Abstract
The performance differential between students from military and civilian schools cannot be attributed only to the school effect, since factors such as family characteristics, accumulation of knowledge and the selection process established by military schools can generate a bias in performance. This work investigates whether there is still a military school effect in empirical models of proficiency a la Hanushek (2016)HANUSHEK, E. What matters for student achievement: updating Coleman on the influence of families and schools. Education Next, v. 16, n. 2, 2016., once these factors are controlled. A longitudinal administrative database for public schools in Ceará was used, following students in fifth and ninth grades, who were selected by CEM (Coarsened Exact Matching) matching method related to performance in base year. It was found that the lack of control for selection bias and former performance increases the military school effect by one and a half times in traditional models. However, even controlling these factors, the military school effect is still significant.
Keywords
military schools; educational performance; matching method
