Educação: Um Escudo Contra o Homicídio?

The objective of this text is to make a preliminary analysis of the relations between educational level and victimization by homicide. To this end, I use the following data sources: the Integrated Mortality System (SIM) between 1999 and 2004; the Demographic Census of 2000 and the National Household Surveys between 1999 and 2004 (except for 2000, year in which there was no household survey). The text consists of both an exploratory analysis comparing homicide rates per 100,000 inhabitants according to age, educational level and sex and a regression analysis to find partial correlation coefficients. The regressions are estimated using cells defined by sex, age, region of residence, skin color and schooling level. The estimation methods used were the linear probability model and logistic regression. My results reinforce three already well-known results. The first is that women suffer homicide rates that are roughly one-tenth that of men. The second is that negroes suffer much higher homicide rates then whites. Finally the most important homicide risk group are youths between 16 and 36. The most important new result is the importance of educational in preventing homicide. Although the exact magnitude depends upon the model being estimated, in all of them schooling is significantly and negatively related to death by homicide. However, caution must be exercised in interpreting the partial correlation coefficients found cannot be interpreted as causal. There are no doubt omitted variables that have a causal relation both with educational level and homicide risk and this may lead to spurious correlation between the two. An analysis using instrumental variables to separate causal and endogenous magnitudes will be the subject of an upcoming study. However, the magnitude of the effect is so strong that even if a fraction is causally due to education, schooling still is one of the most important public policies for reducing homicide.