in the smallest (10th and 25th quantile), in the medians (50th quantile) and in the highest performances (quantile 90).