Heterogeneous Treatment Effects of Educational Interventions by using Random Forests

The aim of this paper is to find heterogeneous treatment effects in different applications regarding educational interventions that have not been located before. We shall use a machine learning technique called the Random Forests algorithm implemented by Athey et al. (2019) to reach our goal. With several hypothesis tests and a variable importance measure of the package grf in R, we can test for heterogeneity and find the most important variables that contribute to this heterogeneity. When implementing this algorithm on the applications of our thesis, we first of all found out that the ability of a student has a high impact on the treatment of camera monitoring on the overall Baccalaureate score in Romania. Also, we found that past math grades and the reading ability have a big importance on the heterogeneity in the application of the grading process in Brazil. Moreover, the number of members and kids in households contribute to the heterogeneity in the normalized math scores of students attending a BRIGHT school in Burkina Faso. All these heterogeneous treatment effects have not been discovered before by past researchers.