In the human capital model, what affects the choice of how much education to acquire? Angrist and Krueger (Quarterly Journal of Economics, 1991) estimate the effects of compulsory schooling in the US using month of birth as an instrument for schooling. Why is a simple OLS regression unlikely to give us causal estimates in this context? Can you think of any possible violation of the key assumptions behind Angrist and Krueger (1991)?