Bénabou [(2002) Econometrica 70, 481–517.] demonstrated how progressive redistribution mitigates unequal opportunity for education to promote growth. We ask how effective that policy is when it affects the average quality of education. So, we add a learning externality in his model of human capital accumulation. It captures how knowledge diffusion through global channels—such as a national educational curriculum—influences children’s learning outcomes and, hence, growth. Whether inequality is a curse or a blessing for economic growth depends on how efficiently an economy designs these channels and to what extent they influence children’s education. The greater this efficiency, the lower is the optimal progressiveness of income redistribution for maximizing the long-run output growth rate. Interestingly, if poorly designed global channels exert greater influence in education, the optimal progressivity may increase. Our quantitative analysis reveals new implications for redistributive policies, including a surprisingly low optimal progressivity (approximately 6–7%) for the calibrated US economy.