This paper evaluates the long-term effects of class size in primary school. We use rich administrative data from Sweden and exploit variation in class size created by a maximum class size rule. Smaller classes in the last three years of primary school (age 10 to 13) are not only beneficial for cognitive test scores at age 13 but also for non-cognitive scores at that age, for cognitive test scores at ages 16 and 18, and for completed education and wages at age 27 to 42. The estimated effect on wages is much larger than any indirect (imputed) estimate of the wage effect, and is large enough to pass a cost-benefit test.