Human resource policies in public education: empirical evidences for Brazil

This thesis brings three empirical exercises on human resources issues in Brazilian public schools, taking advantage of a large policy implemented in São Paulo state school system. This policy raises the wages for teachers working in poor urban schools and its assignment, based on an arbitrary cutoff on a socioeconomic index, allows the identification of causal impacts. In sum, the three papers point that allowances policies are able to, in fact, maintain teachers in disadvantaged schools and this effect, in turn, improves students’ academic performance. Besides, we also find that this policy also reduces teacher absenteeism. However, as a consequence of the policy design, there are no evidence that this allowance improves the profile of teachers allocated in those disadvantaged schools. The first paper evaluates the impacts of this policy on teacher turnover, students’ grades and teachers’ profile. We find that the wage compensation reduced the turnover rate by 7.2 percentage points, which means a drop of 15% over the pre-treatment average. In a reduced form model, we also find evidence that this policy can positively impact students’ performance. The second paper further analyzes the impacts on student learning, focusing on three possible mechanisms: i) the turnover itself; ii) the quality of teachers; iii) the wage increase. Estimates show that the only channel through which this compensatory policy affects students' performance is the reduction in teacher turnover. By reducing turnover rate in one standard deviation, the policy reduced the proportion of low performance students in about 50% of a standard deviation. The third paper evaluates how the wage differentiation created by this policy affects teacher absenteeism. Results show that, after controlling for teachers' and schools' fixed effects, paying a higher wage (on average a raise of 26%) causes a drop in teachers' absent days of 8-22%. Absences that do not lead to salary discount, like for medical leaves, don't respond to the wage differentiation and the impact is larger for teachers that receive a higher incentive.