Uma análise da violência no ambiente das escolas públicas brasileiras

Abstract

School violence is a social problem that affects the human capital of the economy at a decisive stage in school formation. The objective of this dissertation is to analyze violence in the publicschool environment, through three essays. Initially, we sought to analyze whether violence could be transmitted between generations of classes, through the social interaction of students over time. To this end, a School Violence Index (IVE) was constructed using the factor analysis technique, using information from Saeb from 2011-2019 for students in the 9th year of elementary school, based on six original variables, which are: attack on life, threat, robbery, consumption of alcoholic beverages, use of drugs and possession of a weapon or firearm. This indicator was then used as a dependent variable in a school violence equation, estimated using the dynamic panel methodology (GMMsys), in which the lagged dependent variable is incorporated on the right side of the equation. The results corroborate the “inertia effect” of the school violence phenomenon. Secondly, the objective was to verify the existence of spatial behavior of school violence. For this, IVE was used under the exploratory spatial data analysis (AEDE) methodology. Once spatial autocorrelation was confirmed in the territory, the spatial panel data methodology was implemented for the years 2017 and 2019. The existence of an overflow of school violence between neighboring municipalities was confirmed, proving the hypothesis of the “contagion effect” of violence school. Finally, we sought to analyze whether the Happy Child Program (PCF) has the potential to reduce indicators of school violence for students in the 5th and 9th year of elementary school, by strengthening family ties in the community. To achieve this, the instrumented difference-in-differences methodology (DDIV) is implemented. The results indicate that schools in the municipalities that joined the PCF have lower school violence indicators. It is observed that the magnitude of the coefficients is greater in milder cases of violence and with greater impact for the 5th year. The results remain robust to placebo tests and different samples.