This study investigated (1) how students and their families' choices of learning modality differed by student demographics, (2) how much movement (i.e. learning modality changes) occurred during the pandemic school year, and (3) the mediating effect of movement on the relation between initial learning modality choices and students' mathematics performance. The participants involved 4008 7th-grade students from a suburban school district in the southeastern USA. Results of the chi-square tests revealed that initial learning modality choices and movement significantly differed by students' demographic variables. Most White and Hispanic students initially selected in-person learning, whereas most Asian and Black students selected virtual learning. About half of Black (55.6%) and Hispanic (46.7%) students who initially selected virtual learning moved back to in-person classes, while most Asian students remained in virtual learning. Further, we performed a mediation analysis using the subset of the students (n = 2046) who completed both pretest and posttest on algebraic knowledge. The results indicated that the initial choice of learning modality had a significant indirect effect on posttest scores of algebraic knowledge through its association with the change of learning modality, after controlling for students' demographic variables and pretest scores. Our findings suggest that instability in learning modality may be disruptive to student learning and is associated with lower end-of-year math performance.