Multigrading represents the practice of mixing children of different ages in the same classroom. This paper examines the effect of attending a multigrade class in Grade 2 on students’ academic achievement in Grades 2, 5, and 8, respectively, considering Italy as a case study. To address the issue of endogeneity of multigrading (and class size), we adopt an IV identification strategy based on a law that disciplines class composition. We show that multigrading has a positive (16 percent of a standard deviation) short-term effect on academic achievements. However, this effect diminishes over time and becomes negative (-10 percent of a standard deviation) if students spend several years in a multigrade class. Mechanism analysis indicates the fundamental role of teachers and suggests that the negative long-term effect of multigrading is not statistically different from zero when multigrade classes are taught by more experienced teachers. These findings, based on longitudinal data, reconcile contrasting results in the literature, which are based on cross-sectional data and on the short-term effects of multigrading.