Though teacher expectations have been researched for decades, their significance and impact remain topics of debate. Across two studies, we use methodologies designed to address common criticisms of observational research on teacher expectations. In Study 1 (NStudents = 16,176), we use classmates’ perceptions of teacher expectations, while in Study 2, we leverage the random assignment of teachers to classrooms (NSections = 188), to examine how teacher expectations predict students’ math test score gains up to three years later. Employing one of the largest and most diverse datasets ever used in teacher expectations research, we compare the effect size of teacher expectations to key factors known to influence achievement (e.g., socioeconomic status, instructional quality), explore how students’ tendencies to devalue academics mediate these effects, and assess equity in teacher expectations across racial/ethnic groups. Our findings reveal that reflected potential—a measure of perceived teacher belief in students’ potential—predicts achievement gains with effect sizes comparable to those of widely recognized educational variables. We also find that this relationship is mediated by a reduced likelihood of students devaluing academics and that students of color receive relatively less reflected potential from their teachers. These results highlight the role of teacher expectations in fostering student achievement.