We estimate the aggregate long-run elasticity of substitution between more and less educated workers (the slope of the demand curve for more relative to less educated workers) at the US state level. Our data come from the (five)1950-1990 decennial censuses. Our empirical approach allows for...