This thesis is composed of three independent empirical articles. In the first chapter is evaluated to what extent differences in labor supply factors and careers by gender influence the wage gap observed in the brazilian formal labor market. For this, we use a 1 percent representative random sample of the workers in RAIS / MTE between the years 1994 and 2009. From this information is retrieved the career path of individuals in the sample. The empirical strategies exploit the longitudinal feature of this database to generate complementary information about the gender wage gap. The results show that the labor market history measures have economically relevant impacts on individual incomes. Career interruptions reduce average earnings and workers with continuous labor market attachment have higher wages. However, the inclusion into the analysis of these more reliable job experience variables results in a reduction of up to 10 percent in the magnitude of the gender wage gap estimates. This represents a quite reduced influence. The second chapter investigates the relative importance of two dimensions of teacher quality for the learning in mathematics and Portuguese of eighth graders of the elementary school in São Paulo state. For this purpose, we adopt an approach based on the educational production function and the main specification uses a value added model with control for the students past grades. The results show that both the teachers knowledge and pedagogical activities inside the classroom have a positive and statistically significant impact on the acquisition of cognitive skills. However, the teachers knowledge effect has a small economic magnitude. Yet the impacts associated with the frequent application of effective teaching practices are quite large. For instance, the intervention defined by the replacing a math teacher who does not always give homework for another that does it, increases the students proficiency in approximately 12 percent of a standard deviation of the grades distribution. The third chapter examines the relationship between political partisanship and government size for the context of the brazilian municipalities after 2004 local election. In order to achieve this, we use a regression discontinuity research design to estimate the local causal effect on political choices of a municipality being governed by a left-wing party instead of a right-wing one. The results show that left-wing governments spend proportionately less on urbanism and health, and more on administration. Nevertheless, this higher administrative spending is not associated with an excessive hiring of public employees.