Existing growth research provides little explanation for the very large differences in long-run growth performance across OECD countries. We show that cognitive skills can account for growth differences within the OECD, whereas a range of economic institutions and quantitative measures of tertiary education cannot. Under the growth model estimates and plausible projection parameters, school improvements falling within currently observed performance levels yield very large gains. The present value of OECD aggregate gains through 2090 could be as much as $275 trillion, or 13.8 percent of the discounted value of future GDP. Extensive sensitivity analyses indicate that, while differences between model frameworks and alternative parameter choices make a difference, the economic impact of improved educational outcomes remains enormous. Interestingly, the quantitative difference between an endogenous and neoclassical model framework - with improved skills affecting the long-run growth rate versus just the steady-state income level - matters less than academic discussions suggest. We close by discussing evidence on which education policy reforms may be able to bring about the simulated improvements in educational outcomes.