This correlation may bring problems to the analysis of

This correlation may bring problems to the analysis of school performance due to the fact of not knowing what kind of racial ion channel (school or residential) really causes the proficiency gap between races. Card and Rothstein (2007) tried to overcome this difficulty by analyzing the impact of both types of racial segregation on school performance. They found that relative school proficiency of black students is not affected by school racial segregation, controlling by neighborhood segregation. Besides this, when black students moved from a highly segregated neighborhood to an integrated one, the proficiency gap between races diminished by approximately 25%. These findings provided evidence that residential racial segregation has a meaningful impact on the proficiency gap between black and white American students. Another problem which can affect the analysis of racial segregation is the fact that individuals are not allocated randomly in neighborhoods and schools. In general, people choose where to live and, to a lesser extent, where to study. The literature points out two main strategies that have been used to deal with this problem. First, Hoxby (2000) and Hanushek et al. (2002) used a variation in the contact with other races of different cohorts for the same school. Specifically, they used each school grade affirmed as being of random allocation to verify how the effect of the contact with other races varies inside the same school. The second strategy brings as solution of the problem of non-randomness the elimination of this bias through the aggregation of data. Evans et al. (1992) used the variation between the mean of personal characteristics in different metropolitan areas to verify the effects of segregation. According to them, this strategy works because although students of different abilities can be allocated in different schools or neighborhoods inside the city, it is likely that the distribution of abilities between cities is approximately random. Also following the second strategy, Cutler and Glaeser (1997) verified that besides non-random allocation of students among schools of the same city, non-random allocation of people between cities can also occur due to migration. For the authors, in the American context, blacks who succeed in professional life tend to live in richer and “whiter” places. Therefore, they used the difference between races in the same city to extinguish the bias that would emerge from non-random allocation of races between cities. In Brazil, the literature about racial segregation and proficiency gap between races in school tests is still incipient. For this reason, we discuss papers that address this issue in a more general way. Soares and Alves (2003) found that the proficiency of white students is superior to the black students’ proficiency and, to a lesser extent, to mulattos’ proficiency as well, but they did not verify either the reasons for this result or the role of racial segregation in this process. Albernaz et al. (2002) estimated an educational production function to understand the determinants of school achievement of Brazilian students. The results showed that race has a great influence on school performance, with black students showing inferior performance. Ribeiro (2005) showed that a larger spatial and social segregation leads to larger school lag, but did not investigate the relation between racial segregation and performance on standardized tests. In any event, it seems like residential and school segregation are relevant factors in the determination of black students’ school performance relatively to white students’, mainly in the United States. In Brazil there is also a proficiency gap between black and white students, but there are no studies that have measured the impact of any kind of racial segregation. Therefore, this work aims to fill this void and study the effects of racial segregation on the performance of black and white students.