The empirical literature on the causal effects of class-size reduction on academic outcomes is crowded and yields mixed results, with particularly limited and inconsistent findings in the Japanese context. This study examines whether smaller class sizes enhance classroom climate and student-teacher relationships, using large-scale panel data from a student achievement survey conducted in a Japanese prefecture. Employing an instrumental variable approach based on the Maimonides rule, I find that a reduction of 10 students per class yields modest improvements—up to 0.07 standard deviations—in measures of teacher-student relationships. The analysis does not provide sufficient evidence to conclude that class-size reduction is effective in improving students’ well-being.