Black and Hispanic students have lower achievement than White students due to segregation, discrimination, and poverty. If these disadvantages also lead to negative academic attitudes, Black and Hispanic students may disengage from school, compounding the effects of low achievement and limited opportunities. Therefore, my dissertation is organized around two questions: (1) Do racial/ethnic differences in academic attitudes develop in response to educational inequalities? (2) If so, do differences in attitudes translate into differences in educational behavior and decision-making? I answer these questions using elementary and middle school data on mathematics attitudes from the Early Childhood Longitudinal Study, Kindergarten Class of 1998-99 (ECLS-K). Because STEM (science, technology, engineering, and mathematics) professions are the highest paying, racial/ethnic inequalities in mathematics education are particularly consequential for the reproduction of racial/ethnic income inequality. My dissertation has two main contributions. First, I show that Black and Hispanic students’ mathematics self-competence, or self-assessed mathematics ability, declines as they internalize the limitations placed on their achievement by structural racism. In third grade, Black and Hispanic students have high mathematics self-competence relative to White students with similar achievement because they are segregated into underperforming schools. They compare themselves favorably to their low-achieving peers. However, as they get older, Black and Hispanic students’ self-competence falls. By eighth grade, racial/ethnic differences among students with comparable test scores are largely insignificant. Because Black and Hispanic students have lower test scores, on average, this leaves them with lower self-competence overall. These results extend theories on the classic big-fish-little-pond effect by showing that the effect diminishes with age. Second, I demonstrate that, compared to mathematics self-competence, mathematics interest is less dependent on school quality but also less consequential for persistence in STEM. In the second chapter, I find that disadvantaged families are able to buttress their children’s mathematics interest. As a result, Black and Hispanic students end middle school with high mathematics interest relative to their low self-competence. In the third chapter, I show that high self-competence is associated with enrollment in upper-level mathematics courses, whereas high interest motivates more frequent homework completion. Combined, these two chapters demonstrate that mathematics interest is limited as a source of resilience for Black and Hispanic children. Although interest boosts studiousness, the returns to studiousness are lower in the absence of the self-competence to enroll in advanced mathematics courses. Overall, this research advances sociological theory on racial/ethnic differences in academic attitudes. Sociologists of education have disproven the claim that Black and Hispanic communities possess an “oppositional culture” that discourages scholastic achievement as a form of “acting White.” However, these scholars have not posited an alternate theory on the relationship between racial/ethnic educational inequality and academic attitudes. This dissertation shows that Black and Hispanic students’ low achievement leads to negative academic attitudes, not the other way around. Black and Hispanic children enter school with equally positive academic attitudes as White children. Educational disadvantages produce low achievement, which Black and Hispanic students gradually internalize as low self-competence. This low self-competence discourages children from pursuing ambitious academic paths, thereby maintaining racial/ethnic educational inequality.