Per-capita income gaps across US states and Canadian provinces

We propose that per-capita income gaps across US states and Canadian provinces can be explained by university education. Our ordinary least-squares regressions show university education having a robust positive and significant effect on per-capita incomes, when controlling for, e.g. taxes, unionization, government spending, and the sectoral composition of the economy. To control for endogeneity we instrument education today with different historical variables: population density in 1900, railway density in 1900, the sex ratio (men per woman) in 1900, and the fraction slaves in 1850. Our results support a causal link from education to incomes, and the hypothesis that these historical variables are valid instruments. We also find that the Canada dummy is mostly insignificant, and always positive.