Enrico Moretti examines the impact of workers' education on productivity using plant-level production functions. He uses a unique dataset combining the Census of Manufacturers and the Census of Population to analyze how increases in the share of college graduates in cities affect plant productivity. After controlling for plant-specific factors, industry-specific, and state-specific shocks, he finds that plants in cities with higher college shares experience greater productivity gains. These gains are most significant for high-tech plants, while low-tech plants show negligible spillovers. Specification tests indicate that the estimated effects are not spurious, as physical capital density and interactions between plants within the same city do not significantly influence productivity. The estimated productivity differences between cities with high and low human capital align with observed labor cost differences. Human capital spillovers generate productivity gains that are offset by increased labor costs. The paper presents a general equilibrium model showing that cities with higher human capital are more productive, and that spillovers are consistent with wage differences across cities. The results suggest that human capital spillovers contribute to economic growth, but their impact is relatively small. The study provides robust empirical evidence supporting the existence of human capital externalities in U.S. manufacturing.Enrico Moretti examines the impact of workers' education on productivity using plant-level production functions. He uses a unique dataset combining the Census of Manufacturers and the Census of Population to analyze how increases in the share of college graduates in cities affect plant productivity. After controlling for plant-specific factors, industry-specific, and state-specific shocks, he finds that plants in cities with higher college shares experience greater productivity gains. These gains are most significant for high-tech plants, while low-tech plants show negligible spillovers. Specification tests indicate that the estimated effects are not spurious, as physical capital density and interactions between plants within the same city do not significantly influence productivity. The estimated productivity differences between cities with high and low human capital align with observed labor cost differences. Human capital spillovers generate productivity gains that are offset by increased labor costs. The paper presents a general equilibrium model showing that cities with higher human capital are more productive, and that spillovers are consistent with wage differences across cities. The results suggest that human capital spillovers contribute to economic growth, but their impact is relatively small. The study provides robust empirical evidence supporting the existence of human capital externalities in U.S. manufacturing.