The racial achievement gap has been at the center of the educational debate for decades in the United States. Although disparities in educational outcomes have declined in part of the 20^th century, the process has stalled in this decade. For instance, in mathematics the gap in raw scores for students aged 13 has decline from 41 points in 1978 to stable 27-30 points since 1990. In August 2009 results front the SAT scores confirm a widening in the racial achievement gap, thus questioning the success of some policy initiatives. The goal of my dissertation is to examine the contribution of family, school and neighborhood factors to the achievement gap and to highlight promising areas for policy intervention.

This dissertation is structured in three papers. The first paper studies the association between neighborhood socioeconomic composition and student achievement. The findings show that the fraction of college-educated adults and median household income in the neighborhood are positively associated with students' achievement. High levels of poverty have a negative and significant effect only when a threshold of 30% of poor households in the area has been reached. Overall, neighborhood factors can account, on average, for about 5% of student achievement with larger impacts for some subgroups, particularly for Hispanics and Blacks.

The second paper analyzes the effect of enrolling students in Algebra 1 in 8th grade instead of 9th grade. Using a propensity score matching method, this study estimates the average treatment effect on the math section of the tenth-grade North Carolina's High School Comprehensive Test. Results show that low-achieving, low SES, Black students are more likely to take Algebra 1 in ninth-grade than their White peers. Despite the positive average treatment, the effect is small and not statistically significant for low achieving students. Our findings do not suggest that the "Algebra 1, for everyone" policy encouraged since the early 1990s is not equally effective for all students. Students who had low test scores prior to 8th grade did not improve at the same rate or not improve at all, though we can not affirm they are harmed by such policy. The question that remains is how to turn this effect into achievement gains for this group.

The third paper explores factors underlying the achievement gap between White and Hispanic students using the North Carolina Public School Database. We use the Page-Murnane-Willett, Oaxaca—Blinder and Juhn—Murphy—Pierce methods to decompose the gap into school and parental factors. The analysis provides three key results. First, within school factors exceed between school factors. Second, parental education is the most important individual variables: White students have on average better educated parents and that translates to higher test scores. Third, the achievement gap narrows between grades 3 and 10 with the improvement mainly associated with a reduction in within school disparities.