The present research aims to contribute theoretically and empirically to understand the extent to which the endogeneity problem, a major concern frequently observed in educational production processes, affects the estimation of technical efficiency using the Data Envelopment Analysis (DEA) technique. Furthermore, this work combines insights from impact evaluation literature and non-parametric frontier techniques in order to provide potential solutions to deal with this issue in educational empirical applications and obtain accurate efficiency estimates. Three chapters are developed. Chapter 1 analyses theoretically to which extent does the presence of endogeneity in the production process affect DEA estimates in finite samples. In addition, we propose a simple heuristic to detect this phenomenon in empirical applications and, rooted in the Instrumental Variable (IV) approach, we provide a potential solution to deal with this problem in the estimation of technical efficiency using DEA. In Chapter 2 we apply these strategies to the context of Uruguayan public schools. Chapter 3 focuses on the estimation of teachers' technical efficiency and its effect on students' academic results taking into account the presence of school self-selection.