In an empirical analysis, considering 236 U.S. cities in the period 1980-1990, we document a strong positive correlation between local supply of skills and their return. In SMSA's where the average education of workers is high the education premium is also high. This is true both considering the levels of the variables in 1980, 1990 and considering their changes. Technical progress, as well as physical capital investments, may be driven by local pressures to enhance the productivity of those factors which are locally abundant. Therefore we may interpret this regularity as a sign that in cities where educated workers are abundant firms will invest in skill-complementary machines and techniques. Acemoglou (1998) claims that this idea could explain the time evolution of education premia in the 80's. Here I bring some compelling evidence that it may provide an explanation for the behaviour of education premia across cities.