We study the long-run aggregate, distributional, and intergenerational effects of school tracking—the allocation of students to different types of schools—by incorporating school track decisions into a general-equilibrium heterogeneous-agent overlappinggenerations model. The key innovation in our model is the skill production technology during school years with tracking. School tracks endogenously differ in their pace of instruction and the students’ average skills. We show analytically that this technology can rationalize reduced-form evidence on the effects of school tracking on the distribution of end-of-school skills. We then calibrate the model using representative data from Germany, a country with a very early school tracking policy by international standards. Our calibrated model shows that an education reform that postpones the tracking age from ten to fourteen generates improvements in intergenerational mobility but comes at the cost of modest losses in aggregate human capital and economic output, reducing aggregate welfare. This efficiency-mobility trade-off is rooted in the effects of longer comprehensive schooling on learning and depends crucially on the presence of general equilibrium effects in the labor market. Finally, counterfactual analyses suggest that policies that reduce the parental influence in the school track choice can increase both social mobility and aggregate economic output, improving aggregate welfare.